From 099c502bdf02ed9bc34bbfc70a6e786746ecee90 Mon Sep 17 00:00:00 2001 From: Kevin Sala Date: Fri, 8 Aug 2025 10:43:52 -0700 Subject: [PATCH 01/17] [OpenMP] Add parser/semantic support for dyn_groupprivate clause --- clang/include/clang/AST/OpenMPClause.h | 155 ++++++++++++++++++ clang/include/clang/AST/RecursiveASTVisitor.h | 8 + .../clang/Basic/DiagnosticSemaKinds.td | 3 + clang/include/clang/Basic/OpenMPKinds.def | 9 + clang/include/clang/Basic/OpenMPKinds.h | 10 ++ clang/include/clang/Sema/SemaOpenMP.h | 7 + clang/lib/AST/OpenMPClause.cpp | 21 +++ clang/lib/AST/StmtProfile.cpp | 6 + clang/lib/Basic/OpenMPKinds.cpp | 17 ++ clang/lib/Parse/ParseOpenMP.cpp | 47 +++++- clang/lib/Sema/SemaOpenMP.cpp | 73 ++++++++- clang/lib/Sema/TreeTransform.h | 26 +++ clang/lib/Serialization/ASTReader.cpp | 16 ++ clang/lib/Serialization/ASTWriter.cpp | 11 ++ .../target_dyn_groupprivate_messages.cpp | 87 ++++++++++ ...target_teams_dyn_groupprivate_messages.cpp | 87 ++++++++++ clang/tools/libclang/CIndex.cpp | 5 + llvm/include/llvm/Frontend/OpenMP/OMP.td | 25 +++ 18 files changed, 605 insertions(+), 8 deletions(-) create mode 100644 clang/test/OpenMP/target_dyn_groupprivate_messages.cpp create mode 100644 clang/test/OpenMP/target_teams_dyn_groupprivate_messages.cpp diff --git a/clang/include/clang/AST/OpenMPClause.h b/clang/include/clang/AST/OpenMPClause.h index 1118d3e062e68..a3983120df069 100644 --- a/clang/include/clang/AST/OpenMPClause.h +++ b/clang/include/clang/AST/OpenMPClause.h @@ -9768,6 +9768,161 @@ class OMPXDynCGroupMemClause Expr *getSize() const { return getStmtAs(); } }; +/// This represents 'dyn_groupprivate' clause in '#pragma omp target ...' +/// and '#pragma omp teams ...' directives. +/// +/// \code +/// #pragma omp target [...] dyn_groupprivate(a,b: N) +/// \endcode +class OMPDynGroupprivateClause : public OMPClause, public OMPClauseWithPreInit { + friend class OMPClauseReader; + + /// Location of '('. + SourceLocation LParenLoc; + + /// Modifiers for 'dyn_groupprivate' clause. + enum { FIRST, SECOND, NUM_MODIFIERS }; + OpenMPDynGroupprivateClauseModifier Modifiers[NUM_MODIFIERS]; + + /// Locations of modifiers. + SourceLocation ModifiersLoc[NUM_MODIFIERS]; + + /// The size of the dyn_groupprivate. + Expr *Size = nullptr; + + /// Set the first dyn_groupprivate modifier. + /// + /// \param M The modifier. + void setFirstDynGroupprivateModifier(OpenMPDynGroupprivateClauseModifier M) { + Modifiers[FIRST] = M; + } + + /// Set the second dyn_groupprivate modifier. + /// + /// \param M The modifier. + void setSecondDynGroupprivateModifier(OpenMPDynGroupprivateClauseModifier M) { + Modifiers[SECOND] = M; + } + + /// Set location of the first dyn_groupprivate modifier. + void setFirstDynGroupprivateModifierLoc(SourceLocation Loc) { + ModifiersLoc[FIRST] = Loc; + } + + /// Set location of the second dyn_groupprivate modifier. + void setSecondDynGroupprivateModifierLoc(SourceLocation Loc) { + ModifiersLoc[SECOND] = Loc; + } + + /// Set dyn_groupprivate modifier location. + /// + /// \param M The modifier location. + void setDynGroupprivateModifer(OpenMPDynGroupprivateClauseModifier M) { + if (Modifiers[FIRST] == OMPC_DYN_GROUPPRIVATE_unknown) + Modifiers[FIRST] = M; + else { + assert(Modifiers[SECOND] == OMPC_DYN_GROUPPRIVATE_unknown); + Modifiers[SECOND] = M; + } + } + + /// Sets the location of '('. + /// + /// \param Loc Location of '('. + void setLParenLoc(SourceLocation Loc) { LParenLoc = Loc; } + + /// Set size. + /// + /// \param E Size. + void setSize(Expr *E) { Size = E; } + +public: + /// Build 'dyn_groupprivate' clause with a size expression \a Size. + /// + /// \param StartLoc Starting location of the clause. + /// \param LParenLoc Location of '('. + /// \param EndLoc Ending location of the clause. + /// \param Size Size. + /// \param M1 The first modifier applied to 'dyn_groupprivate' clause. + /// \param M1Loc Location of the first modifier. + /// \param M2 The second modifier applied to 'dyn_groupprivate' clause. + /// \param M2Loc Location of the second modifier. + OMPDynGroupprivateClause(SourceLocation StartLoc, SourceLocation LParenLoc, + SourceLocation EndLoc, Expr *Size, Stmt *HelperSize, + OpenMPDirectiveKind CaptureRegion, + OpenMPDynGroupprivateClauseModifier M1, + SourceLocation M1Loc, + OpenMPDynGroupprivateClauseModifier M2, + SourceLocation M2Loc) + : OMPClause(llvm::omp::OMPC_dyn_groupprivate, StartLoc, EndLoc), + OMPClauseWithPreInit(this), LParenLoc(LParenLoc), Size(Size) { + setPreInitStmt(HelperSize, CaptureRegion); + Modifiers[FIRST] = M1; + Modifiers[SECOND] = M2; + ModifiersLoc[FIRST] = M1Loc; + ModifiersLoc[SECOND] = M2Loc; + } + + /// Build an empty clause. + explicit OMPDynGroupprivateClause() + : OMPClause(llvm::omp::OMPC_dyn_groupprivate, SourceLocation(), + SourceLocation()), + OMPClauseWithPreInit(this) { + Modifiers[FIRST] = OMPC_DYN_GROUPPRIVATE_unknown; + Modifiers[SECOND] = OMPC_DYN_GROUPPRIVATE_unknown; + } + + /// Get the first modifier of the clause. + OpenMPDynGroupprivateClauseModifier getFirstDynGroupprivateModifier() const { + return Modifiers[FIRST]; + } + + /// Get the second modifier of the clause. + OpenMPDynGroupprivateClauseModifier getSecondDynGroupprivateModifier() const { + return Modifiers[SECOND]; + } + + /// Get location of '('. + SourceLocation getLParenLoc() { return LParenLoc; } + + /// Get the first modifier location. + SourceLocation getFirstDynGroupprivateModifierLoc() const { + return ModifiersLoc[FIRST]; + } + + /// Get the second modifier location. + SourceLocation getSecondDynGroupprivateModifierLoc() const { + return ModifiersLoc[SECOND]; + } + + /// Get size. + Expr *getSize() { return Size; } + + /// Get size. + const Expr *getSize() const { return Size; } + + child_range children() { + return child_range(reinterpret_cast(&Size), + reinterpret_cast(&Size) + 1); + } + + const_child_range children() const { + auto Children = const_cast(this)->children(); + return const_child_range(Children.begin(), Children.end()); + } + + child_range used_children() { + return child_range(child_iterator(), child_iterator()); + } + const_child_range used_children() const { + return const_child_range(const_child_iterator(), const_child_iterator()); + } + + static bool classof(const OMPClause *T) { + return T->getClauseKind() == llvm::omp::OMPC_dyn_groupprivate; + } +}; + /// This represents the 'doacross' clause for the '#pragma omp ordered' /// directive. /// diff --git a/clang/include/clang/AST/RecursiveASTVisitor.h b/clang/include/clang/AST/RecursiveASTVisitor.h index 5cb2f57edffe4..129115d56fe82 100644 --- a/clang/include/clang/AST/RecursiveASTVisitor.h +++ b/clang/include/clang/AST/RecursiveASTVisitor.h @@ -4060,6 +4060,14 @@ bool RecursiveASTVisitor::VisitOMPXDynCGroupMemClause( return true; } +template +bool RecursiveASTVisitor::VisitOMPDynGroupprivateClause( + OMPDynGroupprivateClause *C) { + TRY_TO(VisitOMPClauseWithPreInit(C)); + TRY_TO(TraverseStmt(C->getSize())); + return true; +} + template bool RecursiveASTVisitor::VisitOMPDoacrossClause( OMPDoacrossClause *C) { diff --git a/clang/include/clang/Basic/DiagnosticSemaKinds.td b/clang/include/clang/Basic/DiagnosticSemaKinds.td index b285309e0b3ca..edfa2d229789a 100644 --- a/clang/include/clang/Basic/DiagnosticSemaKinds.td +++ b/clang/include/clang/Basic/DiagnosticSemaKinds.td @@ -11995,6 +11995,9 @@ def err_omp_unexpected_schedule_modifier : Error< "modifier '%0' cannot be used along with modifier '%1'">; def err_omp_schedule_nonmonotonic_static : Error< "'nonmonotonic' modifier can only be specified with 'dynamic' or 'guided' schedule kind">; +def err_omp_unexpected_dyn_groupprivate_modifier + : Error<"modifier '%0' cannot be used along with modifier '%1' in " + "dyn_groupprivate">; def err_omp_simple_clause_incompatible_with_ordered : Error< "'%0' clause with '%1' modifier cannot be specified if an 'ordered' clause is specified">; def err_omp_ordered_simd : Error< diff --git a/clang/include/clang/Basic/OpenMPKinds.def b/clang/include/clang/Basic/OpenMPKinds.def index 9d6f816eea91f..3321e19cae9b1 100644 --- a/clang/include/clang/Basic/OpenMPKinds.def +++ b/clang/include/clang/Basic/OpenMPKinds.def @@ -83,6 +83,9 @@ #ifndef OPENMP_GRAINSIZE_MODIFIER #define OPENMP_GRAINSIZE_MODIFIER(Name) #endif +#ifndef OPENMP_DYN_GROUPPRIVATE_MODIFIER +#define OPENMP_DYN_GROUPPRIVATE_MODIFIER(Name) +#endif #ifndef OPENMP_NUMTASKS_MODIFIER #define OPENMP_NUMTASKS_MODIFIER(Name) #endif @@ -227,6 +230,11 @@ OPENMP_BIND_KIND(thread) // Modifiers for the 'grainsize' clause. OPENMP_GRAINSIZE_MODIFIER(strict) +// Modifiers for the 'dyn_groupprivate' clause. +OPENMP_DYN_GROUPPRIVATE_MODIFIER(cgroup) +OPENMP_DYN_GROUPPRIVATE_MODIFIER(strict) +OPENMP_DYN_GROUPPRIVATE_MODIFIER(fallback) + // Modifiers for the 'num_tasks' clause. OPENMP_NUMTASKS_MODIFIER(strict) @@ -245,6 +253,7 @@ OPENMP_DOACROSS_MODIFIER(source_omp_cur_iteration) #undef OPENMP_NUMTASKS_MODIFIER #undef OPENMP_NUMTHREADS_MODIFIER +#undef OPENMP_DYN_GROUPPRIVATE_MODIFIER #undef OPENMP_GRAINSIZE_MODIFIER #undef OPENMP_BIND_KIND #undef OPENMP_ADJUST_ARGS_KIND diff --git a/clang/include/clang/Basic/OpenMPKinds.h b/clang/include/clang/Basic/OpenMPKinds.h index f40db4c13c55a..3e164bf1adf22 100644 --- a/clang/include/clang/Basic/OpenMPKinds.h +++ b/clang/include/clang/Basic/OpenMPKinds.h @@ -217,6 +217,16 @@ enum OpenMPGrainsizeClauseModifier { OMPC_GRAINSIZE_unknown }; +enum OpenMPDynGroupprivateClauseModifier { +#define OPENMP_DYN_GROUPPRIVATE_MODIFIER(Name) OMPC_DYN_GROUPPRIVATE_##Name, +#include "clang/Basic/OpenMPKinds.def" + OMPC_DYN_GROUPPRIVATE_unknown +}; + +/// Number of allowed dyn_groupprivate-modifiers. +static constexpr unsigned NumberOfOMPDynGroupprivateClauseModifiers = + OMPC_DYN_GROUPPRIVATE_unknown; + enum OpenMPNumTasksClauseModifier { #define OPENMP_NUMTASKS_MODIFIER(Name) OMPC_NUMTASKS_##Name, #include "clang/Basic/OpenMPKinds.def" diff --git a/clang/include/clang/Sema/SemaOpenMP.h b/clang/include/clang/Sema/SemaOpenMP.h index 91c3d4bd5210e..3b161ff3c7d45 100644 --- a/clang/include/clang/Sema/SemaOpenMP.h +++ b/clang/include/clang/Sema/SemaOpenMP.h @@ -1385,6 +1385,13 @@ class SemaOpenMP : public SemaBase { SourceLocation LParenLoc, SourceLocation EndLoc); + /// Called on a well-formed 'dyn_groupprivate' clause. + OMPClause *ActOnOpenMPDynGroupprivateClause( + OpenMPDynGroupprivateClauseModifier M1, + OpenMPDynGroupprivateClauseModifier M2, Expr *Size, + SourceLocation StartLoc, SourceLocation LParenLoc, SourceLocation M1Loc, + SourceLocation M2Loc, SourceLocation EndLoc); + /// Called on well-formed 'doacross' clause. OMPClause * ActOnOpenMPDoacrossClause(OpenMPDoacrossClauseModifier DepType, diff --git a/clang/lib/AST/OpenMPClause.cpp b/clang/lib/AST/OpenMPClause.cpp index de8b5996818de..a6e79f1e2230b 100644 --- a/clang/lib/AST/OpenMPClause.cpp +++ b/clang/lib/AST/OpenMPClause.cpp @@ -104,6 +104,8 @@ const OMPClauseWithPreInit *OMPClauseWithPreInit::get(const OMPClause *C) { return static_cast(C); case OMPC_ompx_dyn_cgroup_mem: return static_cast(C); + case OMPC_dyn_groupprivate: + return static_cast(C); case OMPC_default: case OMPC_proc_bind: case OMPC_safelen: @@ -2725,6 +2727,25 @@ void OMPClausePrinter::VisitOMPXDynCGroupMemClause( OS << ")"; } +void OMPClausePrinter::VisitOMPDynGroupprivateClause( + OMPDynGroupprivateClause *Node) { + OS << "dyn_groupprivate("; + if (Node->getFirstDynGroupprivateModifier() != + OMPC_DYN_GROUPPRIVATE_unknown) { + OS << getOpenMPSimpleClauseTypeName( + OMPC_dyn_groupprivate, Node->getFirstDynGroupprivateModifier()); + if (Node->getSecondDynGroupprivateModifier() != + OMPC_DYN_GROUPPRIVATE_unknown) { + OS << ", "; + OS << getOpenMPSimpleClauseTypeName( + OMPC_dyn_groupprivate, Node->getSecondDynGroupprivateModifier()); + } + OS << ": "; + } + Node->getSize()->printPretty(OS, nullptr, Policy, 0); + OS << ')'; +} + void OMPClausePrinter::VisitOMPDoacrossClause(OMPDoacrossClause *Node) { OS << "doacross("; OpenMPDoacrossClauseModifier DepType = Node->getDependenceType(); diff --git a/clang/lib/AST/StmtProfile.cpp b/clang/lib/AST/StmtProfile.cpp index c61450e19f1b6..6b1a016649547 100644 --- a/clang/lib/AST/StmtProfile.cpp +++ b/clang/lib/AST/StmtProfile.cpp @@ -957,6 +957,12 @@ void OMPClauseProfiler::VisitOMPXDynCGroupMemClause( if (Expr *Size = C->getSize()) Profiler->VisitStmt(Size); } +void OMPClauseProfiler::VisitOMPDynGroupprivateClause( + const OMPDynGroupprivateClause *C) { + VistOMPClauseWithPreInit(C); + if (auto *Size = C->getSize()) + Profiler->VisitStmt(Size); +} void OMPClauseProfiler::VisitOMPDoacrossClause(const OMPDoacrossClause *C) { VisitOMPClauseList(C); } diff --git a/clang/lib/Basic/OpenMPKinds.cpp b/clang/lib/Basic/OpenMPKinds.cpp index d3d393bd09396..5a2b578cfb33e 100644 --- a/clang/lib/Basic/OpenMPKinds.cpp +++ b/clang/lib/Basic/OpenMPKinds.cpp @@ -171,6 +171,13 @@ unsigned clang::getOpenMPSimpleClauseType(OpenMPClauseKind Kind, StringRef Str, return OMPC_GRAINSIZE_unknown; return Type; } + case OMPC_dyn_groupprivate: { + return llvm::StringSwitch(Str) +#define OPENMP_DYN_GROUPPRIVATE_MODIFIER(Name) \ + .Case(#Name, OMPC_DYN_GROUPPRIVATE_##Name) +#include "clang/Basic/OpenMPKinds.def" + .Default(OMPC_DYN_GROUPPRIVATE_unknown); + } case OMPC_num_tasks: { unsigned Type = llvm::StringSwitch(Str) #define OPENMP_NUMTASKS_MODIFIER(Name) .Case(#Name, OMPC_NUMTASKS_##Name) @@ -508,6 +515,16 @@ const char *clang::getOpenMPSimpleClauseTypeName(OpenMPClauseKind Kind, #include "clang/Basic/OpenMPKinds.def" } llvm_unreachable("Invalid OpenMP 'grainsize' clause modifier"); + case OMPC_dyn_groupprivate: + switch (Type) { + case OMPC_DYN_GROUPPRIVATE_unknown: + return "unknown"; +#define OPENMP_DYN_GROUPPRIVATE_MODIFIER(Name) \ + case OMPC_DYN_GROUPPRIVATE_##Name: \ + return #Name; +#include "clang/Basic/OpenMPKinds.def" + } + llvm_unreachable("Invalid OpenMP 'dyn_groupprivate' clause modifier"); case OMPC_num_tasks: switch (Type) { case OMPC_NUMTASKS_unknown: diff --git a/clang/lib/Parse/ParseOpenMP.cpp b/clang/lib/Parse/ParseOpenMP.cpp index aa6a0c61a2c17..fe3c20b765e52 100644 --- a/clang/lib/Parse/ParseOpenMP.cpp +++ b/clang/lib/Parse/ParseOpenMP.cpp @@ -3039,6 +3039,7 @@ OMPClause *Parser::ParseOpenMPClause(OpenMPDirectiveKind DKind, case OMPC_align: case OMPC_message: case OMPC_ompx_dyn_cgroup_mem: + case OMPC_dyn_groupprivate: // OpenMP [2.5, Restrictions] // At most one num_threads clause can appear on the directive. // OpenMP [2.8.1, simd construct, Restrictions] @@ -3077,7 +3078,7 @@ OMPClause *Parser::ParseOpenMPClause(OpenMPDirectiveKind DKind, PP.LookAhead(/*N=*/0).isNot(tok::l_paren)) Clause = ParseOpenMPClause(CKind, WrongDirective); else if (CKind == OMPC_grainsize || CKind == OMPC_num_tasks || - CKind == OMPC_num_threads) + CKind == OMPC_num_threads || CKind == OMPC_dyn_groupprivate) Clause = ParseOpenMPSingleExprWithArgClause(DKind, CKind, WrongDirective); else Clause = ParseOpenMPSingleExprClause(CKind, WrongDirective); @@ -3835,6 +3836,40 @@ OMPClause *Parser::ParseOpenMPSingleExprWithArgClause(OpenMPDirectiveKind DKind, Arg.push_back(OMPC_GRAINSIZE_unknown); KLoc.emplace_back(); } + } else if (Kind == OMPC_dyn_groupprivate) { + enum { Modifier1, Modifier2, NumberOfElements }; + Arg.resize(NumberOfElements); + KLoc.resize(NumberOfElements); + Arg[Modifier1] = OMPC_DYN_GROUPPRIVATE_unknown; + Arg[Modifier2] = OMPC_DYN_GROUPPRIVATE_unknown; + unsigned Modifier = getOpenMPSimpleClauseType( + Kind, Tok.isAnnotation() ? "" : PP.getSpelling(Tok), getLangOpts()); + + if (Modifier < OMPC_DYN_GROUPPRIVATE_unknown) { + // Parse 'modifier' + Arg[Modifier1] = Modifier; + KLoc[Modifier1] = Tok.getLocation(); + if (Tok.isNot(tok::r_paren) && Tok.isNot(tok::comma) && + Tok.isNot(tok::annot_pragma_openmp_end)) + ConsumeAnyToken(); + if (Tok.is(tok::comma)) { + // Parse ',' 'modifier' + ConsumeAnyToken(); + Modifier = getOpenMPSimpleClauseType( + Kind, Tok.isAnnotation() ? "" : PP.getSpelling(Tok), getLangOpts()); + Arg[Modifier2] = Modifier; + KLoc[Modifier2] = Tok.getLocation(); + if (Tok.isNot(tok::r_paren) && Tok.isNot(tok::comma) && + Tok.isNot(tok::annot_pragma_openmp_end)) + ConsumeAnyToken(); + } + // Parse ':' + if (Tok.is(tok::colon)) + ConsumeAnyToken(); + else + Diag(Tok, diag::warn_pragma_expected_colon) + << "dyn_groupprivate modifier"; + } } else if (Kind == OMPC_num_tasks) { // Parse optional ':' OpenMPNumTasksClauseModifier Modifier = @@ -3909,11 +3944,11 @@ OMPClause *Parser::ParseOpenMPSingleExprWithArgClause(OpenMPDirectiveKind DKind, } } - bool NeedAnExpression = (Kind == OMPC_schedule && DelimLoc.isValid()) || - (Kind == OMPC_dist_schedule && DelimLoc.isValid()) || - Kind == OMPC_if || Kind == OMPC_device || - Kind == OMPC_grainsize || Kind == OMPC_num_tasks || - Kind == OMPC_num_threads; + bool NeedAnExpression = + (Kind == OMPC_schedule && DelimLoc.isValid()) || + (Kind == OMPC_dist_schedule && DelimLoc.isValid()) || Kind == OMPC_if || + Kind == OMPC_device || Kind == OMPC_grainsize || Kind == OMPC_num_tasks || + Kind == OMPC_num_threads || Kind == OMPC_dyn_groupprivate; if (NeedAnExpression) { SourceLocation ELoc = Tok.getLocation(); ExprResult LHS( diff --git a/clang/lib/Sema/SemaOpenMP.cpp b/clang/lib/Sema/SemaOpenMP.cpp index 4ecc9b0d4c5c8..c63310757513e 100644 --- a/clang/lib/Sema/SemaOpenMP.cpp +++ b/clang/lib/Sema/SemaOpenMP.cpp @@ -15560,6 +15560,7 @@ OMPClause *SemaOpenMP::ActOnOpenMPSingleExprClause(OpenMPClauseKind Kind, case OMPC_holds: Res = ActOnOpenMPHoldsClause(Expr, StartLoc, LParenLoc, EndLoc); break; + case OMPC_dyn_groupprivate: case OMPC_grainsize: case OMPC_num_tasks: case OMPC_num_threads: @@ -15686,6 +15687,8 @@ static OpenMPDirectiveKind getOpenMPCaptureRegionForClause( case OMPC_num_teams: case OMPC_thread_limit: case OMPC_ompx_dyn_cgroup_mem: + case OMPC_dyn_groupprivate: + // TODO: This may need to consider teams too. if (Leafs[0] == OMPD_target) return OMPD_target; break; @@ -16646,7 +16649,7 @@ OMPClause *SemaOpenMP::ActOnOpenMPSingleExprWithArgClause( SourceLocation EndLoc) { OMPClause *Res = nullptr; switch (Kind) { - case OMPC_schedule: + case OMPC_schedule: { enum { Modifier1, Modifier2, ScheduleKind, NumberOfElements }; assert(Argument.size() == NumberOfElements && ArgumentLoc.size() == NumberOfElements); @@ -16656,7 +16659,7 @@ OMPClause *SemaOpenMP::ActOnOpenMPSingleExprWithArgClause( static_cast(Argument[ScheduleKind]), Expr, StartLoc, LParenLoc, ArgumentLoc[Modifier1], ArgumentLoc[Modifier2], ArgumentLoc[ScheduleKind], DelimLoc, EndLoc); - break; + } break; case OMPC_if: assert(Argument.size() == 1 && ArgumentLoc.size() == 1); Res = ActOnOpenMPIfClause(static_cast(Argument.back()), @@ -16703,6 +16706,16 @@ OMPClause *SemaOpenMP::ActOnOpenMPSingleExprWithArgClause( static_cast(Argument.back()), Expr, StartLoc, LParenLoc, ArgumentLoc.back(), EndLoc); break; + case OMPC_dyn_groupprivate: { + enum { Modifier1, Modifier2, NumberOfElements }; + assert(Argument.size() == NumberOfElements && + ArgumentLoc.size() == NumberOfElements); + Res = ActOnOpenMPDynGroupprivateClause( + static_cast(Argument[Modifier1]), + static_cast(Argument[Modifier2]), + Expr, StartLoc, LParenLoc, ArgumentLoc[Modifier1], + ArgumentLoc[Modifier2], EndLoc); + } break; case OMPC_num_threads: assert(Argument.size() == 1 && ArgumentLoc.size() == 1 && "Modifier for num_threads clause and its location are expected."); @@ -17056,6 +17069,7 @@ OMPClause *SemaOpenMP::ActOnOpenMPClause(OpenMPClauseKind Kind, case OMPC_affinity: case OMPC_when: case OMPC_ompx_dyn_cgroup_mem: + case OMPC_dyn_groupprivate: default: llvm_unreachable("Clause is not allowed."); } @@ -24143,6 +24157,61 @@ OMPClause *SemaOpenMP::ActOnOpenMPXDynCGroupMemClause(Expr *Size, ValExpr, HelperValStmt, CaptureRegion, StartLoc, LParenLoc, EndLoc); } +OMPClause *SemaOpenMP::ActOnOpenMPDynGroupprivateClause( + OpenMPDynGroupprivateClauseModifier M1, + OpenMPDynGroupprivateClauseModifier M2, Expr *Size, SourceLocation StartLoc, + SourceLocation LParenLoc, SourceLocation M1Loc, SourceLocation M2Loc, + SourceLocation EndLoc) { + + if ((M1Loc.isValid() && M1 == OMPC_DYN_GROUPPRIVATE_unknown) || + (M2Loc.isValid() && M2 == OMPC_DYN_GROUPPRIVATE_unknown)) { + std::string Values = getListOfPossibleValues( + OMPC_dyn_groupprivate, /*First=*/0, OMPC_DYN_GROUPPRIVATE_unknown); + Diag((M1Loc.isValid() && M1 == OMPC_DYN_GROUPPRIVATE_unknown) ? M1Loc + : M2Loc, + diag::err_omp_unexpected_clause_value) + << Values << getOpenMPClauseName(OMPC_dyn_groupprivate); + return nullptr; + } + + if ((M1Loc.isValid() && M2Loc.isValid() && M1 == M2) || + (M1 == OMPC_DYN_GROUPPRIVATE_strict && + M2 == OMPC_DYN_GROUPPRIVATE_fallback) || + (M1 == OMPC_DYN_GROUPPRIVATE_fallback && + M2 == OMPC_DYN_GROUPPRIVATE_strict)) { + + Diag(M2Loc, diag::err_omp_unexpected_dyn_groupprivate_modifier) + << getOpenMPSimpleClauseTypeName(OMPC_dyn_groupprivate, M2) + << getOpenMPSimpleClauseTypeName(OMPC_dyn_groupprivate, M1); + return nullptr; + } + + Expr *ValExpr = Size; + Stmt *HelperValStmt = nullptr; + + // OpenMP [2.5, Restrictions] + // The dyn_groupprivate expression must evaluate to a positive integer + // value. + if (!isNonNegativeIntegerValue(ValExpr, SemaRef, OMPC_dyn_groupprivate, + /*StrictlyPositive=*/false)) + return nullptr; + + OpenMPDirectiveKind DKind = DSAStack->getCurrentDirective(); + OpenMPDirectiveKind CaptureRegion = getOpenMPCaptureRegionForClause( + DKind, OMPC_dyn_groupprivate, getLangOpts().OpenMP); + if (CaptureRegion != OMPD_unknown && + !SemaRef.CurContext->isDependentContext()) { + ValExpr = SemaRef.MakeFullExpr(ValExpr).get(); + llvm::MapVector Captures; + ValExpr = tryBuildCapture(SemaRef, ValExpr, Captures).get(); + HelperValStmt = buildPreInits(getASTContext(), Captures); + } + + return new (getASTContext()) OMPDynGroupprivateClause( + StartLoc, LParenLoc, EndLoc, ValExpr, HelperValStmt, CaptureRegion, M1, + M1Loc, M2, M2Loc); +} + OMPClause *SemaOpenMP::ActOnOpenMPDoacrossClause( OpenMPDoacrossClauseModifier DepType, SourceLocation DepLoc, SourceLocation ColonLoc, ArrayRef VarList, SourceLocation StartLoc, diff --git a/clang/lib/Sema/TreeTransform.h b/clang/lib/Sema/TreeTransform.h index 758012f894a41..6222d63374824 100644 --- a/clang/lib/Sema/TreeTransform.h +++ b/clang/lib/Sema/TreeTransform.h @@ -2430,6 +2430,19 @@ class TreeTransform { LParenLoc, EndLoc); } + /// Build a new OpenMP 'dyn_groupprivate' clause. + /// + /// By default, performs semantic analysis to build the new OpenMP clause. + /// Subclasses may override this routine to provide different behavior. + OMPClause *RebuildOMPDynGroupprivateClause( + OpenMPDynGroupprivateClauseModifier M1, + OpenMPDynGroupprivateClauseModifier M2, Expr *Size, + SourceLocation StartLoc, SourceLocation LParenLoc, SourceLocation M1Loc, + SourceLocation M2Loc, SourceLocation EndLoc) { + return getSema().OpenMP().ActOnOpenMPDynGroupprivateClause( + M1, M2, Size, StartLoc, LParenLoc, M1Loc, M2Loc, EndLoc); + } + /// Build a new OpenMP 'ompx_attribute' clause. /// /// By default, performs semantic analysis to build the new OpenMP clause. @@ -11691,6 +11704,19 @@ OMPClause *TreeTransform::TransformOMPXDynCGroupMemClause( Size.get(), C->getBeginLoc(), C->getLParenLoc(), C->getEndLoc()); } +template +OMPClause *TreeTransform::TransformOMPDynGroupprivateClause( + OMPDynGroupprivateClause *C) { + ExprResult Size = getDerived().TransformExpr(C->getSize()); + if (Size.isInvalid()) + return nullptr; + return getDerived().RebuildOMPDynGroupprivateClause( + C->getFirstDynGroupprivateModifier(), + C->getSecondDynGroupprivateModifier(), Size.get(), C->getBeginLoc(), + C->getLParenLoc(), C->getFirstDynGroupprivateModifierLoc(), + C->getSecondDynGroupprivateModifierLoc(), C->getEndLoc()); +} + template OMPClause * TreeTransform::TransformOMPDoacrossClause(OMPDoacrossClause *C) { diff --git a/clang/lib/Serialization/ASTReader.cpp b/clang/lib/Serialization/ASTReader.cpp index 30e0973149594..84a30fcfcea59 100644 --- a/clang/lib/Serialization/ASTReader.cpp +++ b/clang/lib/Serialization/ASTReader.cpp @@ -11417,6 +11417,9 @@ OMPClause *OMPClauseReader::readClause() { case llvm::omp::OMPC_ompx_dyn_cgroup_mem: C = new (Context) OMPXDynCGroupMemClause(); break; + case llvm::omp::OMPC_dyn_groupprivate: + C = new (Context) OMPDynGroupprivateClause(); + break; case llvm::omp::OMPC_doacross: { unsigned NumVars = Record.readInt(); unsigned NumLoops = Record.readInt(); @@ -12583,6 +12586,19 @@ void OMPClauseReader::VisitOMPXDynCGroupMemClause(OMPXDynCGroupMemClause *C) { C->setLParenLoc(Record.readSourceLocation()); } +void OMPClauseReader::VisitOMPDynGroupprivateClause( + OMPDynGroupprivateClause *C) { + VisitOMPClauseWithPreInit(C); + C->setFirstDynGroupprivateModifier( + static_cast(Record.readInt())); + C->setSecondDynGroupprivateModifier( + static_cast(Record.readInt())); + C->setSize(Record.readSubExpr()); + C->setLParenLoc(Record.readSourceLocation()); + C->setFirstDynGroupprivateModifierLoc(Record.readSourceLocation()); + C->setSecondDynGroupprivateModifierLoc(Record.readSourceLocation()); +} + void OMPClauseReader::VisitOMPDoacrossClause(OMPDoacrossClause *C) { C->setLParenLoc(Record.readSourceLocation()); C->setDependenceType( diff --git a/clang/lib/Serialization/ASTWriter.cpp b/clang/lib/Serialization/ASTWriter.cpp index 1a20fc9595dce..42d1ab91a6879 100644 --- a/clang/lib/Serialization/ASTWriter.cpp +++ b/clang/lib/Serialization/ASTWriter.cpp @@ -8586,6 +8586,17 @@ void OMPClauseWriter::VisitOMPXDynCGroupMemClause(OMPXDynCGroupMemClause *C) { Record.AddSourceLocation(C->getLParenLoc()); } +void OMPClauseWriter::VisitOMPDynGroupprivateClause( + OMPDynGroupprivateClause *C) { + VisitOMPClauseWithPreInit(C); + Record.push_back(C->getFirstDynGroupprivateModifier()); + Record.push_back(C->getSecondDynGroupprivateModifier()); + Record.AddStmt(C->getSize()); + Record.AddSourceLocation(C->getLParenLoc()); + Record.AddSourceLocation(C->getFirstDynGroupprivateModifierLoc()); + Record.AddSourceLocation(C->getSecondDynGroupprivateModifierLoc()); +} + void OMPClauseWriter::VisitOMPDoacrossClause(OMPDoacrossClause *C) { Record.push_back(C->varlist_size()); Record.push_back(C->getNumLoops()); diff --git a/clang/test/OpenMP/target_dyn_groupprivate_messages.cpp b/clang/test/OpenMP/target_dyn_groupprivate_messages.cpp new file mode 100644 index 0000000000000..d5d855ee33e1f --- /dev/null +++ b/clang/test/OpenMP/target_dyn_groupprivate_messages.cpp @@ -0,0 +1,87 @@ +// RUN: %clang_cc1 -verify -fopenmp -fopenmp-version=45 %s -Wuninitialized +// RUN: %clang_cc1 -verify -fopenmp-simd -fopenmp-version=45 %s -Wuninitialized +// RUN: %clang_cc1 -verify -fopenmp %s -Wuninitialized +// RUN: %clang_cc1 -verify -fopenmp-simd %s -Wuninitialized + +void foo() { +} + +bool foobool(int argc) { + return argc; +} + +struct S1; // expected-note {{declared here}} + +template // expected-note {{declared here}} +int tmain(T argc, S **argv) { + T z; + #pragma omp target dyn_groupprivate // expected-error {{expected '(' after 'dyn_groupprivate'}} + foo(); + #pragma omp target dyn_groupprivate ( // expected-error {{expected expression}} expected-error {{expected ')'}} expected-note {{to match this '('}} + foo(); + #pragma omp target dyn_groupprivate () // expected-error {{expected expression}} + foo(); + #pragma omp target dyn_groupprivate (argc // expected-error {{expected ')'}} expected-note {{to match this '('}} + foo(); + #pragma omp target dyn_groupprivate (argc)) // expected-warning {{extra tokens at the end of '#pragma omp target' are ignored}} + foo(); + #pragma omp target dyn_groupprivate (argc > 0 ? argv[1] : argv[2]) // expected-error {{expression must have integral or unscoped enumeration type, not 'char *'}} + foo(); + #pragma omp target dyn_groupprivate (foobool(argc)), dyn_groupprivate (true) // expected-error {{directive '#pragma omp target' cannot contain more than one 'dyn_groupprivate' clause}} + foo(); + #pragma omp target dyn_groupprivate (S) // expected-error {{'S' does not refer to a value}} + foo(); + #pragma omp target dyn_groupprivate (argv[1]=2) // expected-error {{expression must have integral or unscoped enumeration type, not 'char *'}} expected-error {{expected ')'}} expected-note {{to match this '('}} + foo(); + #pragma omp target dyn_groupprivate (argc argc) // expected-error {{expected ')'}} expected-note {{to match this '('}} + foo(); + #pragma omp target dyn_groupprivate(argc+z) + foo(); + return 0; +} + +int main(int argc, char **argv) { +constexpr int n = -1; +int z; + #pragma omp target dyn_groupprivate // expected-error {{expected '(' after 'dyn_groupprivate'}} + foo(); + #pragma omp target dyn_groupprivate ( // expected-error {{expected expression}} expected-error {{expected ')'}} expected-note {{to match this '('}} + foo(); + #pragma omp target dyn_groupprivate () // expected-error {{expected expression}} + foo(); + #pragma omp target dyn_groupprivate (argc // expected-error {{expected ')'}} expected-note {{to match this '('}} + foo(); + #pragma omp target dyn_groupprivate (argc)) // expected-warning {{extra tokens at the end of '#pragma omp target' are ignored}} + foo(); + #pragma omp target dyn_groupprivate (argc > 0 ? argv[1] : argv[2]) // expected-error {{expression must have integral or unscoped enumeration type, not 'char *'}} + foo(); + #pragma omp target dyn_groupprivate (foobool(argc)), dyn_groupprivate (true) // expected-error {{directive '#pragma omp target' cannot contain more than one 'dyn_groupprivate' clause}} + foo(); + #pragma omp target dyn_groupprivate (S1) // expected-error {{'S1' does not refer to a value}} + foo(); + #pragma omp target dyn_groupprivate (argv[1]=2) // expected-error {{expression must have integral or unscoped enumeration type, not 'char *'}} expected-error {{expected ')'}} expected-note {{to match this '('}} + foo(); + #pragma omp target dyn_groupprivate (argc argc) // expected-error {{expected ')'}} expected-note {{to match this '('}} + foo(); + #pragma omp target dyn_groupprivate (1 0) // expected-error {{expected ')'}} expected-note {{to match this '('}} + foo(); + #pragma omp target dyn_groupprivate(dyn_groupprivate(tmain(argc, argv) // expected-error2 {{expected ')'}} expected-note2 {{to match this '('}} expected-note {{in instantiation of function template specialization 'tmain' requested here}} + foo(); + #pragma omp target dyn_groupprivate(-1) // expected-error {{argument to 'dyn_groupprivate' clause must be a non-negative integer value}} + foo(); + #pragma omp target dyn_groupprivate(cgrou) // expected-error {{use of undeclared identifier 'cgrou'}} + foo(); + #pragma omp target dyn_groupprivate(cgrou: argc) // expected-error {{use of undeclared identifier 'cgrou'}} expected-error {{expected ')'}} expected-note {{to match this '('}} + foo(); + #pragma omp target dyn_groupprivate(cgroup,cgroup: argc) // expected-error {{modifier 'cgroup' cannot be used along with modifier 'cgroup' in dyn_groupprivate}} + foo(); + #pragma omp target dyn_groupprivate(fallback,strict: argc) // expected-error {{modifier 'strict' cannot be used along with modifier 'fallback' in dyn_groupprivate}} + foo(); + #pragma omp target dyn_groupprivate(strict,fallback: argc) // expected-error {{modifier 'fallback' cannot be used along with modifier 'strict' in dyn_groupprivate}} + foo(); + #pragma omp target dyn_groupprivate(: argc) // expected-error {{expected ')'}} expected-error {{expected expression}} expected-note {{to match this '('}} + foo(); + + return tmain(argc, argv); +} + diff --git a/clang/test/OpenMP/target_teams_dyn_groupprivate_messages.cpp b/clang/test/OpenMP/target_teams_dyn_groupprivate_messages.cpp new file mode 100644 index 0000000000000..422dff547355c --- /dev/null +++ b/clang/test/OpenMP/target_teams_dyn_groupprivate_messages.cpp @@ -0,0 +1,87 @@ +// RUN: %clang_cc1 -verify -fopenmp -fopenmp-version=45 %s -Wuninitialized +// RUN: %clang_cc1 -verify -fopenmp-simd -fopenmp-version=45 %s -Wuninitialized +// RUN: %clang_cc1 -verify -fopenmp %s -Wuninitialized +// RUN: %clang_cc1 -verify -fopenmp-simd %s -Wuninitialized + +void foo() { +} + +bool foobool(int argc) { + return argc; +} + +struct S1; // expected-note {{declared here}} + +template // expected-note {{declared here}} +int tmain(T argc, S **argv) { + T z; + #pragma omp target teams dyn_groupprivate // expected-error {{expected '(' after 'dyn_groupprivate'}} + foo(); + #pragma omp target teams dyn_groupprivate ( // expected-error {{expected expression}} expected-error {{expected ')'}} expected-note {{to match this '('}} + foo(); + #pragma omp target teams dyn_groupprivate () // expected-error {{expected expression}} + foo(); + #pragma omp target teams dyn_groupprivate (argc // expected-error {{expected ')'}} expected-note {{to match this '('}} + foo(); + #pragma omp target teams dyn_groupprivate (argc)) // expected-warning {{extra tokens at the end of '#pragma omp target teams' are ignored}} + foo(); + #pragma omp target teams dyn_groupprivate (argc > 0 ? argv[1] : argv[2]) // expected-error {{expression must have integral or unscoped enumeration type, not 'char *'}} + foo(); + #pragma omp target teams dyn_groupprivate (foobool(argc)), dyn_groupprivate (true) // expected-error {{directive '#pragma omp target teams' cannot contain more than one 'dyn_groupprivate' clause}} + foo(); + #pragma omp target teams dyn_groupprivate (S) // expected-error {{'S' does not refer to a value}} + foo(); + #pragma omp target teams dyn_groupprivate (argv[1]=2) // expected-error {{expression must have integral or unscoped enumeration type, not 'char *'}} expected-error {{expected ')'}} expected-note {{to match this '('}} + foo(); + #pragma omp target teams dyn_groupprivate (argc argc) // expected-error {{expected ')'}} expected-note {{to match this '('}} + foo(); + #pragma omp target teams dyn_groupprivate(argc+z) + foo(); + return 0; +} + +int main(int argc, char **argv) { +constexpr int n = -1; +int z; + #pragma omp target teams dyn_groupprivate // expected-error {{expected '(' after 'dyn_groupprivate'}} + foo(); + #pragma omp target teams dyn_groupprivate ( // expected-error {{expected expression}} expected-error {{expected ')'}} expected-note {{to match this '('}} + foo(); + #pragma omp target teams dyn_groupprivate () // expected-error {{expected expression}} + foo(); + #pragma omp target teams dyn_groupprivate (argc // expected-error {{expected ')'}} expected-note {{to match this '('}} + foo(); + #pragma omp target teams dyn_groupprivate (argc)) // expected-warning {{extra tokens at the end of '#pragma omp target teams' are ignored}} + foo(); + #pragma omp target teams dyn_groupprivate (argc > 0 ? argv[1] : argv[2]) // expected-error {{expression must have integral or unscoped enumeration type, not 'char *'}} + foo(); + #pragma omp target teams dyn_groupprivate (foobool(argc)), dyn_groupprivate (true) // expected-error {{directive '#pragma omp target teams' cannot contain more than one 'dyn_groupprivate' clause}} + foo(); + #pragma omp target teams dyn_groupprivate (S1) // expected-error {{'S1' does not refer to a value}} + foo(); + #pragma omp target teams dyn_groupprivate (argv[1]=2) // expected-error {{expression must have integral or unscoped enumeration type, not 'char *'}} expected-error {{expected ')'}} expected-note {{to match this '('}} + foo(); + #pragma omp target teams dyn_groupprivate (argc argc) // expected-error {{expected ')'}} expected-note {{to match this '('}} + foo(); + #pragma omp target teams dyn_groupprivate (1 0) // expected-error {{expected ')'}} expected-note {{to match this '('}} + foo(); + #pragma omp target teams dyn_groupprivate(dyn_groupprivate(tmain(argc, argv) // expected-error2 {{expected ')'}} expected-note2 {{to match this '('}} expected-note {{in instantiation of function template specialization 'tmain' requested here}} + foo(); + #pragma omp target teams dyn_groupprivate(-1) // expected-error {{argument to 'dyn_groupprivate' clause must be a non-negative integer value}} + foo(); + #pragma omp target teams dyn_groupprivate(cgrou) // expected-error {{use of undeclared identifier 'cgrou'}} + foo(); + #pragma omp target teams dyn_groupprivate(cgrou: argc) // expected-error {{use of undeclared identifier 'cgrou'}} expected-error {{expected ')'}} expected-note {{to match this '('}} + foo(); + #pragma omp target teams dyn_groupprivate(cgroup,cgroup: argc) // expected-error {{modifier 'cgroup' cannot be used along with modifier 'cgroup' in dyn_groupprivate}} + foo(); + #pragma omp target teams dyn_groupprivate(fallback,strict: argc) // expected-error {{modifier 'strict' cannot be used along with modifier 'fallback' in dyn_groupprivate}} + foo(); + #pragma omp target teams dyn_groupprivate(strict,fallback: argc) // expected-error {{modifier 'fallback' cannot be used along with modifier 'strict' in dyn_groupprivate}} + foo(); + #pragma omp target teams dyn_groupprivate(: argc) // expected-error {{expected ')'}} expected-error {{expected expression}} expected-note {{to match this '('}} + foo(); + + return tmain(argc, argv); +} + diff --git a/clang/tools/libclang/CIndex.cpp b/clang/tools/libclang/CIndex.cpp index 9089984fa4a54..32d75d4cd8d38 100644 --- a/clang/tools/libclang/CIndex.cpp +++ b/clang/tools/libclang/CIndex.cpp @@ -2808,6 +2808,11 @@ void OMPClauseEnqueue::VisitOMPXDynCGroupMemClause( VisitOMPClauseWithPreInit(C); Visitor->AddStmt(C->getSize()); } +void OMPClauseEnqueue::VisitOMPDynGroupprivateClause( + const OMPDynGroupprivateClause *C) { + VisitOMPClauseWithPreInit(C); + Visitor->AddStmt(C->getSize()); +} void OMPClauseEnqueue::VisitOMPDoacrossClause(const OMPDoacrossClause *C) { VisitOMPClauseList(C); } diff --git a/llvm/include/llvm/Frontend/OpenMP/OMP.td b/llvm/include/llvm/Frontend/OpenMP/OMP.td index 1b94657dfae1e..78442a63b3ac4 100644 --- a/llvm/include/llvm/Frontend/OpenMP/OMP.td +++ b/llvm/include/llvm/Frontend/OpenMP/OMP.td @@ -178,6 +178,9 @@ def OMPC_Doacross : Clause<[Spelling<"doacross">]> { def OMPC_DynamicAllocators : Clause<[Spelling<"dynamic_allocators">]> { let clangClass = "OMPDynamicAllocatorsClause"; } +def OMPC_DynGroupprivate : Clause<[Spelling<"dyn_groupprivate">]> { + let clangClass = "OMPDynGroupprivateClause"; +} def OMPC_Enter : Clause<[Spelling<"enter">]> { let flangClass = "OmpObjectList"; } @@ -1104,6 +1107,7 @@ def OMP_Target : Directive<[Spelling<"target">]> { let allowedOnceClauses = [ VersionedClause, VersionedClause, + VersionedClause, VersionedClause, VersionedClause, VersionedClause, @@ -1254,6 +1258,7 @@ def OMP_Teams : Directive<[Spelling<"teams">]> { ]; let allowedOnceClauses = [ VersionedClause, + VersionedClause, VersionedClause, VersionedClause, VersionedClause, @@ -1522,6 +1527,7 @@ def OMP_target_loop : Directive<[Spelling<"target loop">]> { let allowedOnceClauses = [ VersionedClause, VersionedClause, + VersionedClause, VersionedClause, VersionedClause, VersionedClause, @@ -1983,6 +1989,7 @@ def OMP_TargetParallel : Directive<[Spelling<"target parallel">]> { let allowedOnceClauses = [ VersionedClause, VersionedClause, + VersionedClause, VersionedClause, VersionedClause, VersionedClause, @@ -2086,6 +2093,7 @@ def OMP_TargetParallelFor : Directive<[Spelling<"target parallel for">]> { VersionedClause, ]; let allowedOnceClauses = [ + VersionedClause, VersionedClause, VersionedClause, ]; @@ -2126,6 +2134,7 @@ def OMP_TargetParallelForSimd VersionedClause, ]; let allowedOnceClauses = [ + VersionedClause, VersionedClause, VersionedClause, ]; @@ -2155,6 +2164,7 @@ def OMP_target_parallel_loop : Directive<[Spelling<"target parallel loop">]> { VersionedClause, VersionedClause, VersionedClause, + VersionedClause, VersionedClause, VersionedClause, VersionedClause, @@ -2189,6 +2199,7 @@ def OMP_TargetSimd : Directive<[Spelling<"target simd">]> { VersionedClause, VersionedClause, VersionedClause, + VersionedClause, VersionedClause, VersionedClause, VersionedClause, @@ -2220,6 +2231,7 @@ def OMP_TargetTeams : Directive<[Spelling<"target teams">]> { VersionedClause, VersionedClause, VersionedClause, + VersionedClause, VersionedClause, VersionedClause, VersionedClause, @@ -2252,6 +2264,7 @@ def OMP_TargetTeamsDistribute VersionedClause, VersionedClause, VersionedClause, + VersionedClause, VersionedClause, VersionedClause, VersionedClause, @@ -2284,6 +2297,7 @@ def OMP_TargetTeamsDistributeParallelDo VersionedClause, VersionedClause, VersionedClause, + VersionedClause, VersionedClause, VersionedClause, VersionedClause, @@ -2322,6 +2336,7 @@ def OMP_TargetTeamsDistributeParallelDoSimd VersionedClause, VersionedClause, VersionedClause, + VersionedClause, VersionedClause, VersionedClause, VersionedClause, @@ -2367,6 +2382,7 @@ def OMP_TargetTeamsDistributeParallelFor VersionedClause, ]; let allowedOnceClauses = [ + VersionedClause, VersionedClause, ]; let leafConstructs = @@ -2409,6 +2425,7 @@ def OMP_TargetTeamsDistributeParallelForSimd VersionedClause, ]; let allowedOnceClauses = [ + VersionedClause, VersionedClause, ]; let leafConstructs = @@ -2441,6 +2458,7 @@ def OMP_TargetTeamsDistributeSimd VersionedClause, VersionedClause, VersionedClause, + VersionedClause, VersionedClause, VersionedClause, VersionedClause, @@ -2474,6 +2492,7 @@ def OMP_target_teams_loop : Directive<[Spelling<"target teams loop">]> { VersionedClause, VersionedClause, VersionedClause, + VersionedClause, VersionedClause, VersionedClause, VersionedClause, @@ -2532,6 +2551,7 @@ def OMP_TeamsDistribute : Directive<[Spelling<"teams distribute">]> { VersionedClause, ]; let allowedOnceClauses = [ + VersionedClause, VersionedClause, VersionedClause, ]; @@ -2555,6 +2575,7 @@ def OMP_TeamsDistributeParallelDo VersionedClause, VersionedClause, VersionedClause, + VersionedClause, VersionedClause, VersionedClause, VersionedClause, @@ -2584,6 +2605,7 @@ def OMP_TeamsDistributeParallelDoSimd VersionedClause, VersionedClause, VersionedClause, + VersionedClause, VersionedClause, VersionedClause, VersionedClause, @@ -2632,6 +2654,7 @@ def OMP_TeamsDistributeParallelForSimd VersionedClause, VersionedClause, VersionedClause, + VersionedClause, VersionedClause, VersionedClause, VersionedClause, @@ -2673,6 +2696,7 @@ def OMP_TeamsDistributeSimd : Directive<[Spelling<"teams distribute simd">]> { VersionedClause, VersionedClause, VersionedClause, + VersionedClause, VersionedClause, VersionedClause, VersionedClause, @@ -2696,6 +2720,7 @@ def OMP_teams_loop : Directive<[Spelling<"teams loop">]> { VersionedClause, VersionedClause, VersionedClause, + VersionedClause, VersionedClause, VersionedClause, VersionedClause, From f66e5faa93cf2f40bd2a6bd7a95abddf78fb6076 Mon Sep 17 00:00:00 2001 From: Kevin Sala Date: Fri, 8 Aug 2025 11:04:06 -0700 Subject: [PATCH 02/17] [OpenMP] Add codegen support for dyn_groupprivate clause --- clang/lib/CodeGen/CGOpenMPRuntime.cpp | 40 ++++++++++++------- .../llvm/Frontend/OpenMP/OMPIRBuilder.h | 7 +++- llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp | 10 ++++- 3 files changed, 39 insertions(+), 18 deletions(-) diff --git a/clang/lib/CodeGen/CGOpenMPRuntime.cpp b/clang/lib/CodeGen/CGOpenMPRuntime.cpp index a5f2f0efa2c3b..d9121827a813a 100644 --- a/clang/lib/CodeGen/CGOpenMPRuntime.cpp +++ b/clang/lib/CodeGen/CGOpenMPRuntime.cpp @@ -9489,18 +9489,30 @@ static llvm::Value *emitDeviceID( return DeviceID; } -static llvm::Value *emitDynCGGroupMem(const OMPExecutableDirective &D, - CodeGenFunction &CGF) { - llvm::Value *DynCGroupMem = CGF.Builder.getInt32(0); - - if (auto *DynMemClause = D.getSingleClause()) { - CodeGenFunction::RunCleanupsScope DynCGroupMemScope(CGF); - llvm::Value *DynCGroupMemVal = CGF.EmitScalarExpr( - DynMemClause->getSize(), /*IgnoreResultAssign=*/true); - DynCGroupMem = CGF.Builder.CreateIntCast(DynCGroupMemVal, CGF.Int32Ty, - /*isSigned=*/false); - } - return DynCGroupMem; +static std::pair +emitDynCGroupMem(const OMPExecutableDirective &D, CodeGenFunction &CGF) { + llvm::Value *DynGP = CGF.Builder.getInt32(0); + bool DynGPFallback = false; + + if (auto *DynGPClause = D.getSingleClause()) { + CodeGenFunction::RunCleanupsScope DynGPScope(CGF); + llvm::Value *DynGPVal = + CGF.EmitScalarExpr(DynGPClause->getSize(), /*IgnoreResultAssign=*/true); + DynGP = CGF.Builder.CreateIntCast(DynGPVal, CGF.Int32Ty, + /*isSigned=*/false); + DynGPFallback = (DynGPClause->getFirstDynGroupprivateModifier() != + OMPC_DYN_GROUPPRIVATE_strict && + DynGPClause->getSecondDynGroupprivateModifier() != + OMPC_DYN_GROUPPRIVATE_strict); + } else if (auto *OMPXDynCGClause = + D.getSingleClause()) { + CodeGenFunction::RunCleanupsScope DynCGMemScope(CGF); + llvm::Value *DynCGMemVal = CGF.EmitScalarExpr(OMPXDynCGClause->getSize(), + /*IgnoreResultAssign=*/true); + DynGP = CGF.Builder.CreateIntCast(DynCGMemVal, CGF.Int32Ty, + /*isSigned=*/false); + } + return {DynGP, DynGPFallback}; } static void genMapInfoForCaptures( MappableExprsHandler &MEHandler, CodeGenFunction &CGF, @@ -9710,7 +9722,7 @@ static void emitTargetCallKernelLaunch( llvm::Value *RTLoc = OMPRuntime->emitUpdateLocation(CGF, D.getBeginLoc()); llvm::Value *NumIterations = OMPRuntime->emitTargetNumIterationsCall(CGF, D, SizeEmitter); - llvm::Value *DynCGGroupMem = emitDynCGGroupMem(D, CGF); + auto [DynCGroupMem, DynCGroupMemFallback] = emitDynCGroupMem(D, CGF); llvm::OpenMPIRBuilder::InsertPointTy AllocaIP( CGF.AllocaInsertPt->getParent(), CGF.AllocaInsertPt->getIterator()); @@ -9720,7 +9732,7 @@ static void emitTargetCallKernelLaunch( llvm::OpenMPIRBuilder::TargetKernelArgs Args( NumTargetItems, RTArgs, NumIterations, NumTeams, NumThreads, - DynCGGroupMem, HasNoWait); + DynCGroupMem, HasNoWait, DynCGroupMemFallback); llvm::OpenMPIRBuilder::InsertPointTy AfterIP = cantFail(OMPRuntime->getOMPBuilder().emitKernelLaunch( diff --git a/llvm/include/llvm/Frontend/OpenMP/OMPIRBuilder.h b/llvm/include/llvm/Frontend/OpenMP/OMPIRBuilder.h index 19a4058b64382..ebc50eecb551e 100644 --- a/llvm/include/llvm/Frontend/OpenMP/OMPIRBuilder.h +++ b/llvm/include/llvm/Frontend/OpenMP/OMPIRBuilder.h @@ -2341,17 +2341,20 @@ class OpenMPIRBuilder { Value *DynCGGroupMem = nullptr; /// True if the kernel has 'no wait' clause. bool HasNoWait = false; + /// True if the dynamic shared memory may fallback. + bool MayFallbackDynCGroupMem = false; // Constructors for TargetKernelArgs. TargetKernelArgs() {} TargetKernelArgs(unsigned NumTargetItems, TargetDataRTArgs RTArgs, Value *NumIterations, ArrayRef NumTeams, ArrayRef NumThreads, Value *DynCGGroupMem, - bool HasNoWait) + bool HasNoWait, bool MayFallbackDynCGroupMem) : NumTargetItems(NumTargetItems), RTArgs(RTArgs), NumIterations(NumIterations), NumTeams(NumTeams), NumThreads(NumThreads), DynCGGroupMem(DynCGGroupMem), - HasNoWait(HasNoWait) {} + HasNoWait(HasNoWait), + MayFallbackDynCGroupMem(MayFallbackDynCGroupMem) {} }; /// Create the kernel args vector used by emitTargetKernel. This function diff --git a/llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp b/llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp index 170224616ac64..e600508d347cb 100644 --- a/llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp +++ b/llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp @@ -506,7 +506,13 @@ void OpenMPIRBuilder::getKernelArgsVector(TargetKernelArgs &KernelArgs, auto Int32Ty = Type::getInt32Ty(Builder.getContext()); constexpr const size_t MaxDim = 3; Value *ZeroArray = Constant::getNullValue(ArrayType::get(Int32Ty, MaxDim)); - Value *Flags = Builder.getInt64(KernelArgs.HasNoWait); + + Value *HasNoWaitFlag = Builder.getInt64(KernelArgs.HasNoWait); + Value *MayFallbackDynCGroupMemFlag = + Builder.getInt64(KernelArgs.MayFallbackDynCGroupMem); + MayFallbackDynCGroupMemFlag = + Builder.CreateShl(MayFallbackDynCGroupMemFlag, 2); + Value *Flags = Builder.CreateOr(HasNoWaitFlag, MayFallbackDynCGroupMemFlag); assert(!KernelArgs.NumTeams.empty() && !KernelArgs.NumThreads.empty()); @@ -7891,7 +7897,7 @@ static void emitTargetCall( KArgs = OpenMPIRBuilder::TargetKernelArgs(NumTargetItems, RTArgs, TripCount, NumTeamsC, NumThreadsC, - DynCGGroupMem, HasNoWait); + DynCGGroupMem, HasNoWait, false); // Assume no error was returned because TaskBodyCB and // EmitTargetCallFallbackCB don't produce any. From c34e0627d5909a7f6e3822ddc9e7f6a844604f10 Mon Sep 17 00:00:00 2001 From: Kevin Sala Date: Sun, 17 Aug 2025 00:26:46 -0700 Subject: [PATCH 03/17] [OpenMP][Flang] Add empty clause support for dyn_groupprivate in Flang --- flang/include/flang/Lower/OpenMP/Clauses.h | 1 + flang/lib/Lower/OpenMP/Clauses.cpp | 2 ++ flang/lib/Parser/openmp-parsers.cpp | 2 ++ flang/lib/Semantics/check-omp-structure.cpp | 1 + llvm/include/llvm/Frontend/OpenMP/ClauseT.h | 23 ++++++++++++------- .../Frontend/OpenMPDecompositionTest.cpp | 1 + 6 files changed, 22 insertions(+), 8 deletions(-) diff --git a/flang/include/flang/Lower/OpenMP/Clauses.h b/flang/include/flang/Lower/OpenMP/Clauses.h index 7f317f05f67b7..1ab594ffcd209 100644 --- a/flang/include/flang/Lower/OpenMP/Clauses.h +++ b/flang/include/flang/Lower/OpenMP/Clauses.h @@ -219,6 +219,7 @@ using DistSchedule = tomp::clause::DistScheduleT; using Doacross = tomp::clause::DoacrossT; using DynamicAllocators = tomp::clause::DynamicAllocatorsT; +using DynGroupprivate = tomp::clause::DynGroupprivateT; using Enter = tomp::clause::EnterT; using Exclusive = tomp::clause::ExclusiveT; using Fail = tomp::clause::FailT; diff --git a/flang/lib/Lower/OpenMP/Clauses.cpp b/flang/lib/Lower/OpenMP/Clauses.cpp index 22a07219d3a50..1ab1252f33203 100644 --- a/flang/lib/Lower/OpenMP/Clauses.cpp +++ b/flang/lib/Lower/OpenMP/Clauses.cpp @@ -220,6 +220,7 @@ MAKE_EMPTY_CLASS(Acquire, Acquire); MAKE_EMPTY_CLASS(Capture, Capture); MAKE_EMPTY_CLASS(Compare, Compare); MAKE_EMPTY_CLASS(DynamicAllocators, DynamicAllocators); +MAKE_EMPTY_CLASS(DynGroupprivate, DynGroupprivate); MAKE_EMPTY_CLASS(Full, Full); MAKE_EMPTY_CLASS(Inbranch, Inbranch); MAKE_EMPTY_CLASS(Mergeable, Mergeable); @@ -769,6 +770,7 @@ Doacross make(const parser::OmpClause::Doacross &inp, } // DynamicAllocators: empty +// DynGroupprivate: empty Enter make(const parser::OmpClause::Enter &inp, semantics::SemanticsContext &semaCtx) { diff --git a/flang/lib/Parser/openmp-parsers.cpp b/flang/lib/Parser/openmp-parsers.cpp index d70aaab82cbab..54082c6bc02f6 100644 --- a/flang/lib/Parser/openmp-parsers.cpp +++ b/flang/lib/Parser/openmp-parsers.cpp @@ -999,6 +999,8 @@ TYPE_PARSER( // construct(parenthesized(Parser{})) || "DYNAMIC_ALLOCATORS" >> construct(construct()) || + "DYN_GROUPPRIVATE" >> + construct(construct()) || "ENTER" >> construct(construct( parenthesized(Parser{}))) || "EXCLUSIVE" >> construct(construct( diff --git a/flang/lib/Semantics/check-omp-structure.cpp b/flang/lib/Semantics/check-omp-structure.cpp index 2425265e196c6..3675bb6d9d38e 100644 --- a/flang/lib/Semantics/check-omp-structure.cpp +++ b/flang/lib/Semantics/check-omp-structure.cpp @@ -2545,6 +2545,7 @@ CHECK_SIMPLE_CLAUSE(Default, OMPC_default) CHECK_SIMPLE_CLAUSE(Depobj, OMPC_depobj) CHECK_SIMPLE_CLAUSE(DeviceType, OMPC_device_type) CHECK_SIMPLE_CLAUSE(DistSchedule, OMPC_dist_schedule) +CHECK_SIMPLE_CLAUSE(DynGroupprivate, OMPC_dyn_groupprivate) CHECK_SIMPLE_CLAUSE(Exclusive, OMPC_exclusive) CHECK_SIMPLE_CLAUSE(Final, OMPC_final) CHECK_SIMPLE_CLAUSE(Flush, OMPC_flush) diff --git a/llvm/include/llvm/Frontend/OpenMP/ClauseT.h b/llvm/include/llvm/Frontend/OpenMP/ClauseT.h index de888ff86fe91..df75bccc2c867 100644 --- a/llvm/include/llvm/Frontend/OpenMP/ClauseT.h +++ b/llvm/include/llvm/Frontend/OpenMP/ClauseT.h @@ -574,6 +574,12 @@ struct DynamicAllocatorsT { using EmptyTrait = std::true_type; }; +// V6.1: `dyn_groupprivate` clause +template // +struct DynGroupprivateT { + using EmptyTrait = std::true_type; +}; + // V5.2: [5.8.4] `enter` clause template // struct EnterT { @@ -1243,14 +1249,15 @@ using ExtensionClausesT = template using EmptyClausesT = std::variant< AcqRelT, AcquireT, CaptureT, CompareT, - DynamicAllocatorsT, FullT, InbranchT, - MergeableT, NogroupT, NoOpenmpRoutinesT, - NoOpenmpT, NoParallelismT, NotinbranchT, - NowaitT, ReadT, RelaxedT, ReleaseT, - ReverseOffloadT, SeqCstT, SimdT, - ThreadsT, UnifiedAddressT, UnifiedSharedMemoryT, - UnknownT, UntiedT, UseT, WeakT, - WriteT, NoOpenmpConstructsT, SelfMapsT>; + DynamicAllocatorsT, DynGroupprivateT, FullT, + InbranchT, MergeableT, NogroupT, + NoOpenmpRoutinesT, NoOpenmpT, NoParallelismT, + NotinbranchT, NowaitT, ReadT, RelaxedT, + ReleaseT, ReverseOffloadT, SeqCstT, + SimdT, ThreadsT, UnifiedAddressT, + UnifiedSharedMemoryT, UnknownT, UntiedT, + UseT, WeakT, WriteT, + NoOpenmpConstructsT, SelfMapsT>; template using IncompleteClausesT = diff --git a/llvm/unittests/Frontend/OpenMPDecompositionTest.cpp b/llvm/unittests/Frontend/OpenMPDecompositionTest.cpp index 6189d0954891b..79d640968fe86 100644 --- a/llvm/unittests/Frontend/OpenMPDecompositionTest.cpp +++ b/llvm/unittests/Frontend/OpenMPDecompositionTest.cpp @@ -88,6 +88,7 @@ using DistSchedule = tomp::clause::DistScheduleT; using Doacross = tomp::clause::DoacrossT; using DynamicAllocators = tomp::clause::DynamicAllocatorsT; +using DynGroupprivate = tomp::clause::DynGroupprivateT; using Enter = tomp::clause::EnterT; using Exclusive = tomp::clause::ExclusiveT; using Fail = tomp::clause::FailT; From 0a7a96d0c47a7e707fecb7b01d2a8ac262b70990 Mon Sep 17 00:00:00 2001 From: Krzysztof Parzyszek Date: Mon, 18 Aug 2025 16:01:33 -0500 Subject: [PATCH 04/17] Fix merge in ClauseT.h --- llvm/include/llvm/Frontend/OpenMP/ClauseT.h | 17 ++++++++--------- 1 file changed, 8 insertions(+), 9 deletions(-) diff --git a/llvm/include/llvm/Frontend/OpenMP/ClauseT.h b/llvm/include/llvm/Frontend/OpenMP/ClauseT.h index b30c30e590062..8ea50e7e8d416 100644 --- a/llvm/include/llvm/Frontend/OpenMP/ClauseT.h +++ b/llvm/include/llvm/Frontend/OpenMP/ClauseT.h @@ -1254,15 +1254,14 @@ using ExtensionClausesT = template using EmptyClausesT = std::variant< AcqRelT, AcquireT, CaptureT, CompareT, - DynamicAllocatorsT, DynGroupprivateT, FullT, - InbranchT, MergeableT, NogroupT, - NoOpenmpRoutinesT, NoOpenmpT, NoParallelismT, - NotinbranchT, NowaitT, ReadT, RelaxedT, - ReleaseT, ReverseOffloadT, SeqCstT, - SimdT, ThreadsT, UnifiedAddressT, - UnifiedSharedMemoryT, UnknownT, UntiedT, - UseT, WeakT, WriteT, - NoOpenmpConstructsT, SelfMapsT>; + DynamicAllocatorsT, FullT, InbranchT, + MergeableT, NogroupT, NoOpenmpRoutinesT, + NoOpenmpT, NoParallelismT, NotinbranchT, + NowaitT, ReadT, RelaxedT, ReleaseT, + ReverseOffloadT, SeqCstT, SimdT, + ThreadsT, UnifiedAddressT, UnifiedSharedMemoryT, + UnknownT, UntiedT, UseT, WeakT, + WriteT, NoOpenmpConstructsT, SelfMapsT>; template using IncompleteClausesT = From b5b843914647679b0d58b619c652d60b3c1e83e8 Mon Sep 17 00:00:00 2001 From: Krzysztof Parzyszek Date: Mon, 18 Aug 2025 16:03:46 -0500 Subject: [PATCH 05/17] more merge fixes --- flang/lib/Lower/OpenMP/Clauses.cpp | 2 -- llvm/include/llvm/Frontend/OpenMP/OMP.td | 3 ++- 2 files changed, 2 insertions(+), 3 deletions(-) diff --git a/flang/lib/Lower/OpenMP/Clauses.cpp b/flang/lib/Lower/OpenMP/Clauses.cpp index 729dba42a0777..1a16e1c87e250 100644 --- a/flang/lib/Lower/OpenMP/Clauses.cpp +++ b/flang/lib/Lower/OpenMP/Clauses.cpp @@ -220,7 +220,6 @@ MAKE_EMPTY_CLASS(Acquire, Acquire); MAKE_EMPTY_CLASS(Capture, Capture); MAKE_EMPTY_CLASS(Compare, Compare); MAKE_EMPTY_CLASS(DynamicAllocators, DynamicAllocators); -MAKE_EMPTY_CLASS(DynGroupprivate, DynGroupprivate); MAKE_EMPTY_CLASS(Full, Full); MAKE_EMPTY_CLASS(Inbranch, Inbranch); MAKE_EMPTY_CLASS(Mergeable, Mergeable); @@ -772,7 +771,6 @@ Doacross make(const parser::OmpClause::Doacross &inp, } // DynamicAllocators: empty -// DynGroupprivate: empty DynGroupprivate make(const parser::OmpClause::DynGroupprivate &inp, semantics::SemanticsContext &semaCtx) { diff --git a/llvm/include/llvm/Frontend/OpenMP/OMP.td b/llvm/include/llvm/Frontend/OpenMP/OMP.td index b107e7f6f185a..56dd92fc2c023 100644 --- a/llvm/include/llvm/Frontend/OpenMP/OMP.td +++ b/llvm/include/llvm/Frontend/OpenMP/OMP.td @@ -2738,4 +2738,5 @@ def OMP_teams_loop : Directive<[Spelling<"teams loop">]> { ]; let leafConstructs = [OMP_Teams, OMP_loop]; let category = CA_Executable; -} \ No newline at end of file +} + From 6e4c54736e28b34ada5484952c794da54caa34c9 Mon Sep 17 00:00:00 2001 From: Krzysztof Parzyszek Date: Mon, 18 Aug 2025 16:05:27 -0500 Subject: [PATCH 06/17] more merge fixes --- llvm/include/llvm/Frontend/OpenMP/OMP.td | 2 -- 1 file changed, 2 deletions(-) diff --git a/llvm/include/llvm/Frontend/OpenMP/OMP.td b/llvm/include/llvm/Frontend/OpenMP/OMP.td index 56dd92fc2c023..043fc591c9f76 100644 --- a/llvm/include/llvm/Frontend/OpenMP/OMP.td +++ b/llvm/include/llvm/Frontend/OpenMP/OMP.td @@ -2662,7 +2662,6 @@ def OMP_TeamsDistributeParallelForSimd VersionedClause, VersionedClause, VersionedClause, - VersionedClause, VersionedClause, VersionedClause, VersionedClause, @@ -2739,4 +2738,3 @@ def OMP_teams_loop : Directive<[Spelling<"teams loop">]> { let leafConstructs = [OMP_Teams, OMP_loop]; let category = CA_Executable; } - From 84fc963023b8638cf44c68efef21a5f8d0d8128b Mon Sep 17 00:00:00 2001 From: Kevin Sala Date: Sun, 24 Aug 2025 19:47:24 -0700 Subject: [PATCH 07/17] Add fixes and improvements after merge --- clang/lib/AST/StmtProfile.cpp | 2 +- .../target_dyn_groupprivate_messages.cpp | 6 +- ...target_teams_dyn_groupprivate_messages.cpp | 6 +- .../teams_dyn_groupprivate_messages.cpp | 85 +++++++++++++++++++ 4 files changed, 90 insertions(+), 9 deletions(-) create mode 100644 clang/test/OpenMP/teams_dyn_groupprivate_messages.cpp diff --git a/clang/lib/AST/StmtProfile.cpp b/clang/lib/AST/StmtProfile.cpp index 1c8fec1e7328e..714c740c8fe13 100644 --- a/clang/lib/AST/StmtProfile.cpp +++ b/clang/lib/AST/StmtProfile.cpp @@ -958,7 +958,7 @@ void OMPClauseProfiler::VisitOMPXDynCGroupMemClause( } void OMPClauseProfiler::VisitOMPDynGroupprivateClause( const OMPDynGroupprivateClause *C) { - VistOMPClauseWithPreInit(C); + VisitOMPClauseWithPreInit(C); if (auto *Size = C->getSize()) Profiler->VisitStmt(Size); } diff --git a/clang/test/OpenMP/target_dyn_groupprivate_messages.cpp b/clang/test/OpenMP/target_dyn_groupprivate_messages.cpp index d5d855ee33e1f..f924d2bb45eaa 100644 --- a/clang/test/OpenMP/target_dyn_groupprivate_messages.cpp +++ b/clang/test/OpenMP/target_dyn_groupprivate_messages.cpp @@ -1,7 +1,5 @@ -// RUN: %clang_cc1 -verify -fopenmp -fopenmp-version=45 %s -Wuninitialized -// RUN: %clang_cc1 -verify -fopenmp-simd -fopenmp-version=45 %s -Wuninitialized -// RUN: %clang_cc1 -verify -fopenmp %s -Wuninitialized -// RUN: %clang_cc1 -verify -fopenmp-simd %s -Wuninitialized +// RUN: %clang_cc1 -verify -fopenmp -fopenmp-version=61 %s -Wuninitialized +// RUN: %clang_cc1 -verify -fopenmp-simd -fopenmp-version=61 %s -Wuninitialized void foo() { } diff --git a/clang/test/OpenMP/target_teams_dyn_groupprivate_messages.cpp b/clang/test/OpenMP/target_teams_dyn_groupprivate_messages.cpp index 422dff547355c..d05bb433eab1c 100644 --- a/clang/test/OpenMP/target_teams_dyn_groupprivate_messages.cpp +++ b/clang/test/OpenMP/target_teams_dyn_groupprivate_messages.cpp @@ -1,7 +1,5 @@ -// RUN: %clang_cc1 -verify -fopenmp -fopenmp-version=45 %s -Wuninitialized -// RUN: %clang_cc1 -verify -fopenmp-simd -fopenmp-version=45 %s -Wuninitialized -// RUN: %clang_cc1 -verify -fopenmp %s -Wuninitialized -// RUN: %clang_cc1 -verify -fopenmp-simd %s -Wuninitialized +// RUN: %clang_cc1 -verify -fopenmp -fopenmp-version=61 %s -Wuninitialized +// RUN: %clang_cc1 -verify -fopenmp-simd -fopenmp-version=61 %s -Wuninitialized void foo() { } diff --git a/clang/test/OpenMP/teams_dyn_groupprivate_messages.cpp b/clang/test/OpenMP/teams_dyn_groupprivate_messages.cpp new file mode 100644 index 0000000000000..a55a2a570a1b8 --- /dev/null +++ b/clang/test/OpenMP/teams_dyn_groupprivate_messages.cpp @@ -0,0 +1,85 @@ +// RUN: %clang_cc1 -verify -fopenmp -fopenmp-version=61 %s -Wuninitialized +// RUN: %clang_cc1 -verify -fopenmp-simd -fopenmp-version=61 %s -Wuninitialized + +void foo() { +} + +bool foobool(int argc) { + return argc; +} + +struct S1; // expected-note {{declared here}} + +template // expected-note {{declared here}} +int tmain(T argc, S **argv) { + T z; + #pragma omp teams dyn_groupprivate // expected-error {{expected '(' after 'dyn_groupprivate'}} + foo(); + #pragma omp teams dyn_groupprivate ( // expected-error {{expected expression}} expected-error {{expected ')'}} expected-note {{to match this '('}} + foo(); + #pragma omp teams dyn_groupprivate () // expected-error {{expected expression}} + foo(); + #pragma omp teams dyn_groupprivate (argc // expected-error {{expected ')'}} expected-note {{to match this '('}} + foo(); + #pragma omp teams dyn_groupprivate (argc)) // expected-warning {{extra tokens at the end of '#pragma omp teams' are ignored}} + foo(); + #pragma omp teams dyn_groupprivate (argc > 0 ? argv[1] : argv[2]) // expected-error {{expression must have integral or unscoped enumeration type, not 'char *'}} + foo(); + #pragma omp teams dyn_groupprivate (foobool(argc)), dyn_groupprivate (true) // expected-error {{directive '#pragma omp teams' cannot contain more than one 'dyn_groupprivate' clause}} + foo(); + #pragma omp teams dyn_groupprivate (S) // expected-error {{'S' does not refer to a value}} + foo(); + #pragma omp teams dyn_groupprivate (argv[1]=2) // expected-error {{expression must have integral or unscoped enumeration type, not 'char *'}} expected-error {{expected ')'}} expected-note {{to match this '('}} + foo(); + #pragma omp teams dyn_groupprivate (argc argc) // expected-error {{expected ')'}} expected-note {{to match this '('}} + foo(); + #pragma omp teams dyn_groupprivate(argc+z) + foo(); + return 0; +} + +int main(int argc, char **argv) { +constexpr int n = -1; +int z; + #pragma omp teams dyn_groupprivate // expected-error {{expected '(' after 'dyn_groupprivate'}} + foo(); + #pragma omp teams dyn_groupprivate ( // expected-error {{expected expression}} expected-error {{expected ')'}} expected-note {{to match this '('}} + foo(); + #pragma omp teams dyn_groupprivate () // expected-error {{expected expression}} + foo(); + #pragma omp teams dyn_groupprivate (argc // expected-error {{expected ')'}} expected-note {{to match this '('}} + foo(); + #pragma omp teams dyn_groupprivate (argc)) // expected-warning {{extra tokens at the end of '#pragma omp teams' are ignored}} + foo(); + #pragma omp teams dyn_groupprivate (argc > 0 ? argv[1] : argv[2]) // expected-error {{expression must have integral or unscoped enumeration type, not 'char *'}} + foo(); + #pragma omp teams dyn_groupprivate (foobool(argc)), dyn_groupprivate (true) // expected-error {{directive '#pragma omp teams' cannot contain more than one 'dyn_groupprivate' clause}} + foo(); + #pragma omp teams dyn_groupprivate (S1) // expected-error {{'S1' does not refer to a value}} + foo(); + #pragma omp teams dyn_groupprivate (argv[1]=2) // expected-error {{expression must have integral or unscoped enumeration type, not 'char *'}} expected-error {{expected ')'}} expected-note {{to match this '('}} + foo(); + #pragma omp teams dyn_groupprivate (argc argc) // expected-error {{expected ')'}} expected-note {{to match this '('}} + foo(); + #pragma omp teams dyn_groupprivate (1 0) // expected-error {{expected ')'}} expected-note {{to match this '('}} + foo(); + #pragma omp teams dyn_groupprivate(dyn_groupprivate(tmain(argc, argv) // expected-error2 {{expected ')'}} expected-note2 {{to match this '('}} expected-note {{in instantiation of function template specialization 'tmain' requested here}} + foo(); + #pragma omp teams dyn_groupprivate(-1) // expected-error {{argument to 'dyn_groupprivate' clause must be a non-negative integer value}} + foo(); + #pragma omp teams dyn_groupprivate(cgrou) // expected-error {{use of undeclared identifier 'cgrou'}} + foo(); + #pragma omp teams dyn_groupprivate(cgrou: argc) // expected-error {{use of undeclared identifier 'cgrou'}} expected-error {{expected ')'}} expected-note {{to match this '('}} + foo(); + #pragma omp teams dyn_groupprivate(cgroup,cgroup: argc) // expected-error {{modifier 'cgroup' cannot be used along with modifier 'cgroup' in dyn_groupprivate}} + foo(); + #pragma omp teams dyn_groupprivate(fallback,strict: argc) // expected-error {{modifier 'strict' cannot be used along with modifier 'fallback' in dyn_groupprivate}} + foo(); + #pragma omp teams dyn_groupprivate(strict,fallback: argc) // expected-error {{modifier 'fallback' cannot be used along with modifier 'strict' in dyn_groupprivate}} + foo(); + #pragma omp teams dyn_groupprivate(: argc) // expected-error {{expected ')'}} expected-error {{expected expression}} expected-note {{to match this '('}} + foo(); + + return tmain(argc, argv); +} + From 86f0cf0a5fcbea3dfbb07dfcaeb610380596113b Mon Sep 17 00:00:00 2001 From: Kevin Sala Date: Wed, 22 Oct 2025 22:14:16 -0700 Subject: [PATCH 08/17] Update syntax for fallback complex modifier --- clang/include/clang/AST/OpenMPClause.h | 65 +++++++-------- .../clang/Basic/DiagnosticSemaKinds.td | 2 +- clang/include/clang/Basic/OpenMPKinds.def | 11 ++- clang/include/clang/Basic/OpenMPKinds.h | 10 ++- clang/include/clang/Sema/SemaOpenMP.h | 2 +- clang/lib/AST/OpenMPClause.cpp | 13 ++- clang/lib/Basic/OpenMPKinds.cpp | 7 ++ clang/lib/Parse/ParseOpenMP.cpp | 79 ++++++++++++++----- clang/lib/Sema/SemaOpenMP.cpp | 23 ++---- clang/lib/Sema/TreeTransform.h | 10 +-- clang/lib/Serialization/ASTReader.cpp | 11 +-- clang/lib/Serialization/ASTWriter.cpp | 8 +- .../target_dyn_groupprivate_messages.cpp | 8 +- ...target_teams_dyn_groupprivate_messages.cpp | 8 +- .../teams_dyn_groupprivate_messages.cpp | 8 +- 15 files changed, 158 insertions(+), 107 deletions(-) diff --git a/clang/include/clang/AST/OpenMPClause.h b/clang/include/clang/AST/OpenMPClause.h index a3983120df069..f671884280134 100644 --- a/clang/include/clang/AST/OpenMPClause.h +++ b/clang/include/clang/AST/OpenMPClause.h @@ -9781,8 +9781,8 @@ class OMPDynGroupprivateClause : public OMPClause, public OMPClauseWithPreInit { SourceLocation LParenLoc; /// Modifiers for 'dyn_groupprivate' clause. - enum { FIRST, SECOND, NUM_MODIFIERS }; - OpenMPDynGroupprivateClauseModifier Modifiers[NUM_MODIFIERS]; + enum { SIMPLE, FALLBACK, NUM_MODIFIERS }; + unsigned Modifiers[NUM_MODIFIERS]; /// Locations of modifiers. SourceLocation ModifiersLoc[NUM_MODIFIERS]; @@ -9793,37 +9793,26 @@ class OMPDynGroupprivateClause : public OMPClause, public OMPClauseWithPreInit { /// Set the first dyn_groupprivate modifier. /// /// \param M The modifier. - void setFirstDynGroupprivateModifier(OpenMPDynGroupprivateClauseModifier M) { - Modifiers[FIRST] = M; + void setDynGroupprivateModifier(OpenMPDynGroupprivateClauseModifier M) { + Modifiers[SIMPLE] = M; } /// Set the second dyn_groupprivate modifier. /// /// \param M The modifier. - void setSecondDynGroupprivateModifier(OpenMPDynGroupprivateClauseModifier M) { - Modifiers[SECOND] = M; + void setDynGroupprivateFallbackModifier( + OpenMPDynGroupprivateClauseFallbackModifier M) { + Modifiers[FALLBACK] = M; } /// Set location of the first dyn_groupprivate modifier. - void setFirstDynGroupprivateModifierLoc(SourceLocation Loc) { - ModifiersLoc[FIRST] = Loc; + void setDynGroupprivateModifierLoc(SourceLocation Loc) { + ModifiersLoc[SIMPLE] = Loc; } /// Set location of the second dyn_groupprivate modifier. - void setSecondDynGroupprivateModifierLoc(SourceLocation Loc) { - ModifiersLoc[SECOND] = Loc; - } - - /// Set dyn_groupprivate modifier location. - /// - /// \param M The modifier location. - void setDynGroupprivateModifer(OpenMPDynGroupprivateClauseModifier M) { - if (Modifiers[FIRST] == OMPC_DYN_GROUPPRIVATE_unknown) - Modifiers[FIRST] = M; - else { - assert(Modifiers[SECOND] == OMPC_DYN_GROUPPRIVATE_unknown); - Modifiers[SECOND] = M; - } + void setDynGroupprivateFallbackModifierLoc(SourceLocation Loc) { + ModifiersLoc[FALLBACK] = Loc; } /// Sets the location of '('. @@ -9852,15 +9841,15 @@ class OMPDynGroupprivateClause : public OMPClause, public OMPClauseWithPreInit { OpenMPDirectiveKind CaptureRegion, OpenMPDynGroupprivateClauseModifier M1, SourceLocation M1Loc, - OpenMPDynGroupprivateClauseModifier M2, + OpenMPDynGroupprivateClauseFallbackModifier M2, SourceLocation M2Loc) : OMPClause(llvm::omp::OMPC_dyn_groupprivate, StartLoc, EndLoc), OMPClauseWithPreInit(this), LParenLoc(LParenLoc), Size(Size) { setPreInitStmt(HelperSize, CaptureRegion); - Modifiers[FIRST] = M1; - Modifiers[SECOND] = M2; - ModifiersLoc[FIRST] = M1Loc; - ModifiersLoc[SECOND] = M2Loc; + Modifiers[SIMPLE] = M1; + Modifiers[FALLBACK] = M2; + ModifiersLoc[SIMPLE] = M1Loc; + ModifiersLoc[FALLBACK] = M2Loc; } /// Build an empty clause. @@ -9868,31 +9857,33 @@ class OMPDynGroupprivateClause : public OMPClause, public OMPClauseWithPreInit { : OMPClause(llvm::omp::OMPC_dyn_groupprivate, SourceLocation(), SourceLocation()), OMPClauseWithPreInit(this) { - Modifiers[FIRST] = OMPC_DYN_GROUPPRIVATE_unknown; - Modifiers[SECOND] = OMPC_DYN_GROUPPRIVATE_unknown; + Modifiers[SIMPLE] = OMPC_DYN_GROUPPRIVATE_unknown; + Modifiers[FALLBACK] = OMPC_DYN_GROUPPRIVATE_FALLBACK_unknown; } /// Get the first modifier of the clause. - OpenMPDynGroupprivateClauseModifier getFirstDynGroupprivateModifier() const { - return Modifiers[FIRST]; + OpenMPDynGroupprivateClauseModifier getDynGroupprivateModifier() const { + return static_cast(Modifiers[SIMPLE]); } /// Get the second modifier of the clause. - OpenMPDynGroupprivateClauseModifier getSecondDynGroupprivateModifier() const { - return Modifiers[SECOND]; + OpenMPDynGroupprivateClauseFallbackModifier + getDynGroupprivateFallbackModifier() const { + return static_cast( + Modifiers[FALLBACK]); } /// Get location of '('. SourceLocation getLParenLoc() { return LParenLoc; } /// Get the first modifier location. - SourceLocation getFirstDynGroupprivateModifierLoc() const { - return ModifiersLoc[FIRST]; + SourceLocation getDynGroupprivateModifierLoc() const { + return ModifiersLoc[SIMPLE]; } /// Get the second modifier location. - SourceLocation getSecondDynGroupprivateModifierLoc() const { - return ModifiersLoc[SECOND]; + SourceLocation getDynGroupprivateFallbackModifierLoc() const { + return ModifiersLoc[FALLBACK]; } /// Get size. diff --git a/clang/include/clang/Basic/DiagnosticSemaKinds.td b/clang/include/clang/Basic/DiagnosticSemaKinds.td index 47f1e43814fbb..0a66ab7cc3e53 100644 --- a/clang/include/clang/Basic/DiagnosticSemaKinds.td +++ b/clang/include/clang/Basic/DiagnosticSemaKinds.td @@ -12058,7 +12058,7 @@ def err_omp_unexpected_schedule_modifier : Error< "modifier '%0' cannot be used along with modifier '%1'">; def err_omp_schedule_nonmonotonic_static : Error< "'nonmonotonic' modifier can only be specified with 'dynamic' or 'guided' schedule kind">; -def err_omp_unexpected_dyn_groupprivate_modifier +def err_omp_incompatible_dyn_groupprivate_modifier : Error<"modifier '%0' cannot be used along with modifier '%1' in " "dyn_groupprivate">; def err_omp_simple_clause_incompatible_with_ordered : Error< diff --git a/clang/include/clang/Basic/OpenMPKinds.def b/clang/include/clang/Basic/OpenMPKinds.def index 3321e19cae9b1..3dd919cf9d2d8 100644 --- a/clang/include/clang/Basic/OpenMPKinds.def +++ b/clang/include/clang/Basic/OpenMPKinds.def @@ -86,6 +86,9 @@ #ifndef OPENMP_DYN_GROUPPRIVATE_MODIFIER #define OPENMP_DYN_GROUPPRIVATE_MODIFIER(Name) #endif +#ifndef OPENMP_DYN_GROUPPRIVATE_FALLBACK_MODIFIER +#define OPENMP_DYN_GROUPPRIVATE_FALLBACK_MODIFIER(Name) +#endif #ifndef OPENMP_NUMTASKS_MODIFIER #define OPENMP_NUMTASKS_MODIFIER(Name) #endif @@ -232,8 +235,11 @@ OPENMP_GRAINSIZE_MODIFIER(strict) // Modifiers for the 'dyn_groupprivate' clause. OPENMP_DYN_GROUPPRIVATE_MODIFIER(cgroup) -OPENMP_DYN_GROUPPRIVATE_MODIFIER(strict) -OPENMP_DYN_GROUPPRIVATE_MODIFIER(fallback) + +// Fallback modifiers for the 'dyn_groupprivate' clause. +OPENMP_DYN_GROUPPRIVATE_FALLBACK_MODIFIER(abort) +OPENMP_DYN_GROUPPRIVATE_FALLBACK_MODIFIER(null) +OPENMP_DYN_GROUPPRIVATE_FALLBACK_MODIFIER(default_mem) // Modifiers for the 'num_tasks' clause. OPENMP_NUMTASKS_MODIFIER(strict) @@ -254,6 +260,7 @@ OPENMP_DOACROSS_MODIFIER(source_omp_cur_iteration) #undef OPENMP_NUMTASKS_MODIFIER #undef OPENMP_NUMTHREADS_MODIFIER #undef OPENMP_DYN_GROUPPRIVATE_MODIFIER +#undef OPENMP_DYN_GROUPPRIVATE_FALLBACK_MODIFIER #undef OPENMP_GRAINSIZE_MODIFIER #undef OPENMP_BIND_KIND #undef OPENMP_ADJUST_ARGS_KIND diff --git a/clang/include/clang/Basic/OpenMPKinds.h b/clang/include/clang/Basic/OpenMPKinds.h index 3e164bf1adf22..6f9f30aef4311 100644 --- a/clang/include/clang/Basic/OpenMPKinds.h +++ b/clang/include/clang/Basic/OpenMPKinds.h @@ -223,9 +223,13 @@ enum OpenMPDynGroupprivateClauseModifier { OMPC_DYN_GROUPPRIVATE_unknown }; -/// Number of allowed dyn_groupprivate-modifiers. -static constexpr unsigned NumberOfOMPDynGroupprivateClauseModifiers = - OMPC_DYN_GROUPPRIVATE_unknown; +enum OpenMPDynGroupprivateClauseFallbackModifier { + OMPC_DYN_GROUPPRIVATE_FALLBACK_unknown = OMPC_DYN_GROUPPRIVATE_unknown, +#define OPENMP_DYN_GROUPPRIVATE_FALLBACK_MODIFIER(Name) \ + OMPC_DYN_GROUPPRIVATE_FALLBACK_##Name, +#include "clang/Basic/OpenMPKinds.def" + OMPC_DYN_GROUPPRIVATE_FALLBACK_last +}; enum OpenMPNumTasksClauseModifier { #define OPENMP_NUMTASKS_MODIFIER(Name) OMPC_NUMTASKS_##Name, diff --git a/clang/include/clang/Sema/SemaOpenMP.h b/clang/include/clang/Sema/SemaOpenMP.h index 3b161ff3c7d45..b44bc5ff8d71a 100644 --- a/clang/include/clang/Sema/SemaOpenMP.h +++ b/clang/include/clang/Sema/SemaOpenMP.h @@ -1388,7 +1388,7 @@ class SemaOpenMP : public SemaBase { /// Called on a well-formed 'dyn_groupprivate' clause. OMPClause *ActOnOpenMPDynGroupprivateClause( OpenMPDynGroupprivateClauseModifier M1, - OpenMPDynGroupprivateClauseModifier M2, Expr *Size, + OpenMPDynGroupprivateClauseFallbackModifier M2, Expr *Size, SourceLocation StartLoc, SourceLocation LParenLoc, SourceLocation M1Loc, SourceLocation M2Loc, SourceLocation EndLoc); diff --git a/clang/lib/AST/OpenMPClause.cpp b/clang/lib/AST/OpenMPClause.cpp index 8f694bd857b27..749e9d2a7cc34 100644 --- a/clang/lib/AST/OpenMPClause.cpp +++ b/clang/lib/AST/OpenMPClause.cpp @@ -2726,15 +2726,14 @@ void OMPClausePrinter::VisitOMPXDynCGroupMemClause( void OMPClausePrinter::VisitOMPDynGroupprivateClause( OMPDynGroupprivateClause *Node) { OS << "dyn_groupprivate("; - if (Node->getFirstDynGroupprivateModifier() != - OMPC_DYN_GROUPPRIVATE_unknown) { - OS << getOpenMPSimpleClauseTypeName( - OMPC_dyn_groupprivate, Node->getFirstDynGroupprivateModifier()); - if (Node->getSecondDynGroupprivateModifier() != - OMPC_DYN_GROUPPRIVATE_unknown) { + if (Node->getDynGroupprivateModifier() != OMPC_DYN_GROUPPRIVATE_unknown) { + OS << getOpenMPSimpleClauseTypeName(OMPC_dyn_groupprivate, + Node->getDynGroupprivateModifier()); + if (Node->getDynGroupprivateFallbackModifier() != + OMPC_DYN_GROUPPRIVATE_FALLBACK_unknown) { OS << ", "; OS << getOpenMPSimpleClauseTypeName( - OMPC_dyn_groupprivate, Node->getSecondDynGroupprivateModifier()); + OMPC_dyn_groupprivate, Node->getDynGroupprivateFallbackModifier()); } OS << ": "; } diff --git a/clang/lib/Basic/OpenMPKinds.cpp b/clang/lib/Basic/OpenMPKinds.cpp index 59baab0da1655..eb2888f448036 100644 --- a/clang/lib/Basic/OpenMPKinds.cpp +++ b/clang/lib/Basic/OpenMPKinds.cpp @@ -175,6 +175,9 @@ unsigned clang::getOpenMPSimpleClauseType(OpenMPClauseKind Kind, StringRef Str, return llvm::StringSwitch(Str) #define OPENMP_DYN_GROUPPRIVATE_MODIFIER(Name) \ .Case(#Name, OMPC_DYN_GROUPPRIVATE_##Name) +#define OPENMP_DYN_GROUPPRIVATE_FALLBACK_MODIFIER(Name) \ + .Case(#Name, OMPC_DYN_GROUPPRIVATE_FALLBACK_##Name) \ + .Case("fallback(" #Name ")", OMPC_DYN_GROUPPRIVATE_FALLBACK_##Name) #include "clang/Basic/OpenMPKinds.def" .Default(OMPC_DYN_GROUPPRIVATE_unknown); } @@ -518,10 +521,14 @@ const char *clang::getOpenMPSimpleClauseTypeName(OpenMPClauseKind Kind, case OMPC_dyn_groupprivate: switch (Type) { case OMPC_DYN_GROUPPRIVATE_unknown: + case OMPC_DYN_GROUPPRIVATE_FALLBACK_last: return "unknown"; #define OPENMP_DYN_GROUPPRIVATE_MODIFIER(Name) \ case OMPC_DYN_GROUPPRIVATE_##Name: \ return #Name; +#define OPENMP_DYN_GROUPPRIVATE_FALLBACK_MODIFIER(Name) \ + case OMPC_DYN_GROUPPRIVATE_FALLBACK_##Name: \ + return "fallback(" #Name ")"; #include "clang/Basic/OpenMPKinds.def" } llvm_unreachable("Invalid OpenMP 'dyn_groupprivate' clause modifier"); diff --git a/clang/lib/Parse/ParseOpenMP.cpp b/clang/lib/Parse/ParseOpenMP.cpp index c7e6a6267b453..115ed74227175 100644 --- a/clang/lib/Parse/ParseOpenMP.cpp +++ b/clang/lib/Parse/ParseOpenMP.cpp @@ -3837,31 +3837,72 @@ OMPClause *Parser::ParseOpenMPSingleExprWithArgClause(OpenMPDirectiveKind DKind, KLoc.emplace_back(); } } else if (Kind == OMPC_dyn_groupprivate) { - enum { Modifier1, Modifier2, NumberOfElements }; - Arg.resize(NumberOfElements); - KLoc.resize(NumberOfElements); - Arg[Modifier1] = OMPC_DYN_GROUPPRIVATE_unknown; - Arg[Modifier2] = OMPC_DYN_GROUPPRIVATE_unknown; - unsigned Modifier = getOpenMPSimpleClauseType( - Kind, Tok.isAnnotation() ? "" : PP.getSpelling(Tok), getLangOpts()); + enum { SimpleModifier, ComplexModifier, NumberOfModifiers }; + Arg.resize(NumberOfModifiers); + KLoc.resize(NumberOfModifiers); + Arg[SimpleModifier] = OMPC_DYN_GROUPPRIVATE_unknown; + Arg[ComplexModifier] = OMPC_DYN_GROUPPRIVATE_FALLBACK_unknown; + + auto consumeModifier = [&]() { + unsigned Type = NumberOfModifiers; + unsigned Modifier; + SourceLocation Loc; + if (PP.getSpelling(Tok) == "fallback" && NextToken().is(tok::l_paren)) { + ConsumeToken(); + BalancedDelimiterTracker ParenT(*this, tok::l_paren, tok::r_paren); + ParenT.consumeOpen(); - if (Modifier < OMPC_DYN_GROUPPRIVATE_unknown) { - // Parse 'modifier' - Arg[Modifier1] = Modifier; - KLoc[Modifier1] = Tok.getLocation(); - if (Tok.isNot(tok::r_paren) && Tok.isNot(tok::comma) && - Tok.isNot(tok::annot_pragma_openmp_end)) - ConsumeAnyToken(); - if (Tok.is(tok::comma)) { - // Parse ',' 'modifier' - ConsumeAnyToken(); Modifier = getOpenMPSimpleClauseType( Kind, Tok.isAnnotation() ? "" : PP.getSpelling(Tok), getLangOpts()); - Arg[Modifier2] = Modifier; - KLoc[Modifier2] = Tok.getLocation(); + if (Modifier <= OMPC_DYN_GROUPPRIVATE_FALLBACK_unknown || + Modifier >= OMPC_DYN_GROUPPRIVATE_FALLBACK_last) { + Diag(Tok.getLocation(), diag::err_expected) + << "'abort', 'null' or 'default_mem' in fallback modifier"; + SkipUntil(tok::r_paren); + return std::make_tuple(Type, Modifier, Loc); + } + Type = ComplexModifier; + Loc = Tok.getLocation(); if (Tok.isNot(tok::r_paren) && Tok.isNot(tok::comma) && Tok.isNot(tok::annot_pragma_openmp_end)) ConsumeAnyToken(); + ParenT.consumeClose(); + } else { + Modifier = getOpenMPSimpleClauseType( + Kind, Tok.isAnnotation() ? "" : PP.getSpelling(Tok), getLangOpts()); + if (Modifier < OMPC_DYN_GROUPPRIVATE_unknown) { + Type = SimpleModifier; + Loc = Tok.getLocation(); + if (Tok.isNot(tok::r_paren) && Tok.isNot(tok::comma) && + Tok.isNot(tok::annot_pragma_openmp_end)) + ConsumeAnyToken(); + } + } + return std::make_tuple(Type, Modifier, Loc); + }; + + auto saveModifier = [&](unsigned Type, unsigned Modifier, + SourceLocation Loc) { + assert(Type < NumberOfModifiers); + if (!KLoc[Type].isValid()) { + Arg[Type] = Modifier; + KLoc[Type] = Loc; + } else + Diag(Loc, diag::err_omp_incompatible_dyn_groupprivate_modifier) + << getOpenMPSimpleClauseTypeName(OMPC_dyn_groupprivate, Modifier) + << getOpenMPSimpleClauseTypeName(OMPC_dyn_groupprivate, Arg[Type]); + }; + + // Parse 'modifier' + auto [Type1, Mod1, Loc1] = consumeModifier(); + if (Type1 < NumberOfModifiers) { + saveModifier(Type1, Mod1, Loc1); + if (Tok.is(tok::comma)) { + // Parse ',' 'modifier' + ConsumeAnyToken(); + auto [Type2, Mod2, Loc2] = consumeModifier(); + if (Type2 < NumberOfModifiers) + saveModifier(Type2, Mod2, Loc2); } // Parse ':' if (Tok.is(tok::colon)) diff --git a/clang/lib/Sema/SemaOpenMP.cpp b/clang/lib/Sema/SemaOpenMP.cpp index 67d55acca94df..2a5bab5e22355 100644 --- a/clang/lib/Sema/SemaOpenMP.cpp +++ b/clang/lib/Sema/SemaOpenMP.cpp @@ -16731,7 +16731,8 @@ OMPClause *SemaOpenMP::ActOnOpenMPSingleExprWithArgClause( ArgumentLoc.size() == NumberOfElements); Res = ActOnOpenMPDynGroupprivateClause( static_cast(Argument[Modifier1]), - static_cast(Argument[Modifier2]), + static_cast( + Argument[Modifier2]), Expr, StartLoc, LParenLoc, ArgumentLoc[Modifier1], ArgumentLoc[Modifier2], EndLoc); } break; @@ -24179,12 +24180,12 @@ OMPClause *SemaOpenMP::ActOnOpenMPXDynCGroupMemClause(Expr *Size, OMPClause *SemaOpenMP::ActOnOpenMPDynGroupprivateClause( OpenMPDynGroupprivateClauseModifier M1, - OpenMPDynGroupprivateClauseModifier M2, Expr *Size, SourceLocation StartLoc, - SourceLocation LParenLoc, SourceLocation M1Loc, SourceLocation M2Loc, - SourceLocation EndLoc) { + OpenMPDynGroupprivateClauseFallbackModifier M2, Expr *Size, + SourceLocation StartLoc, SourceLocation LParenLoc, SourceLocation M1Loc, + SourceLocation M2Loc, SourceLocation EndLoc) { if ((M1Loc.isValid() && M1 == OMPC_DYN_GROUPPRIVATE_unknown) || - (M2Loc.isValid() && M2 == OMPC_DYN_GROUPPRIVATE_unknown)) { + (M2Loc.isValid() && M2 == OMPC_DYN_GROUPPRIVATE_FALLBACK_unknown)) { std::string Values = getListOfPossibleValues( OMPC_dyn_groupprivate, /*First=*/0, OMPC_DYN_GROUPPRIVATE_unknown); Diag((M1Loc.isValid() && M1 == OMPC_DYN_GROUPPRIVATE_unknown) ? M1Loc @@ -24194,18 +24195,6 @@ OMPClause *SemaOpenMP::ActOnOpenMPDynGroupprivateClause( return nullptr; } - if ((M1Loc.isValid() && M2Loc.isValid() && M1 == M2) || - (M1 == OMPC_DYN_GROUPPRIVATE_strict && - M2 == OMPC_DYN_GROUPPRIVATE_fallback) || - (M1 == OMPC_DYN_GROUPPRIVATE_fallback && - M2 == OMPC_DYN_GROUPPRIVATE_strict)) { - - Diag(M2Loc, diag::err_omp_unexpected_dyn_groupprivate_modifier) - << getOpenMPSimpleClauseTypeName(OMPC_dyn_groupprivate, M2) - << getOpenMPSimpleClauseTypeName(OMPC_dyn_groupprivate, M1); - return nullptr; - } - Expr *ValExpr = Size; Stmt *HelperValStmt = nullptr; diff --git a/clang/lib/Sema/TreeTransform.h b/clang/lib/Sema/TreeTransform.h index 3da81923890ae..e24653069e313 100644 --- a/clang/lib/Sema/TreeTransform.h +++ b/clang/lib/Sema/TreeTransform.h @@ -2465,7 +2465,7 @@ class TreeTransform { /// Subclasses may override this routine to provide different behavior. OMPClause *RebuildOMPDynGroupprivateClause( OpenMPDynGroupprivateClauseModifier M1, - OpenMPDynGroupprivateClauseModifier M2, Expr *Size, + OpenMPDynGroupprivateClauseFallbackModifier M2, Expr *Size, SourceLocation StartLoc, SourceLocation LParenLoc, SourceLocation M1Loc, SourceLocation M2Loc, SourceLocation EndLoc) { return getSema().OpenMP().ActOnOpenMPDynGroupprivateClause( @@ -11727,10 +11727,10 @@ OMPClause *TreeTransform::TransformOMPDynGroupprivateClause( if (Size.isInvalid()) return nullptr; return getDerived().RebuildOMPDynGroupprivateClause( - C->getFirstDynGroupprivateModifier(), - C->getSecondDynGroupprivateModifier(), Size.get(), C->getBeginLoc(), - C->getLParenLoc(), C->getFirstDynGroupprivateModifierLoc(), - C->getSecondDynGroupprivateModifierLoc(), C->getEndLoc()); + C->getDynGroupprivateModifier(), C->getDynGroupprivateFallbackModifier(), + Size.get(), C->getBeginLoc(), C->getLParenLoc(), + C->getDynGroupprivateModifierLoc(), + C->getDynGroupprivateFallbackModifierLoc(), C->getEndLoc()); } template diff --git a/clang/lib/Serialization/ASTReader.cpp b/clang/lib/Serialization/ASTReader.cpp index 61fd0db41c395..f82899beb3194 100644 --- a/clang/lib/Serialization/ASTReader.cpp +++ b/clang/lib/Serialization/ASTReader.cpp @@ -12714,14 +12714,15 @@ void OMPClauseReader::VisitOMPXDynCGroupMemClause(OMPXDynCGroupMemClause *C) { void OMPClauseReader::VisitOMPDynGroupprivateClause( OMPDynGroupprivateClause *C) { VisitOMPClauseWithPreInit(C); - C->setFirstDynGroupprivateModifier( - static_cast(Record.readInt())); - C->setSecondDynGroupprivateModifier( + C->setDynGroupprivateModifier( static_cast(Record.readInt())); + C->setDynGroupprivateFallbackModifier( + static_cast( + Record.readInt())); C->setSize(Record.readSubExpr()); C->setLParenLoc(Record.readSourceLocation()); - C->setFirstDynGroupprivateModifierLoc(Record.readSourceLocation()); - C->setSecondDynGroupprivateModifierLoc(Record.readSourceLocation()); + C->setDynGroupprivateModifierLoc(Record.readSourceLocation()); + C->setDynGroupprivateFallbackModifierLoc(Record.readSourceLocation()); } void OMPClauseReader::VisitOMPDoacrossClause(OMPDoacrossClause *C) { diff --git a/clang/lib/Serialization/ASTWriter.cpp b/clang/lib/Serialization/ASTWriter.cpp index 24fc0ceebb055..f298f03683e66 100644 --- a/clang/lib/Serialization/ASTWriter.cpp +++ b/clang/lib/Serialization/ASTWriter.cpp @@ -8618,12 +8618,12 @@ void OMPClauseWriter::VisitOMPXDynCGroupMemClause(OMPXDynCGroupMemClause *C) { void OMPClauseWriter::VisitOMPDynGroupprivateClause( OMPDynGroupprivateClause *C) { VisitOMPClauseWithPreInit(C); - Record.push_back(C->getFirstDynGroupprivateModifier()); - Record.push_back(C->getSecondDynGroupprivateModifier()); + Record.push_back(C->getDynGroupprivateModifier()); + Record.push_back(C->getDynGroupprivateFallbackModifier()); Record.AddStmt(C->getSize()); Record.AddSourceLocation(C->getLParenLoc()); - Record.AddSourceLocation(C->getFirstDynGroupprivateModifierLoc()); - Record.AddSourceLocation(C->getSecondDynGroupprivateModifierLoc()); + Record.AddSourceLocation(C->getDynGroupprivateModifierLoc()); + Record.AddSourceLocation(C->getDynGroupprivateFallbackModifierLoc()); } void OMPClauseWriter::VisitOMPDoacrossClause(OMPDoacrossClause *C) { diff --git a/clang/test/OpenMP/target_dyn_groupprivate_messages.cpp b/clang/test/OpenMP/target_dyn_groupprivate_messages.cpp index f924d2bb45eaa..385bd5e89829d 100644 --- a/clang/test/OpenMP/target_dyn_groupprivate_messages.cpp +++ b/clang/test/OpenMP/target_dyn_groupprivate_messages.cpp @@ -73,9 +73,13 @@ int z; foo(); #pragma omp target dyn_groupprivate(cgroup,cgroup: argc) // expected-error {{modifier 'cgroup' cannot be used along with modifier 'cgroup' in dyn_groupprivate}} foo(); - #pragma omp target dyn_groupprivate(fallback,strict: argc) // expected-error {{modifier 'strict' cannot be used along with modifier 'fallback' in dyn_groupprivate}} + #pragma omp target dyn_groupprivate(fallback(default_mem),fallback(abort): argc) // expected-error {{modifier 'fallback(abort)' cannot be used along with modifier 'fallback(default_mem)' in dyn_groupprivate}} foo(); - #pragma omp target dyn_groupprivate(strict,fallback: argc) // expected-error {{modifier 'fallback' cannot be used along with modifier 'strict' in dyn_groupprivate}} + #pragma omp target dyn_groupprivate(fallback(abort),fallback(null): argc) // expected-error {{modifier 'fallback(null)' cannot be used along with modifier 'fallback(abort)' in dyn_groupprivate}} + foo(); + #pragma omp target dyn_groupprivate(fallback(cgroup): argc) // expected-error {{expected 'abort', 'null' or 'default_mem' in fallback modifier}} expected-error {{expected expression}} expected-error {{expected ')'}} expected-note {{to match this '('}} + foo(); + #pragma omp target dyn_groupprivate(fallback(): argc) // expected-error {{expected 'abort', 'null' or 'default_mem' in fallback modifier}} expected-error {{expected expression}} expected-error {{expected ')'}} expected-note {{to match this '('}} foo(); #pragma omp target dyn_groupprivate(: argc) // expected-error {{expected ')'}} expected-error {{expected expression}} expected-note {{to match this '('}} foo(); diff --git a/clang/test/OpenMP/target_teams_dyn_groupprivate_messages.cpp b/clang/test/OpenMP/target_teams_dyn_groupprivate_messages.cpp index d05bb433eab1c..ac2cc0dde5073 100644 --- a/clang/test/OpenMP/target_teams_dyn_groupprivate_messages.cpp +++ b/clang/test/OpenMP/target_teams_dyn_groupprivate_messages.cpp @@ -73,9 +73,13 @@ int z; foo(); #pragma omp target teams dyn_groupprivate(cgroup,cgroup: argc) // expected-error {{modifier 'cgroup' cannot be used along with modifier 'cgroup' in dyn_groupprivate}} foo(); - #pragma omp target teams dyn_groupprivate(fallback,strict: argc) // expected-error {{modifier 'strict' cannot be used along with modifier 'fallback' in dyn_groupprivate}} + #pragma omp target dyn_groupprivate(fallback(default_mem),fallback(abort): argc) // expected-error {{modifier 'fallback(abort)' cannot be used along with modifier 'fallback(default_mem)' in dyn_groupprivate}} foo(); - #pragma omp target teams dyn_groupprivate(strict,fallback: argc) // expected-error {{modifier 'fallback' cannot be used along with modifier 'strict' in dyn_groupprivate}} + #pragma omp target dyn_groupprivate(fallback(abort),fallback(null): argc) // expected-error {{modifier 'fallback(null)' cannot be used along with modifier 'fallback(abort)' in dyn_groupprivate}} + foo(); + #pragma omp target dyn_groupprivate(fallback(cgroup): argc) // expected-error {{expected 'abort', 'null' or 'default_mem' in fallback modifier}} expected-error {{expected expression}} expected-error {{expected ')'}} expected-note {{to match this '('}} + foo(); + #pragma omp target dyn_groupprivate(fallback(): argc) // expected-error {{expected 'abort', 'null' or 'default_mem' in fallback modifier}} expected-error {{expected expression}} expected-error {{expected ')'}} expected-note {{to match this '('}} foo(); #pragma omp target teams dyn_groupprivate(: argc) // expected-error {{expected ')'}} expected-error {{expected expression}} expected-note {{to match this '('}} foo(); diff --git a/clang/test/OpenMP/teams_dyn_groupprivate_messages.cpp b/clang/test/OpenMP/teams_dyn_groupprivate_messages.cpp index a55a2a570a1b8..701ebfb43eec6 100644 --- a/clang/test/OpenMP/teams_dyn_groupprivate_messages.cpp +++ b/clang/test/OpenMP/teams_dyn_groupprivate_messages.cpp @@ -73,9 +73,13 @@ int z; foo(); #pragma omp teams dyn_groupprivate(cgroup,cgroup: argc) // expected-error {{modifier 'cgroup' cannot be used along with modifier 'cgroup' in dyn_groupprivate}} foo(); - #pragma omp teams dyn_groupprivate(fallback,strict: argc) // expected-error {{modifier 'strict' cannot be used along with modifier 'fallback' in dyn_groupprivate}} + #pragma omp target dyn_groupprivate(fallback(default_mem),fallback(abort): argc) // expected-error {{modifier 'fallback(abort)' cannot be used along with modifier 'fallback(default_mem)' in dyn_groupprivate}} foo(); - #pragma omp teams dyn_groupprivate(strict,fallback: argc) // expected-error {{modifier 'fallback' cannot be used along with modifier 'strict' in dyn_groupprivate}} + #pragma omp target dyn_groupprivate(fallback(abort),fallback(null): argc) // expected-error {{modifier 'fallback(null)' cannot be used along with modifier 'fallback(abort)' in dyn_groupprivate}} + foo(); + #pragma omp target dyn_groupprivate(fallback(cgroup): argc) // expected-error {{expected 'abort', 'null' or 'default_mem' in fallback modifier}} expected-error {{expected expression}} expected-error {{expected ')'}} expected-note {{to match this '('}} + foo(); + #pragma omp target dyn_groupprivate(fallback(): argc) // expected-error {{expected 'abort', 'null' or 'default_mem' in fallback modifier}} expected-error {{expected expression}} expected-error {{expected ')'}} expected-note {{to match this '('}} foo(); #pragma omp teams dyn_groupprivate(: argc) // expected-error {{expected ')'}} expected-error {{expected expression}} expected-note {{to match this '('}} foo(); From 3a2fe705a6c11e484ed232a166fd332634fd8a70 Mon Sep 17 00:00:00 2001 From: Kevin Sala Date: Fri, 24 Oct 2025 18:57:14 -0700 Subject: [PATCH 09/17] Update for fallback complex modifier --- clang/lib/CodeGen/CGOpenMPRuntime.cpp | 26 ++++++++++++++----- .../llvm/Frontend/OpenMP/OMPConstants.h | 10 +++++++ .../llvm/Frontend/OpenMP/OMPIRBuilder.h | 11 ++++---- llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp | 16 ++++++------ 4 files changed, 44 insertions(+), 19 deletions(-) diff --git a/clang/lib/CodeGen/CGOpenMPRuntime.cpp b/clang/lib/CodeGen/CGOpenMPRuntime.cpp index 48168193778b8..2d08e1a04ff03 100644 --- a/clang/lib/CodeGen/CGOpenMPRuntime.cpp +++ b/clang/lib/CodeGen/CGOpenMPRuntime.cpp @@ -10000,10 +10000,10 @@ static llvm::Value *emitDeviceID( return DeviceID; } -static std::pair +static std::pair emitDynCGroupMem(const OMPExecutableDirective &D, CodeGenFunction &CGF) { llvm::Value *DynGP = CGF.Builder.getInt32(0); - bool DynGPFallback = false; + OMPDynGroupprivateFallbackType DynGPFallback; if (auto *DynGPClause = D.getSingleClause()) { CodeGenFunction::RunCleanupsScope DynGPScope(CGF); @@ -10011,10 +10011,22 @@ emitDynCGroupMem(const OMPExecutableDirective &D, CodeGenFunction &CGF) { CGF.EmitScalarExpr(DynGPClause->getSize(), /*IgnoreResultAssign=*/true); DynGP = CGF.Builder.CreateIntCast(DynGPVal, CGF.Int32Ty, /*isSigned=*/false); - DynGPFallback = (DynGPClause->getFirstDynGroupprivateModifier() != - OMPC_DYN_GROUPPRIVATE_strict && - DynGPClause->getSecondDynGroupprivateModifier() != - OMPC_DYN_GROUPPRIVATE_strict); + auto FallbackModifier = DynGPClause->getDynGroupprivateFallbackModifier(); + switch (FallbackModifier) { + case OMPC_DYN_GROUPPRIVATE_FALLBACK_abort: + DynGPFallback = OMPDynGroupprivateFallbackType::Abort; + break; + case OMPC_DYN_GROUPPRIVATE_FALLBACK_null: + DynGPFallback = OMPDynGroupprivateFallbackType::Null; + break; + case OMPC_DYN_GROUPPRIVATE_FALLBACK_default_mem: + case OMPC_DYN_GROUPPRIVATE_FALLBACK_unknown: + // This is the default for dyn_groupprivate. + DynGPFallback = OMPDynGroupprivateFallbackType::DefaultMem; + break; + default: + llvm_unreachable("Unknown fallback modifier for OpenMP dyn_groupprivate"); + } } else if (auto *OMPXDynCGClause = D.getSingleClause()) { CodeGenFunction::RunCleanupsScope DynCGMemScope(CGF); @@ -10022,9 +10034,11 @@ emitDynCGroupMem(const OMPExecutableDirective &D, CodeGenFunction &CGF) { /*IgnoreResultAssign=*/true); DynGP = CGF.Builder.CreateIntCast(DynCGMemVal, CGF.Int32Ty, /*isSigned=*/false); + DynGPFallback = OMPDynGroupprivateFallbackType::Abort; } return {DynGP, DynGPFallback}; } + static void genMapInfoForCaptures( MappableExprsHandler &MEHandler, CodeGenFunction &CGF, const CapturedStmt &CS, llvm::SmallVectorImpl &CapturedVars, diff --git a/llvm/include/llvm/Frontend/OpenMP/OMPConstants.h b/llvm/include/llvm/Frontend/OpenMP/OMPConstants.h index 7bec7e0c6736d..1ac9ac040468c 100644 --- a/llvm/include/llvm/Frontend/OpenMP/OMPConstants.h +++ b/llvm/include/llvm/Frontend/OpenMP/OMPConstants.h @@ -190,6 +190,16 @@ enum class OMPScheduleType { LLVM_MARK_AS_BITMASK_ENUM(/* LargestValue */ ModifierMask) }; +/// The fallback types for the dyn_groupprivate clause. +enum class OMPDynGroupprivateFallbackType : uint64_t { + /// Abort the execution. + Abort = 0, + /// Return null pointer. + Null = 1, + /// Allocate from a implementation defined memory space. + DefaultMem = 2 +}; + // Default OpenMP mapper name suffix. inline constexpr const char *OmpDefaultMapperName = ".omp.default.mapper"; diff --git a/llvm/include/llvm/Frontend/OpenMP/OMPIRBuilder.h b/llvm/include/llvm/Frontend/OpenMP/OMPIRBuilder.h index f20b58409302f..aa370606c6539 100644 --- a/llvm/include/llvm/Frontend/OpenMP/OMPIRBuilder.h +++ b/llvm/include/llvm/Frontend/OpenMP/OMPIRBuilder.h @@ -2449,20 +2449,21 @@ class OpenMPIRBuilder { Value *DynCGGroupMem = nullptr; /// True if the kernel has 'no wait' clause. bool HasNoWait = false; - /// True if the dynamic shared memory may fallback. - bool MayFallbackDynCGroupMem = false; + /// The fallback mechanism for the shared memory. + omp::OMPDynGroupprivateFallbackType DynCGroupMemFallback = + omp::OMPDynGroupprivateFallbackType::Abort; // Constructors for TargetKernelArgs. TargetKernelArgs() {} TargetKernelArgs(unsigned NumTargetItems, TargetDataRTArgs RTArgs, Value *NumIterations, ArrayRef NumTeams, ArrayRef NumThreads, Value *DynCGGroupMem, - bool HasNoWait, bool MayFallbackDynCGroupMem) + bool HasNoWait, + omp::OMPDynGroupprivateFallbackType DynCGroupMemFallback) : NumTargetItems(NumTargetItems), RTArgs(RTArgs), NumIterations(NumIterations), NumTeams(NumTeams), NumThreads(NumThreads), DynCGGroupMem(DynCGGroupMem), - HasNoWait(HasNoWait), - MayFallbackDynCGroupMem(MayFallbackDynCGroupMem) {} + HasNoWait(HasNoWait), DynCGroupMemFallback(DynCGroupMemFallback) {} }; /// Create the kernel args vector used by emitTargetKernel. This function diff --git a/llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp b/llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp index 4cfe038c36eee..9f80e3eb61bfc 100644 --- a/llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp +++ b/llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp @@ -532,11 +532,11 @@ void OpenMPIRBuilder::getKernelArgsVector(TargetKernelArgs &KernelArgs, Value *ZeroArray = Constant::getNullValue(ArrayType::get(Int32Ty, MaxDim)); Value *HasNoWaitFlag = Builder.getInt64(KernelArgs.HasNoWait); - Value *MayFallbackDynCGroupMemFlag = - Builder.getInt64(KernelArgs.MayFallbackDynCGroupMem); - MayFallbackDynCGroupMemFlag = - Builder.CreateShl(MayFallbackDynCGroupMemFlag, 2); - Value *Flags = Builder.CreateOr(HasNoWaitFlag, MayFallbackDynCGroupMemFlag); + + Value *DynCGroupMemFallbackFlag = + Builder.getInt64(static_cast(KernelArgs.DynCGroupMemFallback)); + DynCGroupMemFallbackFlag = Builder.CreateShl(DynCGroupMemFallbackFlag, 2); + Value *Flags = Builder.CreateOr(HasNoWaitFlag, DynCGroupMemFallbackFlag); assert(!KernelArgs.NumTeams.empty() && !KernelArgs.NumThreads.empty()); @@ -8368,9 +8368,9 @@ static void emitTargetCall( // TODO: Use correct DynCGGroupMem Value *DynCGGroupMem = Builder.getInt32(0); - KArgs = OpenMPIRBuilder::TargetKernelArgs(NumTargetItems, RTArgs, TripCount, - NumTeamsC, NumThreadsC, - DynCGGroupMem, HasNoWait, false); + KArgs = OpenMPIRBuilder::TargetKernelArgs( + NumTargetItems, RTArgs, TripCount, NumTeamsC, NumThreadsC, + DynCGGroupMem, HasNoWait, OMPDynGroupprivateFallbackType::Abort); // Assume no error was returned because TaskBodyCB and // EmitTargetCallFallbackCB don't produce any. From 4662a4ffde75d5cef92f94cceb3c83195d883da7 Mon Sep 17 00:00:00 2001 From: Kevin Sala Date: Sun, 26 Oct 2025 20:05:11 -0700 Subject: [PATCH 10/17] Fix parsing --- clang/lib/Parse/ParseOpenMP.cpp | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/clang/lib/Parse/ParseOpenMP.cpp b/clang/lib/Parse/ParseOpenMP.cpp index 71a5cc9950368..0df6a5ae663a6 100644 --- a/clang/lib/Parse/ParseOpenMP.cpp +++ b/clang/lib/Parse/ParseOpenMP.cpp @@ -4020,7 +4020,7 @@ OMPClause *Parser::ParseOpenMPSingleExprWithArgClause(OpenMPDirectiveKind DKind, unsigned Type = NumberOfModifiers; unsigned Modifier; SourceLocation Loc; - if (PP.getSpelling(Tok) == "fallback" && NextToken().is(tok::l_paren)) { + if (!Tok.isAnnotation() && PP.getSpelling(Tok) == "fallback" && NextToken().is(tok::l_paren)) { ConsumeToken(); BalancedDelimiterTracker ParenT(*this, tok::l_paren, tok::r_paren); ParenT.consumeOpen(); From 64e7abc6d6483be586d089544f1d816c4b514feb Mon Sep 17 00:00:00 2001 From: Kevin Sala Date: Sun, 26 Oct 2025 20:10:03 -0700 Subject: [PATCH 11/17] Fix format --- clang/lib/Parse/ParseOpenMP.cpp | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/clang/lib/Parse/ParseOpenMP.cpp b/clang/lib/Parse/ParseOpenMP.cpp index 0df6a5ae663a6..60d332b5d6532 100644 --- a/clang/lib/Parse/ParseOpenMP.cpp +++ b/clang/lib/Parse/ParseOpenMP.cpp @@ -4020,7 +4020,8 @@ OMPClause *Parser::ParseOpenMPSingleExprWithArgClause(OpenMPDirectiveKind DKind, unsigned Type = NumberOfModifiers; unsigned Modifier; SourceLocation Loc; - if (!Tok.isAnnotation() && PP.getSpelling(Tok) == "fallback" && NextToken().is(tok::l_paren)) { + if (!Tok.isAnnotation() && PP.getSpelling(Tok) == "fallback" && + NextToken().is(tok::l_paren)) { ConsumeToken(); BalancedDelimiterTracker ParenT(*this, tok::l_paren, tok::r_paren); ParenT.consumeOpen(); From 7b53c9a678f6d3bed3befd6c524815717b53d268 Mon Sep 17 00:00:00 2001 From: Kevin Sala Date: Mon, 27 Oct 2025 09:02:55 -0700 Subject: [PATCH 12/17] Fix review comments --- clang/lib/Parse/ParseOpenMP.cpp | 15 ++++++++------- 1 file changed, 8 insertions(+), 7 deletions(-) diff --git a/clang/lib/Parse/ParseOpenMP.cpp b/clang/lib/Parse/ParseOpenMP.cpp index 60d332b5d6532..2c7d3abfc1f7a 100644 --- a/clang/lib/Parse/ParseOpenMP.cpp +++ b/clang/lib/Parse/ParseOpenMP.cpp @@ -4016,7 +4016,7 @@ OMPClause *Parser::ParseOpenMPSingleExprWithArgClause(OpenMPDirectiveKind DKind, Arg[SimpleModifier] = OMPC_DYN_GROUPPRIVATE_unknown; Arg[ComplexModifier] = OMPC_DYN_GROUPPRIVATE_FALLBACK_unknown; - auto consumeModifier = [&]() { + auto ConsumeModifier = [&]() { unsigned Type = NumberOfModifiers; unsigned Modifier; SourceLocation Loc; @@ -4055,28 +4055,29 @@ OMPClause *Parser::ParseOpenMPSingleExprWithArgClause(OpenMPDirectiveKind DKind, return std::make_tuple(Type, Modifier, Loc); }; - auto saveModifier = [&](unsigned Type, unsigned Modifier, + auto SaveModifier = [&](unsigned Type, unsigned Modifier, SourceLocation Loc) { assert(Type < NumberOfModifiers); if (!KLoc[Type].isValid()) { Arg[Type] = Modifier; KLoc[Type] = Loc; - } else + } else { Diag(Loc, diag::err_omp_incompatible_dyn_groupprivate_modifier) << getOpenMPSimpleClauseTypeName(OMPC_dyn_groupprivate, Modifier) << getOpenMPSimpleClauseTypeName(OMPC_dyn_groupprivate, Arg[Type]); + } }; // Parse 'modifier' - auto [Type1, Mod1, Loc1] = consumeModifier(); + auto [Type1, Mod1, Loc1] = ConsumeModifier(); if (Type1 < NumberOfModifiers) { - saveModifier(Type1, Mod1, Loc1); + SaveModifier(Type1, Mod1, Loc1); if (Tok.is(tok::comma)) { // Parse ',' 'modifier' ConsumeAnyToken(); - auto [Type2, Mod2, Loc2] = consumeModifier(); + auto [Type2, Mod2, Loc2] = ConsumeModifier(); if (Type2 < NumberOfModifiers) - saveModifier(Type2, Mod2, Loc2); + SaveModifier(Type2, Mod2, Loc2); } // Parse ':' if (Tok.is(tok::colon)) From 79b34f1994c4f19d2c226b09a352370bbadba756 Mon Sep 17 00:00:00 2001 From: Kevin Sala Date: Mon, 27 Oct 2025 09:06:33 -0700 Subject: [PATCH 13/17] Fix more review comments --- clang/lib/Parse/ParseOpenMP.cpp | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/clang/lib/Parse/ParseOpenMP.cpp b/clang/lib/Parse/ParseOpenMP.cpp index 2c7d3abfc1f7a..a44281f370caa 100644 --- a/clang/lib/Parse/ParseOpenMP.cpp +++ b/clang/lib/Parse/ParseOpenMP.cpp @@ -4057,7 +4057,7 @@ OMPClause *Parser::ParseOpenMPSingleExprWithArgClause(OpenMPDirectiveKind DKind, auto SaveModifier = [&](unsigned Type, unsigned Modifier, SourceLocation Loc) { - assert(Type < NumberOfModifiers); + assert(Type < NumberOfModifiers && "Unexpected modifier type"); if (!KLoc[Type].isValid()) { Arg[Type] = Modifier; KLoc[Type] = Loc; From c4905e056722d56d2b096c6e7e30c9580b2a4461 Mon Sep 17 00:00:00 2001 From: Kevin Sala Date: Thu, 30 Oct 2025 14:30:00 -0700 Subject: [PATCH 14/17] Fix initialization of a fallback variable --- clang/lib/CodeGen/CGOpenMPRuntime.cpp | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) diff --git a/clang/lib/CodeGen/CGOpenMPRuntime.cpp b/clang/lib/CodeGen/CGOpenMPRuntime.cpp index 2d08e1a04ff03..74076eb5a9733 100644 --- a/clang/lib/CodeGen/CGOpenMPRuntime.cpp +++ b/clang/lib/CodeGen/CGOpenMPRuntime.cpp @@ -10003,7 +10003,7 @@ static llvm::Value *emitDeviceID( static std::pair emitDynCGroupMem(const OMPExecutableDirective &D, CodeGenFunction &CGF) { llvm::Value *DynGP = CGF.Builder.getInt32(0); - OMPDynGroupprivateFallbackType DynGPFallback; + auto DynGPFallback = OMPDynGroupprivateFallbackType::Abort; if (auto *DynGPClause = D.getSingleClause()) { CodeGenFunction::RunCleanupsScope DynGPScope(CGF); @@ -10034,7 +10034,6 @@ emitDynCGroupMem(const OMPExecutableDirective &D, CodeGenFunction &CGF) { /*IgnoreResultAssign=*/true); DynGP = CGF.Builder.CreateIntCast(DynCGMemVal, CGF.Int32Ty, /*isSigned=*/false); - DynGPFallback = OMPDynGroupprivateFallbackType::Abort; } return {DynGP, DynGPFallback}; } From c439e4412e7ad983b8f923ec202c46d07f3b683c Mon Sep 17 00:00:00 2001 From: Kevin Sala Date: Thu, 30 Oct 2025 15:13:32 -0700 Subject: [PATCH 15/17] Add codegen test for dyn_groupprivate --- .../target_dyn_groupprivate_codegen.cpp | 2633 +++++++++++++++++ 1 file changed, 2633 insertions(+) create mode 100644 clang/test/OpenMP/target_dyn_groupprivate_codegen.cpp diff --git a/clang/test/OpenMP/target_dyn_groupprivate_codegen.cpp b/clang/test/OpenMP/target_dyn_groupprivate_codegen.cpp new file mode 100644 index 0000000000000..758f35d629ace --- /dev/null +++ b/clang/test/OpenMP/target_dyn_groupprivate_codegen.cpp @@ -0,0 +1,2633 @@ +// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --function-signature --include-generated-funcs --replace-value-regex "__omp_offloading_[0-9a-z]+_[0-9a-z]+" "reduction_size[.].+[.]" "pl_cond[.].+[.|,]" --prefix-filecheck-ir-name _ +// Test host codegen. +// RUN: %clang_cc1 -verify -fopenmp -fopenmp-version=61 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK1 +// RUN: %clang_cc1 -fopenmp -fopenmp-version=61 -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s +// RUN: %clang_cc1 -fopenmp -fopenmp-version=61 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK1 +// RUN: %clang_cc1 -verify -fopenmp -fopenmp-version=61 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK3 +// RUN: %clang_cc1 -fopenmp -fopenmp-version=61 -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s +// RUN: %clang_cc1 -fopenmp -fopenmp-version=61 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK3 + +// RUN: %clang_cc1 -verify -fopenmp-simd -fopenmp-version=61 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}" +// RUN: %clang_cc1 -fopenmp-simd -fopenmp-version=61 -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s +// RUN: %clang_cc1 -fopenmp-simd -fopenmp-version=61 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}" +// RUN: %clang_cc1 -verify -fopenmp-simd -fopenmp-version=61 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}" +// RUN: %clang_cc1 -fopenmp-simd -fopenmp-version=61 -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s +// RUN: %clang_cc1 -fopenmp-simd -fopenmp-version=61 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}" + +// Test target codegen - host bc file has to be created first. +// RUN: %clang_cc1 -verify -fopenmp -fopenmp-version=61 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm-bc %s -o %t-ppc-host.bc +// RUN: %clang_cc1 -verify -fopenmp -fopenmp-version=61 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -fopenmp-is-target-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o - | FileCheck %s --check-prefix=CHECK9 +// RUN: %clang_cc1 -fopenmp -fopenmp-version=61 -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -fopenmp-is-target-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o %t %s +// RUN: %clang_cc1 -fopenmp -fopenmp-version=61 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -fopenmp-is-target-device -fopenmp-host-ir-file-path %t-ppc-host.bc -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK9 +// RUN: %clang_cc1 -verify -fopenmp -fopenmp-version=61 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm-bc %s -o %t-x86-host.bc +// RUN: %clang_cc1 -verify -fopenmp -fopenmp-version=61 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -fopenmp-is-target-device -fopenmp-host-ir-file-path %t-x86-host.bc -o - | FileCheck %s --check-prefix=CHECK11 +// RUN: %clang_cc1 -fopenmp -fopenmp-version=61 -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -fopenmp-is-target-device -fopenmp-host-ir-file-path %t-x86-host.bc -o %t %s +// RUN: %clang_cc1 -fopenmp -fopenmp-version=61 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -fopenmp-is-target-device -fopenmp-host-ir-file-path %t-x86-host.bc -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK11 + +// RUN: %clang_cc1 -verify -fopenmp-simd -fopenmp-version=61 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm-bc %s -o %t-ppc-host.bc +// RUN: %clang_cc1 -verify -fopenmp-simd -fopenmp-version=61 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -fopenmp-is-target-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}" +// RUN: %clang_cc1 -fopenmp-simd -fopenmp-version=61 -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -fopenmp-is-target-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o %t %s +// RUN: %clang_cc1 -fopenmp-simd -fopenmp-version=61 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -fopenmp-is-target-device -fopenmp-host-ir-file-path %t-ppc-host.bc -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}" +// RUN: %clang_cc1 -verify -fopenmp-simd -fopenmp-version=61 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm-bc %s -o %t-x86-host.bc +// RUN: %clang_cc1 -verify -fopenmp-simd -fopenmp-version=61 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -fopenmp-is-target-device -fopenmp-host-ir-file-path %t-x86-host.bc -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}" +// RUN: %clang_cc1 -fopenmp-simd -fopenmp-version=61 -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -fopenmp-is-target-device -fopenmp-host-ir-file-path %t-x86-host.bc -o %t %s +// RUN: %clang_cc1 -fopenmp-simd -fopenmp-version=61 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -fopenmp-is-target-device -fopenmp-host-ir-file-path %t-x86-host.bc -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}" + + +// expected-no-diagnostics +#ifndef HEADER +#define HEADER + + + + +// We have 6 target regions + + + +// Check target registration is registered as a Ctor. + + +template +tx ftemplate(int n) { + tx a = 0; + + #pragma omp target teams dyn_groupprivate(tx(20)) + { + } + + short b = 1; + #pragma omp target teams num_teams(b) dyn_groupprivate(1024) + { + a += b; + } + + return a; +} + +static +int fstatic(int n) { + + #pragma omp target teams distribute parallel for simd num_teams(n) dyn_groupprivate(n*32) + for (int i = 0; i < n ; ++i) { + } + + #pragma omp target teams dyn_groupprivate(fallback(default_mem): 32+n) nowait + { + } + + return n+1; +} + +struct S1 { + double a; + + int r1(int n){ + int b = 1; + + #pragma omp target teams dyn_groupprivate(fallback(null): n-b) + { + this->a = (double)b + 1.5; + } + + #pragma omp target dyn_groupprivate(fallback(abort): 1024) + { + this->a = 2.5; + } + + return (int)a; + } +}; + +int bar(int n){ + int a = 0; + + S1 S; + a += S.r1(n); + + a += fstatic(n); + + a += ftemplate(n); + + return a; +} + + + + + + + + + + + + + + + + + + + + + +// Check that the offloading functions are emitted and that the parallel function +// is appropriately guarded. + + + + + + +#endif + +// CHECK1-LABEL: define {{[^@]+}}@_Z3bari +// CHECK1-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0:[0-9]+]] { +// CHECK1-NEXT: entry: +// CHECK1-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4 +// CHECK1-NEXT: [[A:%.*]] = alloca i32, align 4 +// CHECK1-NEXT: [[S:%.*]] = alloca [[STRUCT_S1:%.*]], align 8 +// CHECK1-NEXT: store i32 [[N]], ptr [[N_ADDR]], align 4 +// CHECK1-NEXT: store i32 0, ptr [[A]], align 4 +// CHECK1-NEXT: [[TMP0:%.*]] = load i32, ptr [[N_ADDR]], align 4 +// CHECK1-NEXT: [[CALL:%.*]] = call noundef signext i32 @_ZN2S12r1Ei(ptr noundef nonnull align 8 dereferenceable(8) [[S]], i32 noundef signext [[TMP0]]) +// CHECK1-NEXT: [[TMP1:%.*]] = load i32, ptr [[A]], align 4 +// CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CALL]] +// CHECK1-NEXT: store i32 [[ADD]], ptr [[A]], align 4 +// CHECK1-NEXT: [[TMP2:%.*]] = load i32, ptr [[N_ADDR]], align 4 +// CHECK1-NEXT: [[CALL1:%.*]] = call noundef signext i32 @_ZL7fstatici(i32 noundef signext [[TMP2]]) +// CHECK1-NEXT: [[TMP3:%.*]] = load i32, ptr [[A]], align 4 +// CHECK1-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP3]], [[CALL1]] +// CHECK1-NEXT: store i32 [[ADD2]], ptr [[A]], align 4 +// CHECK1-NEXT: [[TMP4:%.*]] = load i32, ptr [[N_ADDR]], align 4 +// CHECK1-NEXT: [[CALL3:%.*]] = call noundef signext i32 @_Z9ftemplateIiET_i(i32 noundef signext [[TMP4]]) +// CHECK1-NEXT: [[TMP5:%.*]] = load i32, ptr [[A]], align 4 +// CHECK1-NEXT: [[ADD4:%.*]] = add nsw i32 [[TMP5]], [[CALL3]] +// CHECK1-NEXT: store i32 [[ADD4]], ptr [[A]], align 4 +// CHECK1-NEXT: [[TMP6:%.*]] = load i32, ptr [[A]], align 4 +// CHECK1-NEXT: ret i32 [[TMP6]] +// +// +// CHECK1-LABEL: define {{[^@]+}}@_ZN2S12r1Ei +// CHECK1-SAME: (ptr noundef nonnull align 8 dereferenceable(8) [[THIS:%.*]], i32 noundef signext [[N:%.*]]) #[[ATTR0]] comdat { +// CHECK1-NEXT: entry: +// CHECK1-NEXT: [[THIS_ADDR:%.*]] = alloca ptr, align 8 +// CHECK1-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4 +// CHECK1-NEXT: [[B:%.*]] = alloca i32, align 4 +// CHECK1-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4 +// CHECK1-NEXT: [[B_CASTED:%.*]] = alloca i64, align 8 +// CHECK1-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8 +// CHECK1-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x ptr], align 8 +// CHECK1-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x ptr], align 8 +// CHECK1-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x ptr], align 8 +// CHECK1-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8 +// CHECK1-NEXT: [[DOTOFFLOAD_BASEPTRS3:%.*]] = alloca [1 x ptr], align 8 +// CHECK1-NEXT: [[DOTOFFLOAD_PTRS4:%.*]] = alloca [1 x ptr], align 8 +// CHECK1-NEXT: [[DOTOFFLOAD_MAPPERS5:%.*]] = alloca [1 x ptr], align 8 +// CHECK1-NEXT: [[KERNEL_ARGS6:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8 +// CHECK1-NEXT: store ptr [[THIS]], ptr [[THIS_ADDR]], align 8 +// CHECK1-NEXT: store i32 [[N]], ptr [[N_ADDR]], align 4 +// CHECK1-NEXT: [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8 +// CHECK1-NEXT: store i32 1, ptr [[B]], align 4 +// CHECK1-NEXT: [[TMP0:%.*]] = load i32, ptr [[N_ADDR]], align 4 +// CHECK1-NEXT: [[TMP1:%.*]] = load i32, ptr [[B]], align 4 +// CHECK1-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP0]], [[TMP1]] +// CHECK1-NEXT: store i32 [[SUB]], ptr [[DOTCAPTURE_EXPR_]], align 4 +// CHECK1-NEXT: [[TMP2:%.*]] = load i32, ptr [[B]], align 4 +// CHECK1-NEXT: store i32 [[TMP2]], ptr [[B_CASTED]], align 4 +// CHECK1-NEXT: [[TMP3:%.*]] = load i64, ptr [[B_CASTED]], align 8 +// CHECK1-NEXT: [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4 +// CHECK1-NEXT: store i32 [[TMP4]], ptr [[DOTCAPTURE_EXPR__CASTED]], align 4 +// CHECK1-NEXT: [[TMP5:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR__CASTED]], align 8 +// CHECK1-NEXT: [[A:%.*]] = getelementptr inbounds nuw [[STRUCT_S1:%.*]], ptr [[THIS1]], i32 0, i32 0 +// CHECK1-NEXT: [[TMP6:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 +// CHECK1-NEXT: store ptr [[THIS1]], ptr [[TMP6]], align 8 +// CHECK1-NEXT: [[TMP7:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0 +// CHECK1-NEXT: store ptr [[A]], ptr [[TMP7]], align 8 +// CHECK1-NEXT: [[TMP8:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0 +// CHECK1-NEXT: store ptr null, ptr [[TMP8]], align 8 +// CHECK1-NEXT: [[TMP9:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1 +// CHECK1-NEXT: store i64 [[TMP3]], ptr [[TMP9]], align 8 +// CHECK1-NEXT: [[TMP10:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 1 +// CHECK1-NEXT: store i64 [[TMP3]], ptr [[TMP10]], align 8 +// CHECK1-NEXT: [[TMP11:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1 +// CHECK1-NEXT: store ptr null, ptr [[TMP11]], align 8 +// CHECK1-NEXT: [[TMP12:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2 +// CHECK1-NEXT: store i64 [[TMP5]], ptr [[TMP12]], align 8 +// CHECK1-NEXT: [[TMP13:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 2 +// CHECK1-NEXT: store i64 [[TMP5]], ptr [[TMP13]], align 8 +// CHECK1-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2 +// CHECK1-NEXT: store ptr null, ptr [[TMP14]], align 8 +// CHECK1-NEXT: [[TMP15:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 +// CHECK1-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0 +// CHECK1-NEXT: [[TMP17:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4 +// CHECK1-NEXT: [[TMP18:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0 +// CHECK1-NEXT: store i32 3, ptr [[TMP18]], align 4 +// CHECK1-NEXT: [[TMP19:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1 +// CHECK1-NEXT: store i32 3, ptr [[TMP19]], align 4 +// CHECK1-NEXT: [[TMP20:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2 +// CHECK1-NEXT: store ptr [[TMP15]], ptr [[TMP20]], align 8 +// CHECK1-NEXT: [[TMP21:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3 +// CHECK1-NEXT: store ptr [[TMP16]], ptr [[TMP21]], align 8 +// CHECK1-NEXT: [[TMP22:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4 +// CHECK1-NEXT: store ptr @.offload_sizes, ptr [[TMP22]], align 8 +// CHECK1-NEXT: [[TMP23:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5 +// CHECK1-NEXT: store ptr @.offload_maptypes, ptr [[TMP23]], align 8 +// CHECK1-NEXT: [[TMP24:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6 +// CHECK1-NEXT: store ptr null, ptr [[TMP24]], align 8 +// CHECK1-NEXT: [[TMP25:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7 +// CHECK1-NEXT: store ptr null, ptr [[TMP25]], align 8 +// CHECK1-NEXT: [[TMP26:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8 +// CHECK1-NEXT: store i64 0, ptr [[TMP26]], align 8 +// CHECK1-NEXT: [[TMP27:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9 +// CHECK1-NEXT: store i64 4, ptr [[TMP27]], align 8 +// CHECK1-NEXT: [[TMP28:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10 +// CHECK1-NEXT: store [3 x i32] zeroinitializer, ptr [[TMP28]], align 4 +// CHECK1-NEXT: [[TMP29:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11 +// CHECK1-NEXT: store [3 x i32] zeroinitializer, ptr [[TMP29]], align 4 +// CHECK1-NEXT: [[TMP30:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12 +// CHECK1-NEXT: store i32 [[TMP17]], ptr [[TMP30]], align 4 +// CHECK1-NEXT: [[TMP31:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB1:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l88.region_id, ptr [[KERNEL_ARGS]]) +// CHECK1-NEXT: [[TMP32:%.*]] = icmp ne i32 [[TMP31]], 0 +// CHECK1-NEXT: br i1 [[TMP32]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]] +// CHECK1: omp_offload.failed: +// CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l88(ptr [[THIS1]], i64 [[TMP3]], i64 [[TMP5]]) #[[ATTR2:[0-9]+]] +// CHECK1-NEXT: br label [[OMP_OFFLOAD_CONT]] +// CHECK1: omp_offload.cont: +// CHECK1-NEXT: [[A2:%.*]] = getelementptr inbounds nuw [[STRUCT_S1]], ptr [[THIS1]], i32 0, i32 0 +// CHECK1-NEXT: [[TMP33:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0 +// CHECK1-NEXT: store ptr [[THIS1]], ptr [[TMP33]], align 8 +// CHECK1-NEXT: [[TMP34:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS4]], i32 0, i32 0 +// CHECK1-NEXT: store ptr [[A2]], ptr [[TMP34]], align 8 +// CHECK1-NEXT: [[TMP35:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS5]], i64 0, i64 0 +// CHECK1-NEXT: store ptr null, ptr [[TMP35]], align 8 +// CHECK1-NEXT: [[TMP36:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0 +// CHECK1-NEXT: [[TMP37:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS4]], i32 0, i32 0 +// CHECK1-NEXT: [[TMP38:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS6]], i32 0, i32 0 +// CHECK1-NEXT: store i32 3, ptr [[TMP38]], align 4 +// CHECK1-NEXT: [[TMP39:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS6]], i32 0, i32 1 +// CHECK1-NEXT: store i32 1, ptr [[TMP39]], align 4 +// CHECK1-NEXT: [[TMP40:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS6]], i32 0, i32 2 +// CHECK1-NEXT: store ptr [[TMP36]], ptr [[TMP40]], align 8 +// CHECK1-NEXT: [[TMP41:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS6]], i32 0, i32 3 +// CHECK1-NEXT: store ptr [[TMP37]], ptr [[TMP41]], align 8 +// CHECK1-NEXT: [[TMP42:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS6]], i32 0, i32 4 +// CHECK1-NEXT: store ptr @.offload_sizes.1, ptr [[TMP42]], align 8 +// CHECK1-NEXT: [[TMP43:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS6]], i32 0, i32 5 +// CHECK1-NEXT: store ptr @.offload_maptypes.2, ptr [[TMP43]], align 8 +// CHECK1-NEXT: [[TMP44:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS6]], i32 0, i32 6 +// CHECK1-NEXT: store ptr null, ptr [[TMP44]], align 8 +// CHECK1-NEXT: [[TMP45:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS6]], i32 0, i32 7 +// CHECK1-NEXT: store ptr null, ptr [[TMP45]], align 8 +// CHECK1-NEXT: [[TMP46:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS6]], i32 0, i32 8 +// CHECK1-NEXT: store i64 0, ptr [[TMP46]], align 8 +// CHECK1-NEXT: [[TMP47:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS6]], i32 0, i32 9 +// CHECK1-NEXT: store i64 0, ptr [[TMP47]], align 8 +// CHECK1-NEXT: [[TMP48:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS6]], i32 0, i32 10 +// CHECK1-NEXT: store [3 x i32] [i32 -1, i32 0, i32 0], ptr [[TMP48]], align 4 +// CHECK1-NEXT: [[TMP49:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS6]], i32 0, i32 11 +// CHECK1-NEXT: store [3 x i32] zeroinitializer, ptr [[TMP49]], align 4 +// CHECK1-NEXT: [[TMP50:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS6]], i32 0, i32 12 +// CHECK1-NEXT: store i32 1024, ptr [[TMP50]], align 4 +// CHECK1-NEXT: [[TMP51:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB1]], i64 -1, i32 -1, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l93.region_id, ptr [[KERNEL_ARGS6]]) +// CHECK1-NEXT: [[TMP52:%.*]] = icmp ne i32 [[TMP51]], 0 +// CHECK1-NEXT: br i1 [[TMP52]], label [[OMP_OFFLOAD_FAILED7:%.*]], label [[OMP_OFFLOAD_CONT8:%.*]] +// CHECK1: omp_offload.failed7: +// CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l93(ptr [[THIS1]]) #[[ATTR2]] +// CHECK1-NEXT: br label [[OMP_OFFLOAD_CONT8]] +// CHECK1: omp_offload.cont8: +// CHECK1-NEXT: [[A9:%.*]] = getelementptr inbounds nuw [[STRUCT_S1]], ptr [[THIS1]], i32 0, i32 0 +// CHECK1-NEXT: [[TMP53:%.*]] = load double, ptr [[A9]], align 8 +// CHECK1-NEXT: [[CONV:%.*]] = fptosi double [[TMP53]] to i32 +// CHECK1-NEXT: ret i32 [[CONV]] +// +// +// CHECK1-LABEL: define {{[^@]+}}@_ZL7fstatici +// CHECK1-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0]] { +// CHECK1-NEXT: entry: +// CHECK1-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4 +// CHECK1-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4 +// CHECK1-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4 +// CHECK1-NEXT: [[N_CASTED:%.*]] = alloca i64, align 8 +// CHECK1-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8 +// CHECK1-NEXT: [[DOTCAPTURE_EXPR__CASTED2:%.*]] = alloca i64, align 8 +// CHECK1-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x ptr], align 8 +// CHECK1-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x ptr], align 8 +// CHECK1-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x ptr], align 8 +// CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4 +// CHECK1-NEXT: [[DOTCAPTURE_EXPR_3:%.*]] = alloca i32, align 4 +// CHECK1-NEXT: [[DOTCAPTURE_EXPR_4:%.*]] = alloca i32, align 4 +// CHECK1-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8 +// CHECK1-NEXT: [[DOTCAPTURE_EXPR_6:%.*]] = alloca i32, align 4 +// CHECK1-NEXT: [[DOTCAPTURE_EXPR__CASTED8:%.*]] = alloca i64, align 8 +// CHECK1-NEXT: [[DOTOFFLOAD_BASEPTRS9:%.*]] = alloca [1 x ptr], align 8 +// CHECK1-NEXT: [[DOTOFFLOAD_PTRS10:%.*]] = alloca [1 x ptr], align 8 +// CHECK1-NEXT: [[DOTOFFLOAD_MAPPERS11:%.*]] = alloca [1 x ptr], align 8 +// CHECK1-NEXT: [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 4 +// CHECK1-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]]) +// CHECK1-NEXT: store i32 [[N]], ptr [[N_ADDR]], align 4 +// CHECK1-NEXT: [[TMP1:%.*]] = load i32, ptr [[N_ADDR]], align 4 +// CHECK1-NEXT: store i32 [[TMP1]], ptr [[DOTCAPTURE_EXPR_]], align 4 +// CHECK1-NEXT: [[TMP2:%.*]] = load i32, ptr [[N_ADDR]], align 4 +// CHECK1-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP2]], 32 +// CHECK1-NEXT: store i32 [[MUL]], ptr [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK1-NEXT: [[TMP3:%.*]] = load i32, ptr [[N_ADDR]], align 4 +// CHECK1-NEXT: store i32 [[TMP3]], ptr [[N_CASTED]], align 4 +// CHECK1-NEXT: [[TMP4:%.*]] = load i64, ptr [[N_CASTED]], align 8 +// CHECK1-NEXT: [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4 +// CHECK1-NEXT: store i32 [[TMP5]], ptr [[DOTCAPTURE_EXPR__CASTED]], align 4 +// CHECK1-NEXT: [[TMP6:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR__CASTED]], align 8 +// CHECK1-NEXT: [[TMP7:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK1-NEXT: store i32 [[TMP7]], ptr [[DOTCAPTURE_EXPR__CASTED2]], align 4 +// CHECK1-NEXT: [[TMP8:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR__CASTED2]], align 8 +// CHECK1-NEXT: [[TMP9:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 +// CHECK1-NEXT: store i64 [[TMP4]], ptr [[TMP9]], align 8 +// CHECK1-NEXT: [[TMP10:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0 +// CHECK1-NEXT: store i64 [[TMP4]], ptr [[TMP10]], align 8 +// CHECK1-NEXT: [[TMP11:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0 +// CHECK1-NEXT: store ptr null, ptr [[TMP11]], align 8 +// CHECK1-NEXT: [[TMP12:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1 +// CHECK1-NEXT: store i64 [[TMP6]], ptr [[TMP12]], align 8 +// CHECK1-NEXT: [[TMP13:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 1 +// CHECK1-NEXT: store i64 [[TMP6]], ptr [[TMP13]], align 8 +// CHECK1-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1 +// CHECK1-NEXT: store ptr null, ptr [[TMP14]], align 8 +// CHECK1-NEXT: [[TMP15:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2 +// CHECK1-NEXT: store i64 [[TMP8]], ptr [[TMP15]], align 8 +// CHECK1-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 2 +// CHECK1-NEXT: store i64 [[TMP8]], ptr [[TMP16]], align 8 +// CHECK1-NEXT: [[TMP17:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2 +// CHECK1-NEXT: store ptr null, ptr [[TMP17]], align 8 +// CHECK1-NEXT: [[TMP18:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 +// CHECK1-NEXT: [[TMP19:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0 +// CHECK1-NEXT: [[TMP20:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4 +// CHECK1-NEXT: [[TMP21:%.*]] = load i32, ptr [[N_ADDR]], align 4 +// CHECK1-NEXT: store i32 [[TMP21]], ptr [[DOTCAPTURE_EXPR_3]], align 4 +// CHECK1-NEXT: [[TMP22:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_3]], align 4 +// CHECK1-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP22]], 0 +// CHECK1-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1 +// CHECK1-NEXT: [[SUB5:%.*]] = sub nsw i32 [[DIV]], 1 +// CHECK1-NEXT: store i32 [[SUB5]], ptr [[DOTCAPTURE_EXPR_4]], align 4 +// CHECK1-NEXT: [[TMP23:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4 +// CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP23]], 1 +// CHECK1-NEXT: [[TMP24:%.*]] = zext i32 [[ADD]] to i64 +// CHECK1-NEXT: [[TMP25:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK1-NEXT: [[TMP26:%.*]] = insertvalue [3 x i32] zeroinitializer, i32 [[TMP20]], 0 +// CHECK1-NEXT: [[TMP27:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0 +// CHECK1-NEXT: store i32 3, ptr [[TMP27]], align 4 +// CHECK1-NEXT: [[TMP28:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1 +// CHECK1-NEXT: store i32 3, ptr [[TMP28]], align 4 +// CHECK1-NEXT: [[TMP29:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2 +// CHECK1-NEXT: store ptr [[TMP18]], ptr [[TMP29]], align 8 +// CHECK1-NEXT: [[TMP30:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3 +// CHECK1-NEXT: store ptr [[TMP19]], ptr [[TMP30]], align 8 +// CHECK1-NEXT: [[TMP31:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4 +// CHECK1-NEXT: store ptr @.offload_sizes.3, ptr [[TMP31]], align 8 +// CHECK1-NEXT: [[TMP32:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5 +// CHECK1-NEXT: store ptr @.offload_maptypes.4, ptr [[TMP32]], align 8 +// CHECK1-NEXT: [[TMP33:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6 +// CHECK1-NEXT: store ptr null, ptr [[TMP33]], align 8 +// CHECK1-NEXT: [[TMP34:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7 +// CHECK1-NEXT: store ptr null, ptr [[TMP34]], align 8 +// CHECK1-NEXT: [[TMP35:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8 +// CHECK1-NEXT: store i64 [[TMP24]], ptr [[TMP35]], align 8 +// CHECK1-NEXT: [[TMP36:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9 +// CHECK1-NEXT: store i64 8, ptr [[TMP36]], align 8 +// CHECK1-NEXT: [[TMP37:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10 +// CHECK1-NEXT: store [3 x i32] [[TMP26]], ptr [[TMP37]], align 4 +// CHECK1-NEXT: [[TMP38:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11 +// CHECK1-NEXT: store [3 x i32] zeroinitializer, ptr [[TMP38]], align 4 +// CHECK1-NEXT: [[TMP39:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12 +// CHECK1-NEXT: store i32 [[TMP25]], ptr [[TMP39]], align 4 +// CHECK1-NEXT: [[TMP40:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB1]], i64 -1, i32 [[TMP20]], i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l71.region_id, ptr [[KERNEL_ARGS]]) +// CHECK1-NEXT: [[TMP41:%.*]] = icmp ne i32 [[TMP40]], 0 +// CHECK1-NEXT: br i1 [[TMP41]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]] +// CHECK1: omp_offload.failed: +// CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l71(i64 [[TMP4]], i64 [[TMP6]], i64 [[TMP8]]) #[[ATTR2]] +// CHECK1-NEXT: br label [[OMP_OFFLOAD_CONT]] +// CHECK1: omp_offload.cont: +// CHECK1-NEXT: [[TMP42:%.*]] = load i32, ptr [[N_ADDR]], align 4 +// CHECK1-NEXT: [[ADD7:%.*]] = add nsw i32 32, [[TMP42]] +// CHECK1-NEXT: store i32 [[ADD7]], ptr [[DOTCAPTURE_EXPR_6]], align 4 +// CHECK1-NEXT: [[TMP43:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_6]], align 4 +// CHECK1-NEXT: store i32 [[TMP43]], ptr [[DOTCAPTURE_EXPR__CASTED8]], align 4 +// CHECK1-NEXT: [[TMP44:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR__CASTED8]], align 8 +// CHECK1-NEXT: [[TMP45:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS9]], i32 0, i32 0 +// CHECK1-NEXT: store i64 [[TMP44]], ptr [[TMP45]], align 8 +// CHECK1-NEXT: [[TMP46:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS10]], i32 0, i32 0 +// CHECK1-NEXT: store i64 [[TMP44]], ptr [[TMP46]], align 8 +// CHECK1-NEXT: [[TMP47:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS11]], i64 0, i64 0 +// CHECK1-NEXT: store ptr null, ptr [[TMP47]], align 8 +// CHECK1-NEXT: [[TMP48:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS9]], i32 0, i32 0 +// CHECK1-NEXT: [[TMP49:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS10]], i32 0, i32 0 +// CHECK1-NEXT: [[TMP50:%.*]] = getelementptr inbounds nuw [[STRUCT_ANON]], ptr [[AGG_CAPTURED]], i32 0, i32 0 +// CHECK1-NEXT: [[TMP51:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_6]], align 4 +// CHECK1-NEXT: store i32 [[TMP51]], ptr [[TMP50]], align 4 +// CHECK1-NEXT: [[TMP52:%.*]] = call ptr @__kmpc_omp_target_task_alloc(ptr @[[GLOB1]], i32 [[TMP0]], i32 1, i64 64, i64 4, ptr @.omp_task_entry., i64 -1) +// CHECK1-NEXT: [[TMP53:%.*]] = getelementptr inbounds nuw [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], ptr [[TMP52]], i32 0, i32 0 +// CHECK1-NEXT: [[TMP54:%.*]] = getelementptr inbounds nuw [[STRUCT_KMP_TASK_T:%.*]], ptr [[TMP53]], i32 0, i32 0 +// CHECK1-NEXT: [[TMP55:%.*]] = load ptr, ptr [[TMP54]], align 8 +// CHECK1-NEXT: call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[TMP55]], ptr align 4 [[AGG_CAPTURED]], i64 4, i1 false) +// CHECK1-NEXT: [[TMP56:%.*]] = getelementptr inbounds nuw [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], ptr [[TMP52]], i32 0, i32 1 +// CHECK1-NEXT: [[TMP57:%.*]] = getelementptr inbounds nuw [[STRUCT__KMP_PRIVATES_T:%.*]], ptr [[TMP56]], i32 0, i32 0 +// CHECK1-NEXT: call void @llvm.memcpy.p0.p0.i64(ptr align 8 [[TMP57]], ptr align 8 [[TMP48]], i64 8, i1 false) +// CHECK1-NEXT: [[TMP58:%.*]] = getelementptr inbounds nuw [[STRUCT__KMP_PRIVATES_T]], ptr [[TMP56]], i32 0, i32 1 +// CHECK1-NEXT: call void @llvm.memcpy.p0.p0.i64(ptr align 8 [[TMP58]], ptr align 8 [[TMP49]], i64 8, i1 false) +// CHECK1-NEXT: [[TMP59:%.*]] = getelementptr inbounds nuw [[STRUCT__KMP_PRIVATES_T]], ptr [[TMP56]], i32 0, i32 2 +// CHECK1-NEXT: call void @llvm.memcpy.p0.p0.i64(ptr align 8 [[TMP59]], ptr align 8 @.offload_sizes.5, i64 8, i1 false) +// CHECK1-NEXT: [[TMP60:%.*]] = call i32 @__kmpc_omp_task(ptr @[[GLOB1]], i32 [[TMP0]], ptr [[TMP52]]) +// CHECK1-NEXT: [[TMP61:%.*]] = load i32, ptr [[N_ADDR]], align 4 +// CHECK1-NEXT: [[ADD12:%.*]] = add nsw i32 [[TMP61]], 1 +// CHECK1-NEXT: ret i32 [[ADD12]] +// +// +// CHECK1-LABEL: define {{[^@]+}}@_Z9ftemplateIiET_i +// CHECK1-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0]] comdat { +// CHECK1-NEXT: entry: +// CHECK1-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4 +// CHECK1-NEXT: [[A:%.*]] = alloca i32, align 4 +// CHECK1-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8 +// CHECK1-NEXT: [[B:%.*]] = alloca i16, align 2 +// CHECK1-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i16, align 2 +// CHECK1-NEXT: [[A_CASTED:%.*]] = alloca i64, align 8 +// CHECK1-NEXT: [[B_CASTED:%.*]] = alloca i64, align 8 +// CHECK1-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8 +// CHECK1-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x ptr], align 8 +// CHECK1-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x ptr], align 8 +// CHECK1-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x ptr], align 8 +// CHECK1-NEXT: [[KERNEL_ARGS1:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8 +// CHECK1-NEXT: store i32 [[N]], ptr [[N_ADDR]], align 4 +// CHECK1-NEXT: store i32 0, ptr [[A]], align 4 +// CHECK1-NEXT: [[TMP0:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0 +// CHECK1-NEXT: store i32 3, ptr [[TMP0]], align 4 +// CHECK1-NEXT: [[TMP1:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1 +// CHECK1-NEXT: store i32 0, ptr [[TMP1]], align 4 +// CHECK1-NEXT: [[TMP2:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2 +// CHECK1-NEXT: store ptr null, ptr [[TMP2]], align 8 +// CHECK1-NEXT: [[TMP3:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3 +// CHECK1-NEXT: store ptr null, ptr [[TMP3]], align 8 +// CHECK1-NEXT: [[TMP4:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4 +// CHECK1-NEXT: store ptr null, ptr [[TMP4]], align 8 +// CHECK1-NEXT: [[TMP5:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5 +// CHECK1-NEXT: store ptr null, ptr [[TMP5]], align 8 +// CHECK1-NEXT: [[TMP6:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6 +// CHECK1-NEXT: store ptr null, ptr [[TMP6]], align 8 +// CHECK1-NEXT: [[TMP7:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7 +// CHECK1-NEXT: store ptr null, ptr [[TMP7]], align 8 +// CHECK1-NEXT: [[TMP8:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8 +// CHECK1-NEXT: store i64 0, ptr [[TMP8]], align 8 +// CHECK1-NEXT: [[TMP9:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9 +// CHECK1-NEXT: store i64 8, ptr [[TMP9]], align 8 +// CHECK1-NEXT: [[TMP10:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10 +// CHECK1-NEXT: store [3 x i32] zeroinitializer, ptr [[TMP10]], align 4 +// CHECK1-NEXT: [[TMP11:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11 +// CHECK1-NEXT: store [3 x i32] zeroinitializer, ptr [[TMP11]], align 4 +// CHECK1-NEXT: [[TMP12:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12 +// CHECK1-NEXT: store i32 20, ptr [[TMP12]], align 4 +// CHECK1-NEXT: [[TMP13:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB1]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l55.region_id, ptr [[KERNEL_ARGS]]) +// CHECK1-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0 +// CHECK1-NEXT: br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]] +// CHECK1: omp_offload.failed: +// CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l55() #[[ATTR2]] +// CHECK1-NEXT: br label [[OMP_OFFLOAD_CONT]] +// CHECK1: omp_offload.cont: +// CHECK1-NEXT: store i16 1, ptr [[B]], align 2 +// CHECK1-NEXT: [[TMP15:%.*]] = load i16, ptr [[B]], align 2 +// CHECK1-NEXT: store i16 [[TMP15]], ptr [[DOTCAPTURE_EXPR_]], align 2 +// CHECK1-NEXT: [[TMP16:%.*]] = load i32, ptr [[A]], align 4 +// CHECK1-NEXT: store i32 [[TMP16]], ptr [[A_CASTED]], align 4 +// CHECK1-NEXT: [[TMP17:%.*]] = load i64, ptr [[A_CASTED]], align 8 +// CHECK1-NEXT: [[TMP18:%.*]] = load i16, ptr [[B]], align 2 +// CHECK1-NEXT: store i16 [[TMP18]], ptr [[B_CASTED]], align 2 +// CHECK1-NEXT: [[TMP19:%.*]] = load i64, ptr [[B_CASTED]], align 8 +// CHECK1-NEXT: [[TMP20:%.*]] = load i16, ptr [[DOTCAPTURE_EXPR_]], align 2 +// CHECK1-NEXT: store i16 [[TMP20]], ptr [[DOTCAPTURE_EXPR__CASTED]], align 2 +// CHECK1-NEXT: [[TMP21:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR__CASTED]], align 8 +// CHECK1-NEXT: [[TMP22:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 +// CHECK1-NEXT: store i64 [[TMP17]], ptr [[TMP22]], align 8 +// CHECK1-NEXT: [[TMP23:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0 +// CHECK1-NEXT: store i64 [[TMP17]], ptr [[TMP23]], align 8 +// CHECK1-NEXT: [[TMP24:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0 +// CHECK1-NEXT: store ptr null, ptr [[TMP24]], align 8 +// CHECK1-NEXT: [[TMP25:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1 +// CHECK1-NEXT: store i64 [[TMP19]], ptr [[TMP25]], align 8 +// CHECK1-NEXT: [[TMP26:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 1 +// CHECK1-NEXT: store i64 [[TMP19]], ptr [[TMP26]], align 8 +// CHECK1-NEXT: [[TMP27:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1 +// CHECK1-NEXT: store ptr null, ptr [[TMP27]], align 8 +// CHECK1-NEXT: [[TMP28:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2 +// CHECK1-NEXT: store i64 [[TMP21]], ptr [[TMP28]], align 8 +// CHECK1-NEXT: [[TMP29:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 2 +// CHECK1-NEXT: store i64 [[TMP21]], ptr [[TMP29]], align 8 +// CHECK1-NEXT: [[TMP30:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2 +// CHECK1-NEXT: store ptr null, ptr [[TMP30]], align 8 +// CHECK1-NEXT: [[TMP31:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 +// CHECK1-NEXT: [[TMP32:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0 +// CHECK1-NEXT: [[TMP33:%.*]] = load i16, ptr [[DOTCAPTURE_EXPR_]], align 2 +// CHECK1-NEXT: [[TMP34:%.*]] = sext i16 [[TMP33]] to i32 +// CHECK1-NEXT: [[TMP35:%.*]] = insertvalue [3 x i32] zeroinitializer, i32 [[TMP34]], 0 +// CHECK1-NEXT: [[TMP36:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS1]], i32 0, i32 0 +// CHECK1-NEXT: store i32 3, ptr [[TMP36]], align 4 +// CHECK1-NEXT: [[TMP37:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS1]], i32 0, i32 1 +// CHECK1-NEXT: store i32 3, ptr [[TMP37]], align 4 +// CHECK1-NEXT: [[TMP38:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS1]], i32 0, i32 2 +// CHECK1-NEXT: store ptr [[TMP31]], ptr [[TMP38]], align 8 +// CHECK1-NEXT: [[TMP39:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS1]], i32 0, i32 3 +// CHECK1-NEXT: store ptr [[TMP32]], ptr [[TMP39]], align 8 +// CHECK1-NEXT: [[TMP40:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS1]], i32 0, i32 4 +// CHECK1-NEXT: store ptr @.offload_sizes.7, ptr [[TMP40]], align 8 +// CHECK1-NEXT: [[TMP41:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS1]], i32 0, i32 5 +// CHECK1-NEXT: store ptr @.offload_maptypes.8, ptr [[TMP41]], align 8 +// CHECK1-NEXT: [[TMP42:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS1]], i32 0, i32 6 +// CHECK1-NEXT: store ptr null, ptr [[TMP42]], align 8 +// CHECK1-NEXT: [[TMP43:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS1]], i32 0, i32 7 +// CHECK1-NEXT: store ptr null, ptr [[TMP43]], align 8 +// CHECK1-NEXT: [[TMP44:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS1]], i32 0, i32 8 +// CHECK1-NEXT: store i64 0, ptr [[TMP44]], align 8 +// CHECK1-NEXT: [[TMP45:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS1]], i32 0, i32 9 +// CHECK1-NEXT: store i64 8, ptr [[TMP45]], align 8 +// CHECK1-NEXT: [[TMP46:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS1]], i32 0, i32 10 +// CHECK1-NEXT: store [3 x i32] [[TMP35]], ptr [[TMP46]], align 4 +// CHECK1-NEXT: [[TMP47:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS1]], i32 0, i32 11 +// CHECK1-NEXT: store [3 x i32] zeroinitializer, ptr [[TMP47]], align 4 +// CHECK1-NEXT: [[TMP48:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS1]], i32 0, i32 12 +// CHECK1-NEXT: store i32 1024, ptr [[TMP48]], align 4 +// CHECK1-NEXT: [[TMP49:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB1]], i64 -1, i32 [[TMP34]], i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l60.region_id, ptr [[KERNEL_ARGS1]]) +// CHECK1-NEXT: [[TMP50:%.*]] = icmp ne i32 [[TMP49]], 0 +// CHECK1-NEXT: br i1 [[TMP50]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]] +// CHECK1: omp_offload.failed2: +// CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l60(i64 [[TMP17]], i64 [[TMP19]], i64 [[TMP21]]) #[[ATTR2]] +// CHECK1-NEXT: br label [[OMP_OFFLOAD_CONT3]] +// CHECK1: omp_offload.cont3: +// CHECK1-NEXT: [[TMP51:%.*]] = load i32, ptr [[A]], align 4 +// CHECK1-NEXT: ret i32 [[TMP51]] +// +// +// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l88 +// CHECK1-SAME: (ptr noundef [[THIS:%.*]], i64 noundef [[B:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1:[0-9]+]] { +// CHECK1-NEXT: entry: +// CHECK1-NEXT: [[THIS_ADDR:%.*]] = alloca ptr, align 8 +// CHECK1-NEXT: [[B_ADDR:%.*]] = alloca i64, align 8 +// CHECK1-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8 +// CHECK1-NEXT: [[B_CASTED:%.*]] = alloca i64, align 8 +// CHECK1-NEXT: store ptr [[THIS]], ptr [[THIS_ADDR]], align 8 +// CHECK1-NEXT: store i64 [[B]], ptr [[B_ADDR]], align 8 +// CHECK1-NEXT: store i64 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 8 +// CHECK1-NEXT: [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8 +// CHECK1-NEXT: [[TMP1:%.*]] = load i32, ptr [[B_ADDR]], align 4 +// CHECK1-NEXT: store i32 [[TMP1]], ptr [[B_CASTED]], align 4 +// CHECK1-NEXT: [[TMP2:%.*]] = load i64, ptr [[B_CASTED]], align 8 +// CHECK1-NEXT: call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB1]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l88.omp_outlined, ptr [[TMP0]], i64 [[TMP2]]) +// CHECK1-NEXT: ret void +// +// +// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l88.omp_outlined +// CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef [[THIS:%.*]], i64 noundef [[B:%.*]]) #[[ATTR1]] { +// CHECK1-NEXT: entry: +// CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8 +// CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8 +// CHECK1-NEXT: [[THIS_ADDR:%.*]] = alloca ptr, align 8 +// CHECK1-NEXT: [[B_ADDR:%.*]] = alloca i64, align 8 +// CHECK1-NEXT: store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8 +// CHECK1-NEXT: store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8 +// CHECK1-NEXT: store ptr [[THIS]], ptr [[THIS_ADDR]], align 8 +// CHECK1-NEXT: store i64 [[B]], ptr [[B_ADDR]], align 8 +// CHECK1-NEXT: [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8 +// CHECK1-NEXT: [[TMP1:%.*]] = load i32, ptr [[B_ADDR]], align 4 +// CHECK1-NEXT: [[CONV:%.*]] = sitofp i32 [[TMP1]] to double +// CHECK1-NEXT: [[ADD:%.*]] = fadd double [[CONV]], 1.500000e+00 +// CHECK1-NEXT: [[A:%.*]] = getelementptr inbounds nuw [[STRUCT_S1:%.*]], ptr [[TMP0]], i32 0, i32 0 +// CHECK1-NEXT: store double [[ADD]], ptr [[A]], align 8 +// CHECK1-NEXT: ret void +// +// +// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l93 +// CHECK1-SAME: (ptr noundef [[THIS:%.*]]) #[[ATTR1]] { +// CHECK1-NEXT: entry: +// CHECK1-NEXT: [[THIS_ADDR:%.*]] = alloca ptr, align 8 +// CHECK1-NEXT: store ptr [[THIS]], ptr [[THIS_ADDR]], align 8 +// CHECK1-NEXT: [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8 +// CHECK1-NEXT: [[A:%.*]] = getelementptr inbounds nuw [[STRUCT_S1:%.*]], ptr [[TMP0]], i32 0, i32 0 +// CHECK1-NEXT: store double 2.500000e+00, ptr [[A]], align 8 +// CHECK1-NEXT: ret void +// +// +// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l71 +// CHECK1-SAME: (i64 noundef [[N:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]], i64 noundef [[DOTCAPTURE_EXPR_1:%.*]]) #[[ATTR1]] { +// CHECK1-NEXT: entry: +// CHECK1-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8 +// CHECK1-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8 +// CHECK1-NEXT: [[DOTCAPTURE_EXPR__ADDR2:%.*]] = alloca i64, align 8 +// CHECK1-NEXT: [[N_CASTED:%.*]] = alloca i64, align 8 +// CHECK1-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]]) +// CHECK1-NEXT: store i64 [[N]], ptr [[N_ADDR]], align 8 +// CHECK1-NEXT: store i64 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 8 +// CHECK1-NEXT: store i64 [[DOTCAPTURE_EXPR_1]], ptr [[DOTCAPTURE_EXPR__ADDR2]], align 8 +// CHECK1-NEXT: [[TMP1:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ADDR]], align 4 +// CHECK1-NEXT: call void @__kmpc_push_num_teams(ptr @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]], i32 0) +// CHECK1-NEXT: [[TMP2:%.*]] = load i32, ptr [[N_ADDR]], align 4 +// CHECK1-NEXT: store i32 [[TMP2]], ptr [[N_CASTED]], align 4 +// CHECK1-NEXT: [[TMP3:%.*]] = load i64, ptr [[N_CASTED]], align 8 +// CHECK1-NEXT: call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB1]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l71.omp_outlined, i64 [[TMP3]]) +// CHECK1-NEXT: ret void +// +// +// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l71.omp_outlined +// CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[N:%.*]]) #[[ATTR1]] { +// CHECK1-NEXT: entry: +// CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8 +// CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8 +// CHECK1-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8 +// CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 +// CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4 +// CHECK1-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4 +// CHECK1-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4 +// CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4 +// CHECK1-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4 +// CHECK1-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4 +// CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 +// CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 +// CHECK1-NEXT: [[I3:%.*]] = alloca i32, align 4 +// CHECK1-NEXT: [[N_CASTED:%.*]] = alloca i64, align 8 +// CHECK1-NEXT: store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8 +// CHECK1-NEXT: store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8 +// CHECK1-NEXT: store i64 [[N]], ptr [[N_ADDR]], align 8 +// CHECK1-NEXT: [[TMP0:%.*]] = load i32, ptr [[N_ADDR]], align 4 +// CHECK1-NEXT: store i32 [[TMP0]], ptr [[DOTCAPTURE_EXPR_]], align 4 +// CHECK1-NEXT: [[TMP1:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4 +// CHECK1-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP1]], 0 +// CHECK1-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1 +// CHECK1-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1 +// CHECK1-NEXT: store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK1-NEXT: store i32 0, ptr [[I]], align 4 +// CHECK1-NEXT: [[TMP2:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4 +// CHECK1-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP2]] +// CHECK1-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]] +// CHECK1: omp.precond.then: +// CHECK1-NEXT: store i32 0, ptr [[DOTOMP_COMB_LB]], align 4 +// CHECK1-NEXT: [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK1-NEXT: store i32 [[TMP3]], ptr [[DOTOMP_COMB_UB]], align 4 +// CHECK1-NEXT: store i32 1, ptr [[DOTOMP_STRIDE]], align 4 +// CHECK1-NEXT: store i32 0, ptr [[DOTOMP_IS_LAST]], align 4 +// CHECK1-NEXT: [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8 +// CHECK1-NEXT: [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4 +// CHECK1-NEXT: call void @__kmpc_for_static_init_4(ptr @[[GLOB2:[0-9]+]], i32 [[TMP5]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1) +// CHECK1-NEXT: [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4 +// CHECK1-NEXT: [[TMP7:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK1-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP6]], [[TMP7]] +// CHECK1-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] +// CHECK1: cond.true: +// CHECK1-NEXT: [[TMP8:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK1-NEXT: br label [[COND_END:%.*]] +// CHECK1: cond.false: +// CHECK1-NEXT: [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4 +// CHECK1-NEXT: br label [[COND_END]] +// CHECK1: cond.end: +// CHECK1-NEXT: [[COND:%.*]] = phi i32 [ [[TMP8]], [[COND_TRUE]] ], [ [[TMP9]], [[COND_FALSE]] ] +// CHECK1-NEXT: store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4 +// CHECK1-NEXT: [[TMP10:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4 +// CHECK1-NEXT: store i32 [[TMP10]], ptr [[DOTOMP_IV]], align 4 +// CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] +// CHECK1: omp.inner.for.cond: +// CHECK1-NEXT: [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP17:![0-9]+]] +// CHECK1-NEXT: [[TMP12:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP17]] +// CHECK1-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]] +// CHECK1-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] +// CHECK1: omp.inner.for.body: +// CHECK1-NEXT: [[TMP13:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP17]] +// CHECK1-NEXT: [[TMP14:%.*]] = zext i32 [[TMP13]] to i64 +// CHECK1-NEXT: [[TMP15:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP17]] +// CHECK1-NEXT: [[TMP16:%.*]] = zext i32 [[TMP15]] to i64 +// CHECK1-NEXT: [[TMP17:%.*]] = load i32, ptr [[N_ADDR]], align 4, !llvm.access.group [[ACC_GRP17]] +// CHECK1-NEXT: store i32 [[TMP17]], ptr [[N_CASTED]], align 4, !llvm.access.group [[ACC_GRP17]] +// CHECK1-NEXT: [[TMP18:%.*]] = load i64, ptr [[N_CASTED]], align 8, !llvm.access.group [[ACC_GRP17]] +// CHECK1-NEXT: call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB1]], i32 3, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l71.omp_outlined.omp_outlined, i64 [[TMP14]], i64 [[TMP16]], i64 [[TMP18]]), !llvm.access.group [[ACC_GRP17]] +// CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] +// CHECK1: omp.inner.for.inc: +// CHECK1-NEXT: [[TMP19:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP17]] +// CHECK1-NEXT: [[TMP20:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP17]] +// CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP19]], [[TMP20]] +// CHECK1-NEXT: store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP17]] +// CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP18:![0-9]+]] +// CHECK1: omp.inner.for.end: +// CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]] +// CHECK1: omp.loop.exit: +// CHECK1-NEXT: [[TMP21:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8 +// CHECK1-NEXT: [[TMP22:%.*]] = load i32, ptr [[TMP21]], align 4 +// CHECK1-NEXT: call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP22]]) +// CHECK1-NEXT: [[TMP23:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4 +// CHECK1-NEXT: [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0 +// CHECK1-NEXT: br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]] +// CHECK1: .omp.final.then: +// CHECK1-NEXT: [[TMP25:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4 +// CHECK1-NEXT: [[SUB6:%.*]] = sub nsw i32 [[TMP25]], 0 +// CHECK1-NEXT: [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1 +// CHECK1-NEXT: [[MUL:%.*]] = mul nsw i32 [[DIV7]], 1 +// CHECK1-NEXT: [[ADD8:%.*]] = add nsw i32 0, [[MUL]] +// CHECK1-NEXT: store i32 [[ADD8]], ptr [[I3]], align 4 +// CHECK1-NEXT: br label [[DOTOMP_FINAL_DONE]] +// CHECK1: .omp.final.done: +// CHECK1-NEXT: br label [[OMP_PRECOND_END]] +// CHECK1: omp.precond.end: +// CHECK1-NEXT: ret void +// +// +// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l71.omp_outlined.omp_outlined +// CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i64 noundef [[N:%.*]]) #[[ATTR1]] { +// CHECK1-NEXT: entry: +// CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8 +// CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8 +// CHECK1-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8 +// CHECK1-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8 +// CHECK1-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8 +// CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 +// CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4 +// CHECK1-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4 +// CHECK1-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4 +// CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4 +// CHECK1-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4 +// CHECK1-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4 +// CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 +// CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 +// CHECK1-NEXT: [[I4:%.*]] = alloca i32, align 4 +// CHECK1-NEXT: store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8 +// CHECK1-NEXT: store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8 +// CHECK1-NEXT: store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8 +// CHECK1-NEXT: store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8 +// CHECK1-NEXT: store i64 [[N]], ptr [[N_ADDR]], align 8 +// CHECK1-NEXT: [[TMP0:%.*]] = load i32, ptr [[N_ADDR]], align 4 +// CHECK1-NEXT: store i32 [[TMP0]], ptr [[DOTCAPTURE_EXPR_]], align 4 +// CHECK1-NEXT: [[TMP1:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4 +// CHECK1-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP1]], 0 +// CHECK1-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1 +// CHECK1-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1 +// CHECK1-NEXT: store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK1-NEXT: store i32 0, ptr [[I]], align 4 +// CHECK1-NEXT: [[TMP2:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4 +// CHECK1-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP2]] +// CHECK1-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]] +// CHECK1: omp.precond.then: +// CHECK1-NEXT: store i32 0, ptr [[DOTOMP_LB]], align 4 +// CHECK1-NEXT: [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK1-NEXT: store i32 [[TMP3]], ptr [[DOTOMP_UB]], align 4 +// CHECK1-NEXT: [[TMP4:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8 +// CHECK1-NEXT: [[CONV:%.*]] = trunc i64 [[TMP4]] to i32 +// CHECK1-NEXT: [[TMP5:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8 +// CHECK1-NEXT: [[CONV3:%.*]] = trunc i64 [[TMP5]] to i32 +// CHECK1-NEXT: store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4 +// CHECK1-NEXT: store i32 [[CONV3]], ptr [[DOTOMP_UB]], align 4 +// CHECK1-NEXT: store i32 1, ptr [[DOTOMP_STRIDE]], align 4 +// CHECK1-NEXT: store i32 0, ptr [[DOTOMP_IS_LAST]], align 4 +// CHECK1-NEXT: [[TMP6:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8 +// CHECK1-NEXT: [[TMP7:%.*]] = load i32, ptr [[TMP6]], align 4 +// CHECK1-NEXT: call void @__kmpc_for_static_init_4(ptr @[[GLOB3:[0-9]+]], i32 [[TMP7]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1) +// CHECK1-NEXT: [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4 +// CHECK1-NEXT: [[TMP9:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK1-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP8]], [[TMP9]] +// CHECK1-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] +// CHECK1: cond.true: +// CHECK1-NEXT: [[TMP10:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK1-NEXT: br label [[COND_END:%.*]] +// CHECK1: cond.false: +// CHECK1-NEXT: [[TMP11:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4 +// CHECK1-NEXT: br label [[COND_END]] +// CHECK1: cond.end: +// CHECK1-NEXT: [[COND:%.*]] = phi i32 [ [[TMP10]], [[COND_TRUE]] ], [ [[TMP11]], [[COND_FALSE]] ] +// CHECK1-NEXT: store i32 [[COND]], ptr [[DOTOMP_UB]], align 4 +// CHECK1-NEXT: [[TMP12:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4 +// CHECK1-NEXT: store i32 [[TMP12]], ptr [[DOTOMP_IV]], align 4 +// CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] +// CHECK1: omp.inner.for.cond: +// CHECK1-NEXT: [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP21:![0-9]+]] +// CHECK1-NEXT: [[TMP14:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP21]] +// CHECK1-NEXT: [[CMP6:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]] +// CHECK1-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] +// CHECK1: omp.inner.for.body: +// CHECK1-NEXT: [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP21]] +// CHECK1-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1 +// CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]] +// CHECK1-NEXT: store i32 [[ADD]], ptr [[I4]], align 4, !llvm.access.group [[ACC_GRP21]] +// CHECK1-NEXT: br label [[OMP_BODY_CONTINUE:%.*]] +// CHECK1: omp.body.continue: +// CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] +// CHECK1: omp.inner.for.inc: +// CHECK1-NEXT: [[TMP16:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP21]] +// CHECK1-NEXT: [[ADD7:%.*]] = add nsw i32 [[TMP16]], 1 +// CHECK1-NEXT: store i32 [[ADD7]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP21]] +// CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP22:![0-9]+]] +// CHECK1: omp.inner.for.end: +// CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]] +// CHECK1: omp.loop.exit: +// CHECK1-NEXT: [[TMP17:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8 +// CHECK1-NEXT: [[TMP18:%.*]] = load i32, ptr [[TMP17]], align 4 +// CHECK1-NEXT: call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP18]]) +// CHECK1-NEXT: [[TMP19:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4 +// CHECK1-NEXT: [[TMP20:%.*]] = icmp ne i32 [[TMP19]], 0 +// CHECK1-NEXT: br i1 [[TMP20]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]] +// CHECK1: .omp.final.then: +// CHECK1-NEXT: [[TMP21:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4 +// CHECK1-NEXT: [[SUB8:%.*]] = sub nsw i32 [[TMP21]], 0 +// CHECK1-NEXT: [[DIV9:%.*]] = sdiv i32 [[SUB8]], 1 +// CHECK1-NEXT: [[MUL10:%.*]] = mul nsw i32 [[DIV9]], 1 +// CHECK1-NEXT: [[ADD11:%.*]] = add nsw i32 0, [[MUL10]] +// CHECK1-NEXT: store i32 [[ADD11]], ptr [[I4]], align 4 +// CHECK1-NEXT: br label [[DOTOMP_FINAL_DONE]] +// CHECK1: .omp.final.done: +// CHECK1-NEXT: br label [[OMP_PRECOND_END]] +// CHECK1: omp.precond.end: +// CHECK1-NEXT: ret void +// +// +// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l75 +// CHECK1-SAME: (i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] { +// CHECK1-NEXT: entry: +// CHECK1-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8 +// CHECK1-NEXT: store i64 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 8 +// CHECK1-NEXT: call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB1]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l75.omp_outlined) +// CHECK1-NEXT: ret void +// +// +// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l75.omp_outlined +// CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] { +// CHECK1-NEXT: entry: +// CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8 +// CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8 +// CHECK1-NEXT: store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8 +// CHECK1-NEXT: store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8 +// CHECK1-NEXT: ret void +// +// +// CHECK1-LABEL: define {{[^@]+}}@.omp_task_privates_map. +// CHECK1-SAME: (ptr noalias noundef [[TMP0:%.*]], ptr noalias noundef [[TMP1:%.*]], ptr noalias noundef [[TMP2:%.*]], ptr noalias noundef [[TMP3:%.*]]) #[[ATTR3:[0-9]+]] { +// CHECK1-NEXT: entry: +// CHECK1-NEXT: [[DOTADDR:%.*]] = alloca ptr, align 8 +// CHECK1-NEXT: [[DOTADDR1:%.*]] = alloca ptr, align 8 +// CHECK1-NEXT: [[DOTADDR2:%.*]] = alloca ptr, align 8 +// CHECK1-NEXT: [[DOTADDR3:%.*]] = alloca ptr, align 8 +// CHECK1-NEXT: store ptr [[TMP0]], ptr [[DOTADDR]], align 8 +// CHECK1-NEXT: store ptr [[TMP1]], ptr [[DOTADDR1]], align 8 +// CHECK1-NEXT: store ptr [[TMP2]], ptr [[DOTADDR2]], align 8 +// CHECK1-NEXT: store ptr [[TMP3]], ptr [[DOTADDR3]], align 8 +// CHECK1-NEXT: [[TMP4:%.*]] = load ptr, ptr [[DOTADDR]], align 8 +// CHECK1-NEXT: [[TMP5:%.*]] = getelementptr inbounds nuw [[STRUCT__KMP_PRIVATES_T:%.*]], ptr [[TMP4]], i32 0, i32 0 +// CHECK1-NEXT: [[TMP6:%.*]] = load ptr, ptr [[DOTADDR1]], align 8 +// CHECK1-NEXT: store ptr [[TMP5]], ptr [[TMP6]], align 8 +// CHECK1-NEXT: [[TMP7:%.*]] = getelementptr inbounds nuw [[STRUCT__KMP_PRIVATES_T]], ptr [[TMP4]], i32 0, i32 1 +// CHECK1-NEXT: [[TMP8:%.*]] = load ptr, ptr [[DOTADDR2]], align 8 +// CHECK1-NEXT: store ptr [[TMP7]], ptr [[TMP8]], align 8 +// CHECK1-NEXT: [[TMP9:%.*]] = getelementptr inbounds nuw [[STRUCT__KMP_PRIVATES_T]], ptr [[TMP4]], i32 0, i32 2 +// CHECK1-NEXT: [[TMP10:%.*]] = load ptr, ptr [[DOTADDR3]], align 8 +// CHECK1-NEXT: store ptr [[TMP9]], ptr [[TMP10]], align 8 +// CHECK1-NEXT: ret void +// +// +// CHECK1-LABEL: define {{[^@]+}}@.omp_task_entry. +// CHECK1-SAME: (i32 noundef signext [[TMP0:%.*]], ptr noalias noundef [[TMP1:%.*]]) #[[ATTR4:[0-9]+]] { +// CHECK1-NEXT: entry: +// CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4 +// CHECK1-NEXT: [[DOTPART_ID__ADDR_I:%.*]] = alloca ptr, align 8 +// CHECK1-NEXT: [[DOTPRIVATES__ADDR_I:%.*]] = alloca ptr, align 8 +// CHECK1-NEXT: [[DOTCOPY_FN__ADDR_I:%.*]] = alloca ptr, align 8 +// CHECK1-NEXT: [[DOTTASK_T__ADDR_I:%.*]] = alloca ptr, align 8 +// CHECK1-NEXT: [[__CONTEXT_ADDR_I:%.*]] = alloca ptr, align 8 +// CHECK1-NEXT: [[DOTFIRSTPRIV_PTR_ADDR_I:%.*]] = alloca ptr, align 8 +// CHECK1-NEXT: [[DOTFIRSTPRIV_PTR_ADDR1_I:%.*]] = alloca ptr, align 8 +// CHECK1-NEXT: [[DOTFIRSTPRIV_PTR_ADDR2_I:%.*]] = alloca ptr, align 8 +// CHECK1-NEXT: [[KERNEL_ARGS_I:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8 +// CHECK1-NEXT: [[DOTCAPTURE_EXPR__CASTED_I:%.*]] = alloca i64, align 8 +// CHECK1-NEXT: [[DOTADDR:%.*]] = alloca i32, align 4 +// CHECK1-NEXT: [[DOTADDR1:%.*]] = alloca ptr, align 8 +// CHECK1-NEXT: store i32 [[TMP0]], ptr [[DOTADDR]], align 4 +// CHECK1-NEXT: store ptr [[TMP1]], ptr [[DOTADDR1]], align 8 +// CHECK1-NEXT: [[TMP2:%.*]] = load i32, ptr [[DOTADDR]], align 4 +// CHECK1-NEXT: [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 8 +// CHECK1-NEXT: [[TMP4:%.*]] = getelementptr inbounds nuw [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], ptr [[TMP3]], i32 0, i32 0 +// CHECK1-NEXT: [[TMP5:%.*]] = getelementptr inbounds nuw [[STRUCT_KMP_TASK_T:%.*]], ptr [[TMP4]], i32 0, i32 2 +// CHECK1-NEXT: [[TMP6:%.*]] = getelementptr inbounds nuw [[STRUCT_KMP_TASK_T]], ptr [[TMP4]], i32 0, i32 0 +// CHECK1-NEXT: [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8 +// CHECK1-NEXT: [[TMP8:%.*]] = getelementptr inbounds nuw [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], ptr [[TMP3]], i32 0, i32 1 +// CHECK1-NEXT: call void @llvm.experimental.noalias.scope.decl(metadata [[META24:![0-9]+]]) +// CHECK1-NEXT: call void @llvm.experimental.noalias.scope.decl(metadata [[META27:![0-9]+]]) +// CHECK1-NEXT: call void @llvm.experimental.noalias.scope.decl(metadata [[META29:![0-9]+]]) +// CHECK1-NEXT: call void @llvm.experimental.noalias.scope.decl(metadata [[META31:![0-9]+]]) +// CHECK1-NEXT: store i32 [[TMP2]], ptr [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias [[META33:![0-9]+]] +// CHECK1-NEXT: store ptr [[TMP5]], ptr [[DOTPART_ID__ADDR_I]], align 8, !noalias [[META33]] +// CHECK1-NEXT: store ptr [[TMP8]], ptr [[DOTPRIVATES__ADDR_I]], align 8, !noalias [[META33]] +// CHECK1-NEXT: store ptr @.omp_task_privates_map., ptr [[DOTCOPY_FN__ADDR_I]], align 8, !noalias [[META33]] +// CHECK1-NEXT: store ptr [[TMP3]], ptr [[DOTTASK_T__ADDR_I]], align 8, !noalias [[META33]] +// CHECK1-NEXT: store ptr [[TMP7]], ptr [[__CONTEXT_ADDR_I]], align 8, !noalias [[META33]] +// CHECK1-NEXT: [[TMP9:%.*]] = load ptr, ptr [[__CONTEXT_ADDR_I]], align 8, !noalias [[META33]] +// CHECK1-NEXT: [[TMP10:%.*]] = load ptr, ptr [[DOTCOPY_FN__ADDR_I]], align 8, !noalias [[META33]] +// CHECK1-NEXT: [[TMP11:%.*]] = load ptr, ptr [[DOTPRIVATES__ADDR_I]], align 8, !noalias [[META33]] +// CHECK1-NEXT: call void [[TMP10]](ptr [[TMP11]], ptr [[DOTFIRSTPRIV_PTR_ADDR_I]], ptr [[DOTFIRSTPRIV_PTR_ADDR1_I]], ptr [[DOTFIRSTPRIV_PTR_ADDR2_I]]) #[[ATTR2]] +// CHECK1-NEXT: [[TMP12:%.*]] = load ptr, ptr [[DOTFIRSTPRIV_PTR_ADDR_I]], align 8, !noalias [[META33]] +// CHECK1-NEXT: [[TMP13:%.*]] = load ptr, ptr [[DOTFIRSTPRIV_PTR_ADDR1_I]], align 8, !noalias [[META33]] +// CHECK1-NEXT: [[TMP14:%.*]] = load ptr, ptr [[DOTFIRSTPRIV_PTR_ADDR2_I]], align 8, !noalias [[META33]] +// CHECK1-NEXT: [[TMP15:%.*]] = load i32, ptr [[TMP9]], align 4 +// CHECK1-NEXT: store i32 3, ptr [[KERNEL_ARGS_I]], align 4, !noalias [[META33]] +// CHECK1-NEXT: [[TMP16:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS_I]], i32 0, i32 1 +// CHECK1-NEXT: store i32 1, ptr [[TMP16]], align 4, !noalias [[META33]] +// CHECK1-NEXT: [[TMP17:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS_I]], i32 0, i32 2 +// CHECK1-NEXT: store ptr [[TMP12]], ptr [[TMP17]], align 8, !noalias [[META33]] +// CHECK1-NEXT: [[TMP18:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS_I]], i32 0, i32 3 +// CHECK1-NEXT: store ptr [[TMP13]], ptr [[TMP18]], align 8, !noalias [[META33]] +// CHECK1-NEXT: [[TMP19:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS_I]], i32 0, i32 4 +// CHECK1-NEXT: store ptr [[TMP14]], ptr [[TMP19]], align 8, !noalias [[META33]] +// CHECK1-NEXT: [[TMP20:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS_I]], i32 0, i32 5 +// CHECK1-NEXT: store ptr @.offload_maptypes.6, ptr [[TMP20]], align 8, !noalias [[META33]] +// CHECK1-NEXT: [[TMP21:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS_I]], i32 0, i32 6 +// CHECK1-NEXT: store ptr null, ptr [[TMP21]], align 8, !noalias [[META33]] +// CHECK1-NEXT: [[TMP22:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS_I]], i32 0, i32 7 +// CHECK1-NEXT: store ptr null, ptr [[TMP22]], align 8, !noalias [[META33]] +// CHECK1-NEXT: [[TMP23:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS_I]], i32 0, i32 8 +// CHECK1-NEXT: store i64 0, ptr [[TMP23]], align 8, !noalias [[META33]] +// CHECK1-NEXT: [[TMP24:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS_I]], i32 0, i32 9 +// CHECK1-NEXT: store i64 9, ptr [[TMP24]], align 8, !noalias [[META33]] +// CHECK1-NEXT: [[TMP25:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS_I]], i32 0, i32 10 +// CHECK1-NEXT: store [3 x i32] zeroinitializer, ptr [[TMP25]], align 4, !noalias [[META33]] +// CHECK1-NEXT: [[TMP26:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS_I]], i32 0, i32 11 +// CHECK1-NEXT: store [3 x i32] zeroinitializer, ptr [[TMP26]], align 4, !noalias [[META33]] +// CHECK1-NEXT: [[TMP27:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS_I]], i32 0, i32 12 +// CHECK1-NEXT: store i32 [[TMP15]], ptr [[TMP27]], align 4, !noalias [[META33]] +// CHECK1-NEXT: [[TMP28:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB1]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l75.region_id, ptr [[KERNEL_ARGS_I]]) +// CHECK1-NEXT: [[TMP29:%.*]] = icmp ne i32 [[TMP28]], 0 +// CHECK1-NEXT: br i1 [[TMP29]], label [[OMP_OFFLOAD_FAILED_I:%.*]], label [[DOTOMP_OUTLINED__EXIT:%.*]] +// CHECK1: omp_offload.failed.i: +// CHECK1-NEXT: [[TMP30:%.*]] = load i32, ptr [[TMP9]], align 4 +// CHECK1-NEXT: store i32 [[TMP30]], ptr [[DOTCAPTURE_EXPR__CASTED_I]], align 4, !noalias [[META33]] +// CHECK1-NEXT: [[TMP31:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR__CASTED_I]], align 8, !noalias [[META33]] +// CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l75(i64 [[TMP31]]) #[[ATTR2]] +// CHECK1-NEXT: br label [[DOTOMP_OUTLINED__EXIT]] +// CHECK1: .omp_outlined..exit: +// CHECK1-NEXT: ret i32 0 +// +// +// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l55 +// CHECK1-SAME: () #[[ATTR1]] { +// CHECK1-NEXT: entry: +// CHECK1-NEXT: call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB1]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l55.omp_outlined) +// CHECK1-NEXT: ret void +// +// +// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l55.omp_outlined +// CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] { +// CHECK1-NEXT: entry: +// CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8 +// CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8 +// CHECK1-NEXT: store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8 +// CHECK1-NEXT: store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8 +// CHECK1-NEXT: ret void +// +// +// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l60 +// CHECK1-SAME: (i64 noundef [[A:%.*]], i64 noundef [[B:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] { +// CHECK1-NEXT: entry: +// CHECK1-NEXT: [[A_ADDR:%.*]] = alloca i64, align 8 +// CHECK1-NEXT: [[B_ADDR:%.*]] = alloca i64, align 8 +// CHECK1-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8 +// CHECK1-NEXT: [[A_CASTED:%.*]] = alloca i64, align 8 +// CHECK1-NEXT: [[B_CASTED:%.*]] = alloca i64, align 8 +// CHECK1-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]]) +// CHECK1-NEXT: store i64 [[A]], ptr [[A_ADDR]], align 8 +// CHECK1-NEXT: store i64 [[B]], ptr [[B_ADDR]], align 8 +// CHECK1-NEXT: store i64 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 8 +// CHECK1-NEXT: [[TMP1:%.*]] = load i16, ptr [[DOTCAPTURE_EXPR__ADDR]], align 2 +// CHECK1-NEXT: [[TMP2:%.*]] = sext i16 [[TMP1]] to i32 +// CHECK1-NEXT: call void @__kmpc_push_num_teams(ptr @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]], i32 0) +// CHECK1-NEXT: [[TMP3:%.*]] = load i32, ptr [[A_ADDR]], align 4 +// CHECK1-NEXT: store i32 [[TMP3]], ptr [[A_CASTED]], align 4 +// CHECK1-NEXT: [[TMP4:%.*]] = load i64, ptr [[A_CASTED]], align 8 +// CHECK1-NEXT: [[TMP5:%.*]] = load i16, ptr [[B_ADDR]], align 2 +// CHECK1-NEXT: store i16 [[TMP5]], ptr [[B_CASTED]], align 2 +// CHECK1-NEXT: [[TMP6:%.*]] = load i64, ptr [[B_CASTED]], align 8 +// CHECK1-NEXT: call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB1]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l60.omp_outlined, i64 [[TMP4]], i64 [[TMP6]]) +// CHECK1-NEXT: ret void +// +// +// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l60.omp_outlined +// CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]], i64 noundef [[B:%.*]]) #[[ATTR1]] { +// CHECK1-NEXT: entry: +// CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8 +// CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8 +// CHECK1-NEXT: [[A_ADDR:%.*]] = alloca i64, align 8 +// CHECK1-NEXT: [[B_ADDR:%.*]] = alloca i64, align 8 +// CHECK1-NEXT: store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8 +// CHECK1-NEXT: store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8 +// CHECK1-NEXT: store i64 [[A]], ptr [[A_ADDR]], align 8 +// CHECK1-NEXT: store i64 [[B]], ptr [[B_ADDR]], align 8 +// CHECK1-NEXT: [[TMP0:%.*]] = load i16, ptr [[B_ADDR]], align 2 +// CHECK1-NEXT: [[CONV:%.*]] = sext i16 [[TMP0]] to i32 +// CHECK1-NEXT: [[TMP1:%.*]] = load i32, ptr [[A_ADDR]], align 4 +// CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CONV]] +// CHECK1-NEXT: store i32 [[ADD]], ptr [[A_ADDR]], align 4 +// CHECK1-NEXT: ret void +// +// +// CHECK3-LABEL: define {{[^@]+}}@_Z3bari +// CHECK3-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0:[0-9]+]] { +// CHECK3-NEXT: entry: +// CHECK3-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: [[A:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: [[S:%.*]] = alloca [[STRUCT_S1:%.*]], align 4 +// CHECK3-NEXT: store i32 [[N]], ptr [[N_ADDR]], align 4 +// CHECK3-NEXT: store i32 0, ptr [[A]], align 4 +// CHECK3-NEXT: [[TMP0:%.*]] = load i32, ptr [[N_ADDR]], align 4 +// CHECK3-NEXT: [[CALL:%.*]] = call noundef i32 @_ZN2S12r1Ei(ptr noundef nonnull align 4 dereferenceable(8) [[S]], i32 noundef [[TMP0]]) +// CHECK3-NEXT: [[TMP1:%.*]] = load i32, ptr [[A]], align 4 +// CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CALL]] +// CHECK3-NEXT: store i32 [[ADD]], ptr [[A]], align 4 +// CHECK3-NEXT: [[TMP2:%.*]] = load i32, ptr [[N_ADDR]], align 4 +// CHECK3-NEXT: [[CALL1:%.*]] = call noundef i32 @_ZL7fstatici(i32 noundef [[TMP2]]) +// CHECK3-NEXT: [[TMP3:%.*]] = load i32, ptr [[A]], align 4 +// CHECK3-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP3]], [[CALL1]] +// CHECK3-NEXT: store i32 [[ADD2]], ptr [[A]], align 4 +// CHECK3-NEXT: [[TMP4:%.*]] = load i32, ptr [[N_ADDR]], align 4 +// CHECK3-NEXT: [[CALL3:%.*]] = call noundef i32 @_Z9ftemplateIiET_i(i32 noundef [[TMP4]]) +// CHECK3-NEXT: [[TMP5:%.*]] = load i32, ptr [[A]], align 4 +// CHECK3-NEXT: [[ADD4:%.*]] = add nsw i32 [[TMP5]], [[CALL3]] +// CHECK3-NEXT: store i32 [[ADD4]], ptr [[A]], align 4 +// CHECK3-NEXT: [[TMP6:%.*]] = load i32, ptr [[A]], align 4 +// CHECK3-NEXT: ret i32 [[TMP6]] +// +// +// CHECK3-LABEL: define {{[^@]+}}@_ZN2S12r1Ei +// CHECK3-SAME: (ptr noundef nonnull align 4 dereferenceable(8) [[THIS:%.*]], i32 noundef [[N:%.*]]) #[[ATTR0]] comdat align 2 { +// CHECK3-NEXT: entry: +// CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca ptr, align 4 +// CHECK3-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: [[B:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: [[B_CASTED:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x ptr], align 4 +// CHECK3-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x ptr], align 4 +// CHECK3-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x ptr], align 4 +// CHECK3-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8 +// CHECK3-NEXT: [[DOTOFFLOAD_BASEPTRS3:%.*]] = alloca [1 x ptr], align 4 +// CHECK3-NEXT: [[DOTOFFLOAD_PTRS4:%.*]] = alloca [1 x ptr], align 4 +// CHECK3-NEXT: [[DOTOFFLOAD_MAPPERS5:%.*]] = alloca [1 x ptr], align 4 +// CHECK3-NEXT: [[KERNEL_ARGS6:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8 +// CHECK3-NEXT: store ptr [[THIS]], ptr [[THIS_ADDR]], align 4 +// CHECK3-NEXT: store i32 [[N]], ptr [[N_ADDR]], align 4 +// CHECK3-NEXT: [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4 +// CHECK3-NEXT: store i32 1, ptr [[B]], align 4 +// CHECK3-NEXT: [[TMP0:%.*]] = load i32, ptr [[N_ADDR]], align 4 +// CHECK3-NEXT: [[TMP1:%.*]] = load i32, ptr [[B]], align 4 +// CHECK3-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP0]], [[TMP1]] +// CHECK3-NEXT: store i32 [[SUB]], ptr [[DOTCAPTURE_EXPR_]], align 4 +// CHECK3-NEXT: [[TMP2:%.*]] = load i32, ptr [[B]], align 4 +// CHECK3-NEXT: store i32 [[TMP2]], ptr [[B_CASTED]], align 4 +// CHECK3-NEXT: [[TMP3:%.*]] = load i32, ptr [[B_CASTED]], align 4 +// CHECK3-NEXT: [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4 +// CHECK3-NEXT: store i32 [[TMP4]], ptr [[DOTCAPTURE_EXPR__CASTED]], align 4 +// CHECK3-NEXT: [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__CASTED]], align 4 +// CHECK3-NEXT: [[A:%.*]] = getelementptr inbounds nuw [[STRUCT_S1:%.*]], ptr [[THIS1]], i32 0, i32 0 +// CHECK3-NEXT: [[TMP6:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 +// CHECK3-NEXT: store ptr [[THIS1]], ptr [[TMP6]], align 4 +// CHECK3-NEXT: [[TMP7:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0 +// CHECK3-NEXT: store ptr [[A]], ptr [[TMP7]], align 4 +// CHECK3-NEXT: [[TMP8:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0 +// CHECK3-NEXT: store ptr null, ptr [[TMP8]], align 4 +// CHECK3-NEXT: [[TMP9:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1 +// CHECK3-NEXT: store i32 [[TMP3]], ptr [[TMP9]], align 4 +// CHECK3-NEXT: [[TMP10:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 1 +// CHECK3-NEXT: store i32 [[TMP3]], ptr [[TMP10]], align 4 +// CHECK3-NEXT: [[TMP11:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1 +// CHECK3-NEXT: store ptr null, ptr [[TMP11]], align 4 +// CHECK3-NEXT: [[TMP12:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2 +// CHECK3-NEXT: store i32 [[TMP5]], ptr [[TMP12]], align 4 +// CHECK3-NEXT: [[TMP13:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 2 +// CHECK3-NEXT: store i32 [[TMP5]], ptr [[TMP13]], align 4 +// CHECK3-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2 +// CHECK3-NEXT: store ptr null, ptr [[TMP14]], align 4 +// CHECK3-NEXT: [[TMP15:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 +// CHECK3-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0 +// CHECK3-NEXT: [[TMP17:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4 +// CHECK3-NEXT: [[TMP18:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0 +// CHECK3-NEXT: store i32 3, ptr [[TMP18]], align 4 +// CHECK3-NEXT: [[TMP19:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1 +// CHECK3-NEXT: store i32 3, ptr [[TMP19]], align 4 +// CHECK3-NEXT: [[TMP20:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2 +// CHECK3-NEXT: store ptr [[TMP15]], ptr [[TMP20]], align 4 +// CHECK3-NEXT: [[TMP21:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3 +// CHECK3-NEXT: store ptr [[TMP16]], ptr [[TMP21]], align 4 +// CHECK3-NEXT: [[TMP22:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4 +// CHECK3-NEXT: store ptr @.offload_sizes, ptr [[TMP22]], align 4 +// CHECK3-NEXT: [[TMP23:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5 +// CHECK3-NEXT: store ptr @.offload_maptypes, ptr [[TMP23]], align 4 +// CHECK3-NEXT: [[TMP24:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6 +// CHECK3-NEXT: store ptr null, ptr [[TMP24]], align 4 +// CHECK3-NEXT: [[TMP25:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7 +// CHECK3-NEXT: store ptr null, ptr [[TMP25]], align 4 +// CHECK3-NEXT: [[TMP26:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8 +// CHECK3-NEXT: store i64 0, ptr [[TMP26]], align 8 +// CHECK3-NEXT: [[TMP27:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9 +// CHECK3-NEXT: store i64 4, ptr [[TMP27]], align 8 +// CHECK3-NEXT: [[TMP28:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10 +// CHECK3-NEXT: store [3 x i32] zeroinitializer, ptr [[TMP28]], align 4 +// CHECK3-NEXT: [[TMP29:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11 +// CHECK3-NEXT: store [3 x i32] zeroinitializer, ptr [[TMP29]], align 4 +// CHECK3-NEXT: [[TMP30:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12 +// CHECK3-NEXT: store i32 [[TMP17]], ptr [[TMP30]], align 4 +// CHECK3-NEXT: [[TMP31:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB1:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l88.region_id, ptr [[KERNEL_ARGS]]) +// CHECK3-NEXT: [[TMP32:%.*]] = icmp ne i32 [[TMP31]], 0 +// CHECK3-NEXT: br i1 [[TMP32]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]] +// CHECK3: omp_offload.failed: +// CHECK3-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l88(ptr [[THIS1]], i32 [[TMP3]], i32 [[TMP5]]) #[[ATTR2:[0-9]+]] +// CHECK3-NEXT: br label [[OMP_OFFLOAD_CONT]] +// CHECK3: omp_offload.cont: +// CHECK3-NEXT: [[A2:%.*]] = getelementptr inbounds nuw [[STRUCT_S1]], ptr [[THIS1]], i32 0, i32 0 +// CHECK3-NEXT: [[TMP33:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0 +// CHECK3-NEXT: store ptr [[THIS1]], ptr [[TMP33]], align 4 +// CHECK3-NEXT: [[TMP34:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS4]], i32 0, i32 0 +// CHECK3-NEXT: store ptr [[A2]], ptr [[TMP34]], align 4 +// CHECK3-NEXT: [[TMP35:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS5]], i32 0, i32 0 +// CHECK3-NEXT: store ptr null, ptr [[TMP35]], align 4 +// CHECK3-NEXT: [[TMP36:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0 +// CHECK3-NEXT: [[TMP37:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS4]], i32 0, i32 0 +// CHECK3-NEXT: [[TMP38:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS6]], i32 0, i32 0 +// CHECK3-NEXT: store i32 3, ptr [[TMP38]], align 4 +// CHECK3-NEXT: [[TMP39:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS6]], i32 0, i32 1 +// CHECK3-NEXT: store i32 1, ptr [[TMP39]], align 4 +// CHECK3-NEXT: [[TMP40:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS6]], i32 0, i32 2 +// CHECK3-NEXT: store ptr [[TMP36]], ptr [[TMP40]], align 4 +// CHECK3-NEXT: [[TMP41:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS6]], i32 0, i32 3 +// CHECK3-NEXT: store ptr [[TMP37]], ptr [[TMP41]], align 4 +// CHECK3-NEXT: [[TMP42:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS6]], i32 0, i32 4 +// CHECK3-NEXT: store ptr @.offload_sizes.1, ptr [[TMP42]], align 4 +// CHECK3-NEXT: [[TMP43:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS6]], i32 0, i32 5 +// CHECK3-NEXT: store ptr @.offload_maptypes.2, ptr [[TMP43]], align 4 +// CHECK3-NEXT: [[TMP44:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS6]], i32 0, i32 6 +// CHECK3-NEXT: store ptr null, ptr [[TMP44]], align 4 +// CHECK3-NEXT: [[TMP45:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS6]], i32 0, i32 7 +// CHECK3-NEXT: store ptr null, ptr [[TMP45]], align 4 +// CHECK3-NEXT: [[TMP46:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS6]], i32 0, i32 8 +// CHECK3-NEXT: store i64 0, ptr [[TMP46]], align 8 +// CHECK3-NEXT: [[TMP47:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS6]], i32 0, i32 9 +// CHECK3-NEXT: store i64 0, ptr [[TMP47]], align 8 +// CHECK3-NEXT: [[TMP48:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS6]], i32 0, i32 10 +// CHECK3-NEXT: store [3 x i32] [i32 -1, i32 0, i32 0], ptr [[TMP48]], align 4 +// CHECK3-NEXT: [[TMP49:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS6]], i32 0, i32 11 +// CHECK3-NEXT: store [3 x i32] zeroinitializer, ptr [[TMP49]], align 4 +// CHECK3-NEXT: [[TMP50:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS6]], i32 0, i32 12 +// CHECK3-NEXT: store i32 1024, ptr [[TMP50]], align 4 +// CHECK3-NEXT: [[TMP51:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB1]], i64 -1, i32 -1, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l93.region_id, ptr [[KERNEL_ARGS6]]) +// CHECK3-NEXT: [[TMP52:%.*]] = icmp ne i32 [[TMP51]], 0 +// CHECK3-NEXT: br i1 [[TMP52]], label [[OMP_OFFLOAD_FAILED7:%.*]], label [[OMP_OFFLOAD_CONT8:%.*]] +// CHECK3: omp_offload.failed7: +// CHECK3-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l93(ptr [[THIS1]]) #[[ATTR2]] +// CHECK3-NEXT: br label [[OMP_OFFLOAD_CONT8]] +// CHECK3: omp_offload.cont8: +// CHECK3-NEXT: [[A9:%.*]] = getelementptr inbounds nuw [[STRUCT_S1]], ptr [[THIS1]], i32 0, i32 0 +// CHECK3-NEXT: [[TMP53:%.*]] = load double, ptr [[A9]], align 4 +// CHECK3-NEXT: [[CONV:%.*]] = fptosi double [[TMP53]] to i32 +// CHECK3-NEXT: ret i32 [[CONV]] +// +// +// CHECK3-LABEL: define {{[^@]+}}@_ZL7fstatici +// CHECK3-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0]] { +// CHECK3-NEXT: entry: +// CHECK3-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: [[N_CASTED:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: [[DOTCAPTURE_EXPR__CASTED2:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x ptr], align 4 +// CHECK3-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x ptr], align 4 +// CHECK3-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x ptr], align 4 +// CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: [[DOTCAPTURE_EXPR_3:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: [[DOTCAPTURE_EXPR_4:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8 +// CHECK3-NEXT: [[DOTCAPTURE_EXPR_6:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: [[DOTCAPTURE_EXPR__CASTED8:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: [[DOTOFFLOAD_BASEPTRS9:%.*]] = alloca [1 x ptr], align 4 +// CHECK3-NEXT: [[DOTOFFLOAD_PTRS10:%.*]] = alloca [1 x ptr], align 4 +// CHECK3-NEXT: [[DOTOFFLOAD_MAPPERS11:%.*]] = alloca [1 x ptr], align 4 +// CHECK3-NEXT: [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 4 +// CHECK3-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]]) +// CHECK3-NEXT: store i32 [[N]], ptr [[N_ADDR]], align 4 +// CHECK3-NEXT: [[TMP1:%.*]] = load i32, ptr [[N_ADDR]], align 4 +// CHECK3-NEXT: store i32 [[TMP1]], ptr [[DOTCAPTURE_EXPR_]], align 4 +// CHECK3-NEXT: [[TMP2:%.*]] = load i32, ptr [[N_ADDR]], align 4 +// CHECK3-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP2]], 32 +// CHECK3-NEXT: store i32 [[MUL]], ptr [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK3-NEXT: [[TMP3:%.*]] = load i32, ptr [[N_ADDR]], align 4 +// CHECK3-NEXT: store i32 [[TMP3]], ptr [[N_CASTED]], align 4 +// CHECK3-NEXT: [[TMP4:%.*]] = load i32, ptr [[N_CASTED]], align 4 +// CHECK3-NEXT: [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4 +// CHECK3-NEXT: store i32 [[TMP5]], ptr [[DOTCAPTURE_EXPR__CASTED]], align 4 +// CHECK3-NEXT: [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__CASTED]], align 4 +// CHECK3-NEXT: [[TMP7:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK3-NEXT: store i32 [[TMP7]], ptr [[DOTCAPTURE_EXPR__CASTED2]], align 4 +// CHECK3-NEXT: [[TMP8:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__CASTED2]], align 4 +// CHECK3-NEXT: [[TMP9:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 +// CHECK3-NEXT: store i32 [[TMP4]], ptr [[TMP9]], align 4 +// CHECK3-NEXT: [[TMP10:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0 +// CHECK3-NEXT: store i32 [[TMP4]], ptr [[TMP10]], align 4 +// CHECK3-NEXT: [[TMP11:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0 +// CHECK3-NEXT: store ptr null, ptr [[TMP11]], align 4 +// CHECK3-NEXT: [[TMP12:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1 +// CHECK3-NEXT: store i32 [[TMP6]], ptr [[TMP12]], align 4 +// CHECK3-NEXT: [[TMP13:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 1 +// CHECK3-NEXT: store i32 [[TMP6]], ptr [[TMP13]], align 4 +// CHECK3-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1 +// CHECK3-NEXT: store ptr null, ptr [[TMP14]], align 4 +// CHECK3-NEXT: [[TMP15:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2 +// CHECK3-NEXT: store i32 [[TMP8]], ptr [[TMP15]], align 4 +// CHECK3-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 2 +// CHECK3-NEXT: store i32 [[TMP8]], ptr [[TMP16]], align 4 +// CHECK3-NEXT: [[TMP17:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2 +// CHECK3-NEXT: store ptr null, ptr [[TMP17]], align 4 +// CHECK3-NEXT: [[TMP18:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 +// CHECK3-NEXT: [[TMP19:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0 +// CHECK3-NEXT: [[TMP20:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4 +// CHECK3-NEXT: [[TMP21:%.*]] = load i32, ptr [[N_ADDR]], align 4 +// CHECK3-NEXT: store i32 [[TMP21]], ptr [[DOTCAPTURE_EXPR_3]], align 4 +// CHECK3-NEXT: [[TMP22:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_3]], align 4 +// CHECK3-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP22]], 0 +// CHECK3-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1 +// CHECK3-NEXT: [[SUB5:%.*]] = sub nsw i32 [[DIV]], 1 +// CHECK3-NEXT: store i32 [[SUB5]], ptr [[DOTCAPTURE_EXPR_4]], align 4 +// CHECK3-NEXT: [[TMP23:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4 +// CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP23]], 1 +// CHECK3-NEXT: [[TMP24:%.*]] = zext i32 [[ADD]] to i64 +// CHECK3-NEXT: [[TMP25:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK3-NEXT: [[TMP26:%.*]] = insertvalue [3 x i32] zeroinitializer, i32 [[TMP20]], 0 +// CHECK3-NEXT: [[TMP27:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0 +// CHECK3-NEXT: store i32 3, ptr [[TMP27]], align 4 +// CHECK3-NEXT: [[TMP28:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1 +// CHECK3-NEXT: store i32 3, ptr [[TMP28]], align 4 +// CHECK3-NEXT: [[TMP29:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2 +// CHECK3-NEXT: store ptr [[TMP18]], ptr [[TMP29]], align 4 +// CHECK3-NEXT: [[TMP30:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3 +// CHECK3-NEXT: store ptr [[TMP19]], ptr [[TMP30]], align 4 +// CHECK3-NEXT: [[TMP31:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4 +// CHECK3-NEXT: store ptr @.offload_sizes.3, ptr [[TMP31]], align 4 +// CHECK3-NEXT: [[TMP32:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5 +// CHECK3-NEXT: store ptr @.offload_maptypes.4, ptr [[TMP32]], align 4 +// CHECK3-NEXT: [[TMP33:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6 +// CHECK3-NEXT: store ptr null, ptr [[TMP33]], align 4 +// CHECK3-NEXT: [[TMP34:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7 +// CHECK3-NEXT: store ptr null, ptr [[TMP34]], align 4 +// CHECK3-NEXT: [[TMP35:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8 +// CHECK3-NEXT: store i64 [[TMP24]], ptr [[TMP35]], align 8 +// CHECK3-NEXT: [[TMP36:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9 +// CHECK3-NEXT: store i64 8, ptr [[TMP36]], align 8 +// CHECK3-NEXT: [[TMP37:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10 +// CHECK3-NEXT: store [3 x i32] [[TMP26]], ptr [[TMP37]], align 4 +// CHECK3-NEXT: [[TMP38:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11 +// CHECK3-NEXT: store [3 x i32] zeroinitializer, ptr [[TMP38]], align 4 +// CHECK3-NEXT: [[TMP39:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12 +// CHECK3-NEXT: store i32 [[TMP25]], ptr [[TMP39]], align 4 +// CHECK3-NEXT: [[TMP40:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB1]], i64 -1, i32 [[TMP20]], i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l71.region_id, ptr [[KERNEL_ARGS]]) +// CHECK3-NEXT: [[TMP41:%.*]] = icmp ne i32 [[TMP40]], 0 +// CHECK3-NEXT: br i1 [[TMP41]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]] +// CHECK3: omp_offload.failed: +// CHECK3-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l71(i32 [[TMP4]], i32 [[TMP6]], i32 [[TMP8]]) #[[ATTR2]] +// CHECK3-NEXT: br label [[OMP_OFFLOAD_CONT]] +// CHECK3: omp_offload.cont: +// CHECK3-NEXT: [[TMP42:%.*]] = load i32, ptr [[N_ADDR]], align 4 +// CHECK3-NEXT: [[ADD7:%.*]] = add nsw i32 32, [[TMP42]] +// CHECK3-NEXT: store i32 [[ADD7]], ptr [[DOTCAPTURE_EXPR_6]], align 4 +// CHECK3-NEXT: [[TMP43:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_6]], align 4 +// CHECK3-NEXT: store i32 [[TMP43]], ptr [[DOTCAPTURE_EXPR__CASTED8]], align 4 +// CHECK3-NEXT: [[TMP44:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__CASTED8]], align 4 +// CHECK3-NEXT: [[TMP45:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS9]], i32 0, i32 0 +// CHECK3-NEXT: store i32 [[TMP44]], ptr [[TMP45]], align 4 +// CHECK3-NEXT: [[TMP46:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS10]], i32 0, i32 0 +// CHECK3-NEXT: store i32 [[TMP44]], ptr [[TMP46]], align 4 +// CHECK3-NEXT: [[TMP47:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS11]], i32 0, i32 0 +// CHECK3-NEXT: store ptr null, ptr [[TMP47]], align 4 +// CHECK3-NEXT: [[TMP48:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS9]], i32 0, i32 0 +// CHECK3-NEXT: [[TMP49:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS10]], i32 0, i32 0 +// CHECK3-NEXT: [[TMP50:%.*]] = getelementptr inbounds nuw [[STRUCT_ANON]], ptr [[AGG_CAPTURED]], i32 0, i32 0 +// CHECK3-NEXT: [[TMP51:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_6]], align 4 +// CHECK3-NEXT: store i32 [[TMP51]], ptr [[TMP50]], align 4 +// CHECK3-NEXT: [[TMP52:%.*]] = call ptr @__kmpc_omp_target_task_alloc(ptr @[[GLOB1]], i32 [[TMP0]], i32 1, i32 36, i32 4, ptr @.omp_task_entry., i64 -1) +// CHECK3-NEXT: [[TMP53:%.*]] = getelementptr inbounds nuw [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], ptr [[TMP52]], i32 0, i32 0 +// CHECK3-NEXT: [[TMP54:%.*]] = getelementptr inbounds nuw [[STRUCT_KMP_TASK_T:%.*]], ptr [[TMP53]], i32 0, i32 0 +// CHECK3-NEXT: [[TMP55:%.*]] = load ptr, ptr [[TMP54]], align 4 +// CHECK3-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[TMP55]], ptr align 4 [[AGG_CAPTURED]], i32 4, i1 false) +// CHECK3-NEXT: [[TMP56:%.*]] = getelementptr inbounds nuw [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], ptr [[TMP52]], i32 0, i32 1 +// CHECK3-NEXT: [[TMP57:%.*]] = getelementptr inbounds nuw [[STRUCT__KMP_PRIVATES_T:%.*]], ptr [[TMP56]], i32 0, i32 0 +// CHECK3-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[TMP57]], ptr align 4 @.offload_sizes.5, i32 8, i1 false) +// CHECK3-NEXT: [[TMP58:%.*]] = getelementptr inbounds nuw [[STRUCT__KMP_PRIVATES_T]], ptr [[TMP56]], i32 0, i32 1 +// CHECK3-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[TMP58]], ptr align 4 [[TMP48]], i32 4, i1 false) +// CHECK3-NEXT: [[TMP59:%.*]] = getelementptr inbounds nuw [[STRUCT__KMP_PRIVATES_T]], ptr [[TMP56]], i32 0, i32 2 +// CHECK3-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[TMP59]], ptr align 4 [[TMP49]], i32 4, i1 false) +// CHECK3-NEXT: [[TMP60:%.*]] = call i32 @__kmpc_omp_task(ptr @[[GLOB1]], i32 [[TMP0]], ptr [[TMP52]]) +// CHECK3-NEXT: [[TMP61:%.*]] = load i32, ptr [[N_ADDR]], align 4 +// CHECK3-NEXT: [[ADD12:%.*]] = add nsw i32 [[TMP61]], 1 +// CHECK3-NEXT: ret i32 [[ADD12]] +// +// +// CHECK3-LABEL: define {{[^@]+}}@_Z9ftemplateIiET_i +// CHECK3-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0]] comdat { +// CHECK3-NEXT: entry: +// CHECK3-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: [[A:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8 +// CHECK3-NEXT: [[B:%.*]] = alloca i16, align 2 +// CHECK3-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i16, align 2 +// CHECK3-NEXT: [[A_CASTED:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: [[B_CASTED:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x ptr], align 4 +// CHECK3-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x ptr], align 4 +// CHECK3-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x ptr], align 4 +// CHECK3-NEXT: [[KERNEL_ARGS1:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8 +// CHECK3-NEXT: store i32 [[N]], ptr [[N_ADDR]], align 4 +// CHECK3-NEXT: store i32 0, ptr [[A]], align 4 +// CHECK3-NEXT: [[TMP0:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0 +// CHECK3-NEXT: store i32 3, ptr [[TMP0]], align 4 +// CHECK3-NEXT: [[TMP1:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1 +// CHECK3-NEXT: store i32 0, ptr [[TMP1]], align 4 +// CHECK3-NEXT: [[TMP2:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2 +// CHECK3-NEXT: store ptr null, ptr [[TMP2]], align 4 +// CHECK3-NEXT: [[TMP3:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3 +// CHECK3-NEXT: store ptr null, ptr [[TMP3]], align 4 +// CHECK3-NEXT: [[TMP4:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4 +// CHECK3-NEXT: store ptr null, ptr [[TMP4]], align 4 +// CHECK3-NEXT: [[TMP5:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5 +// CHECK3-NEXT: store ptr null, ptr [[TMP5]], align 4 +// CHECK3-NEXT: [[TMP6:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6 +// CHECK3-NEXT: store ptr null, ptr [[TMP6]], align 4 +// CHECK3-NEXT: [[TMP7:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7 +// CHECK3-NEXT: store ptr null, ptr [[TMP7]], align 4 +// CHECK3-NEXT: [[TMP8:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8 +// CHECK3-NEXT: store i64 0, ptr [[TMP8]], align 8 +// CHECK3-NEXT: [[TMP9:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9 +// CHECK3-NEXT: store i64 8, ptr [[TMP9]], align 8 +// CHECK3-NEXT: [[TMP10:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10 +// CHECK3-NEXT: store [3 x i32] zeroinitializer, ptr [[TMP10]], align 4 +// CHECK3-NEXT: [[TMP11:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11 +// CHECK3-NEXT: store [3 x i32] zeroinitializer, ptr [[TMP11]], align 4 +// CHECK3-NEXT: [[TMP12:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12 +// CHECK3-NEXT: store i32 20, ptr [[TMP12]], align 4 +// CHECK3-NEXT: [[TMP13:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB1]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l55.region_id, ptr [[KERNEL_ARGS]]) +// CHECK3-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0 +// CHECK3-NEXT: br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]] +// CHECK3: omp_offload.failed: +// CHECK3-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l55() #[[ATTR2]] +// CHECK3-NEXT: br label [[OMP_OFFLOAD_CONT]] +// CHECK3: omp_offload.cont: +// CHECK3-NEXT: store i16 1, ptr [[B]], align 2 +// CHECK3-NEXT: [[TMP15:%.*]] = load i16, ptr [[B]], align 2 +// CHECK3-NEXT: store i16 [[TMP15]], ptr [[DOTCAPTURE_EXPR_]], align 2 +// CHECK3-NEXT: [[TMP16:%.*]] = load i32, ptr [[A]], align 4 +// CHECK3-NEXT: store i32 [[TMP16]], ptr [[A_CASTED]], align 4 +// CHECK3-NEXT: [[TMP17:%.*]] = load i32, ptr [[A_CASTED]], align 4 +// CHECK3-NEXT: [[TMP18:%.*]] = load i16, ptr [[B]], align 2 +// CHECK3-NEXT: store i16 [[TMP18]], ptr [[B_CASTED]], align 2 +// CHECK3-NEXT: [[TMP19:%.*]] = load i32, ptr [[B_CASTED]], align 4 +// CHECK3-NEXT: [[TMP20:%.*]] = load i16, ptr [[DOTCAPTURE_EXPR_]], align 2 +// CHECK3-NEXT: store i16 [[TMP20]], ptr [[DOTCAPTURE_EXPR__CASTED]], align 2 +// CHECK3-NEXT: [[TMP21:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__CASTED]], align 4 +// CHECK3-NEXT: [[TMP22:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 +// CHECK3-NEXT: store i32 [[TMP17]], ptr [[TMP22]], align 4 +// CHECK3-NEXT: [[TMP23:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0 +// CHECK3-NEXT: store i32 [[TMP17]], ptr [[TMP23]], align 4 +// CHECK3-NEXT: [[TMP24:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0 +// CHECK3-NEXT: store ptr null, ptr [[TMP24]], align 4 +// CHECK3-NEXT: [[TMP25:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1 +// CHECK3-NEXT: store i32 [[TMP19]], ptr [[TMP25]], align 4 +// CHECK3-NEXT: [[TMP26:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 1 +// CHECK3-NEXT: store i32 [[TMP19]], ptr [[TMP26]], align 4 +// CHECK3-NEXT: [[TMP27:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1 +// CHECK3-NEXT: store ptr null, ptr [[TMP27]], align 4 +// CHECK3-NEXT: [[TMP28:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2 +// CHECK3-NEXT: store i32 [[TMP21]], ptr [[TMP28]], align 4 +// CHECK3-NEXT: [[TMP29:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 2 +// CHECK3-NEXT: store i32 [[TMP21]], ptr [[TMP29]], align 4 +// CHECK3-NEXT: [[TMP30:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2 +// CHECK3-NEXT: store ptr null, ptr [[TMP30]], align 4 +// CHECK3-NEXT: [[TMP31:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 +// CHECK3-NEXT: [[TMP32:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0 +// CHECK3-NEXT: [[TMP33:%.*]] = load i16, ptr [[DOTCAPTURE_EXPR_]], align 2 +// CHECK3-NEXT: [[TMP34:%.*]] = sext i16 [[TMP33]] to i32 +// CHECK3-NEXT: [[TMP35:%.*]] = insertvalue [3 x i32] zeroinitializer, i32 [[TMP34]], 0 +// CHECK3-NEXT: [[TMP36:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS1]], i32 0, i32 0 +// CHECK3-NEXT: store i32 3, ptr [[TMP36]], align 4 +// CHECK3-NEXT: [[TMP37:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS1]], i32 0, i32 1 +// CHECK3-NEXT: store i32 3, ptr [[TMP37]], align 4 +// CHECK3-NEXT: [[TMP38:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS1]], i32 0, i32 2 +// CHECK3-NEXT: store ptr [[TMP31]], ptr [[TMP38]], align 4 +// CHECK3-NEXT: [[TMP39:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS1]], i32 0, i32 3 +// CHECK3-NEXT: store ptr [[TMP32]], ptr [[TMP39]], align 4 +// CHECK3-NEXT: [[TMP40:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS1]], i32 0, i32 4 +// CHECK3-NEXT: store ptr @.offload_sizes.7, ptr [[TMP40]], align 4 +// CHECK3-NEXT: [[TMP41:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS1]], i32 0, i32 5 +// CHECK3-NEXT: store ptr @.offload_maptypes.8, ptr [[TMP41]], align 4 +// CHECK3-NEXT: [[TMP42:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS1]], i32 0, i32 6 +// CHECK3-NEXT: store ptr null, ptr [[TMP42]], align 4 +// CHECK3-NEXT: [[TMP43:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS1]], i32 0, i32 7 +// CHECK3-NEXT: store ptr null, ptr [[TMP43]], align 4 +// CHECK3-NEXT: [[TMP44:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS1]], i32 0, i32 8 +// CHECK3-NEXT: store i64 0, ptr [[TMP44]], align 8 +// CHECK3-NEXT: [[TMP45:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS1]], i32 0, i32 9 +// CHECK3-NEXT: store i64 8, ptr [[TMP45]], align 8 +// CHECK3-NEXT: [[TMP46:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS1]], i32 0, i32 10 +// CHECK3-NEXT: store [3 x i32] [[TMP35]], ptr [[TMP46]], align 4 +// CHECK3-NEXT: [[TMP47:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS1]], i32 0, i32 11 +// CHECK3-NEXT: store [3 x i32] zeroinitializer, ptr [[TMP47]], align 4 +// CHECK3-NEXT: [[TMP48:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS1]], i32 0, i32 12 +// CHECK3-NEXT: store i32 1024, ptr [[TMP48]], align 4 +// CHECK3-NEXT: [[TMP49:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB1]], i64 -1, i32 [[TMP34]], i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l60.region_id, ptr [[KERNEL_ARGS1]]) +// CHECK3-NEXT: [[TMP50:%.*]] = icmp ne i32 [[TMP49]], 0 +// CHECK3-NEXT: br i1 [[TMP50]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]] +// CHECK3: omp_offload.failed2: +// CHECK3-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l60(i32 [[TMP17]], i32 [[TMP19]], i32 [[TMP21]]) #[[ATTR2]] +// CHECK3-NEXT: br label [[OMP_OFFLOAD_CONT3]] +// CHECK3: omp_offload.cont3: +// CHECK3-NEXT: [[TMP51:%.*]] = load i32, ptr [[A]], align 4 +// CHECK3-NEXT: ret i32 [[TMP51]] +// +// +// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l88 +// CHECK3-SAME: (ptr noundef [[THIS:%.*]], i32 noundef [[B:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1:[0-9]+]] { +// CHECK3-NEXT: entry: +// CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca ptr, align 4 +// CHECK3-NEXT: [[B_ADDR:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: [[B_CASTED:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: store ptr [[THIS]], ptr [[THIS_ADDR]], align 4 +// CHECK3-NEXT: store i32 [[B]], ptr [[B_ADDR]], align 4 +// CHECK3-NEXT: store i32 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 4 +// CHECK3-NEXT: [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4 +// CHECK3-NEXT: [[TMP1:%.*]] = load i32, ptr [[B_ADDR]], align 4 +// CHECK3-NEXT: store i32 [[TMP1]], ptr [[B_CASTED]], align 4 +// CHECK3-NEXT: [[TMP2:%.*]] = load i32, ptr [[B_CASTED]], align 4 +// CHECK3-NEXT: call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB1]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l88.omp_outlined, ptr [[TMP0]], i32 [[TMP2]]) +// CHECK3-NEXT: ret void +// +// +// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l88.omp_outlined +// CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef [[THIS:%.*]], i32 noundef [[B:%.*]]) #[[ATTR1]] { +// CHECK3-NEXT: entry: +// CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4 +// CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4 +// CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca ptr, align 4 +// CHECK3-NEXT: [[B_ADDR:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4 +// CHECK3-NEXT: store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4 +// CHECK3-NEXT: store ptr [[THIS]], ptr [[THIS_ADDR]], align 4 +// CHECK3-NEXT: store i32 [[B]], ptr [[B_ADDR]], align 4 +// CHECK3-NEXT: [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4 +// CHECK3-NEXT: [[TMP1:%.*]] = load i32, ptr [[B_ADDR]], align 4 +// CHECK3-NEXT: [[CONV:%.*]] = sitofp i32 [[TMP1]] to double +// CHECK3-NEXT: [[ADD:%.*]] = fadd double [[CONV]], 1.500000e+00 +// CHECK3-NEXT: [[A:%.*]] = getelementptr inbounds nuw [[STRUCT_S1:%.*]], ptr [[TMP0]], i32 0, i32 0 +// CHECK3-NEXT: store double [[ADD]], ptr [[A]], align 4 +// CHECK3-NEXT: ret void +// +// +// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l93 +// CHECK3-SAME: (ptr noundef [[THIS:%.*]]) #[[ATTR1]] { +// CHECK3-NEXT: entry: +// CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca ptr, align 4 +// CHECK3-NEXT: store ptr [[THIS]], ptr [[THIS_ADDR]], align 4 +// CHECK3-NEXT: [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4 +// CHECK3-NEXT: [[A:%.*]] = getelementptr inbounds nuw [[STRUCT_S1:%.*]], ptr [[TMP0]], i32 0, i32 0 +// CHECK3-NEXT: store double 2.500000e+00, ptr [[A]], align 4 +// CHECK3-NEXT: ret void +// +// +// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l71 +// CHECK3-SAME: (i32 noundef [[N:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]], i32 noundef [[DOTCAPTURE_EXPR_1:%.*]]) #[[ATTR1]] { +// CHECK3-NEXT: entry: +// CHECK3-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: [[DOTCAPTURE_EXPR__ADDR2:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: [[N_CASTED:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]]) +// CHECK3-NEXT: store i32 [[N]], ptr [[N_ADDR]], align 4 +// CHECK3-NEXT: store i32 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 4 +// CHECK3-NEXT: store i32 [[DOTCAPTURE_EXPR_1]], ptr [[DOTCAPTURE_EXPR__ADDR2]], align 4 +// CHECK3-NEXT: [[TMP1:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ADDR]], align 4 +// CHECK3-NEXT: call void @__kmpc_push_num_teams(ptr @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]], i32 0) +// CHECK3-NEXT: [[TMP2:%.*]] = load i32, ptr [[N_ADDR]], align 4 +// CHECK3-NEXT: store i32 [[TMP2]], ptr [[N_CASTED]], align 4 +// CHECK3-NEXT: [[TMP3:%.*]] = load i32, ptr [[N_CASTED]], align 4 +// CHECK3-NEXT: call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB1]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l71.omp_outlined, i32 [[TMP3]]) +// CHECK3-NEXT: ret void +// +// +// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l71.omp_outlined +// CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[N:%.*]]) #[[ATTR1]] { +// CHECK3-NEXT: entry: +// CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4 +// CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4 +// CHECK3-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: [[I:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: [[I3:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: [[N_CASTED:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4 +// CHECK3-NEXT: store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4 +// CHECK3-NEXT: store i32 [[N]], ptr [[N_ADDR]], align 4 +// CHECK3-NEXT: [[TMP0:%.*]] = load i32, ptr [[N_ADDR]], align 4 +// CHECK3-NEXT: store i32 [[TMP0]], ptr [[DOTCAPTURE_EXPR_]], align 4 +// CHECK3-NEXT: [[TMP1:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4 +// CHECK3-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP1]], 0 +// CHECK3-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1 +// CHECK3-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1 +// CHECK3-NEXT: store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK3-NEXT: store i32 0, ptr [[I]], align 4 +// CHECK3-NEXT: [[TMP2:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4 +// CHECK3-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP2]] +// CHECK3-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]] +// CHECK3: omp.precond.then: +// CHECK3-NEXT: store i32 0, ptr [[DOTOMP_COMB_LB]], align 4 +// CHECK3-NEXT: [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK3-NEXT: store i32 [[TMP3]], ptr [[DOTOMP_COMB_UB]], align 4 +// CHECK3-NEXT: store i32 1, ptr [[DOTOMP_STRIDE]], align 4 +// CHECK3-NEXT: store i32 0, ptr [[DOTOMP_IS_LAST]], align 4 +// CHECK3-NEXT: [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4 +// CHECK3-NEXT: [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4 +// CHECK3-NEXT: call void @__kmpc_for_static_init_4(ptr @[[GLOB2:[0-9]+]], i32 [[TMP5]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1) +// CHECK3-NEXT: [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4 +// CHECK3-NEXT: [[TMP7:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK3-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP6]], [[TMP7]] +// CHECK3-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] +// CHECK3: cond.true: +// CHECK3-NEXT: [[TMP8:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK3-NEXT: br label [[COND_END:%.*]] +// CHECK3: cond.false: +// CHECK3-NEXT: [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4 +// CHECK3-NEXT: br label [[COND_END]] +// CHECK3: cond.end: +// CHECK3-NEXT: [[COND:%.*]] = phi i32 [ [[TMP8]], [[COND_TRUE]] ], [ [[TMP9]], [[COND_FALSE]] ] +// CHECK3-NEXT: store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4 +// CHECK3-NEXT: [[TMP10:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4 +// CHECK3-NEXT: store i32 [[TMP10]], ptr [[DOTOMP_IV]], align 4 +// CHECK3-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] +// CHECK3: omp.inner.for.cond: +// CHECK3-NEXT: [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP18:![0-9]+]] +// CHECK3-NEXT: [[TMP12:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP18]] +// CHECK3-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]] +// CHECK3-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] +// CHECK3: omp.inner.for.body: +// CHECK3-NEXT: [[TMP13:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP18]] +// CHECK3-NEXT: [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP18]] +// CHECK3-NEXT: [[TMP15:%.*]] = load i32, ptr [[N_ADDR]], align 4, !llvm.access.group [[ACC_GRP18]] +// CHECK3-NEXT: store i32 [[TMP15]], ptr [[N_CASTED]], align 4, !llvm.access.group [[ACC_GRP18]] +// CHECK3-NEXT: [[TMP16:%.*]] = load i32, ptr [[N_CASTED]], align 4, !llvm.access.group [[ACC_GRP18]] +// CHECK3-NEXT: call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB1]], i32 3, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l71.omp_outlined.omp_outlined, i32 [[TMP13]], i32 [[TMP14]], i32 [[TMP16]]), !llvm.access.group [[ACC_GRP18]] +// CHECK3-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] +// CHECK3: omp.inner.for.inc: +// CHECK3-NEXT: [[TMP17:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP18]] +// CHECK3-NEXT: [[TMP18:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP18]] +// CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP17]], [[TMP18]] +// CHECK3-NEXT: store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP18]] +// CHECK3-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP19:![0-9]+]] +// CHECK3: omp.inner.for.end: +// CHECK3-NEXT: br label [[OMP_LOOP_EXIT:%.*]] +// CHECK3: omp.loop.exit: +// CHECK3-NEXT: [[TMP19:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4 +// CHECK3-NEXT: [[TMP20:%.*]] = load i32, ptr [[TMP19]], align 4 +// CHECK3-NEXT: call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP20]]) +// CHECK3-NEXT: [[TMP21:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4 +// CHECK3-NEXT: [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0 +// CHECK3-NEXT: br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]] +// CHECK3: .omp.final.then: +// CHECK3-NEXT: [[TMP23:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4 +// CHECK3-NEXT: [[SUB6:%.*]] = sub nsw i32 [[TMP23]], 0 +// CHECK3-NEXT: [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1 +// CHECK3-NEXT: [[MUL:%.*]] = mul nsw i32 [[DIV7]], 1 +// CHECK3-NEXT: [[ADD8:%.*]] = add nsw i32 0, [[MUL]] +// CHECK3-NEXT: store i32 [[ADD8]], ptr [[I3]], align 4 +// CHECK3-NEXT: br label [[DOTOMP_FINAL_DONE]] +// CHECK3: .omp.final.done: +// CHECK3-NEXT: br label [[OMP_PRECOND_END]] +// CHECK3: omp.precond.end: +// CHECK3-NEXT: ret void +// +// +// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l71.omp_outlined.omp_outlined +// CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], i32 noundef [[N:%.*]]) #[[ATTR1]] { +// CHECK3-NEXT: entry: +// CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4 +// CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4 +// CHECK3-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: [[I:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: [[I3:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4 +// CHECK3-NEXT: store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4 +// CHECK3-NEXT: store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4 +// CHECK3-NEXT: store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4 +// CHECK3-NEXT: store i32 [[N]], ptr [[N_ADDR]], align 4 +// CHECK3-NEXT: [[TMP0:%.*]] = load i32, ptr [[N_ADDR]], align 4 +// CHECK3-NEXT: store i32 [[TMP0]], ptr [[DOTCAPTURE_EXPR_]], align 4 +// CHECK3-NEXT: [[TMP1:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4 +// CHECK3-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP1]], 0 +// CHECK3-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1 +// CHECK3-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1 +// CHECK3-NEXT: store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK3-NEXT: store i32 0, ptr [[I]], align 4 +// CHECK3-NEXT: [[TMP2:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4 +// CHECK3-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP2]] +// CHECK3-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]] +// CHECK3: omp.precond.then: +// CHECK3-NEXT: store i32 0, ptr [[DOTOMP_LB]], align 4 +// CHECK3-NEXT: [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK3-NEXT: store i32 [[TMP3]], ptr [[DOTOMP_UB]], align 4 +// CHECK3-NEXT: [[TMP4:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4 +// CHECK3-NEXT: [[TMP5:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4 +// CHECK3-NEXT: store i32 [[TMP4]], ptr [[DOTOMP_LB]], align 4 +// CHECK3-NEXT: store i32 [[TMP5]], ptr [[DOTOMP_UB]], align 4 +// CHECK3-NEXT: store i32 1, ptr [[DOTOMP_STRIDE]], align 4 +// CHECK3-NEXT: store i32 0, ptr [[DOTOMP_IS_LAST]], align 4 +// CHECK3-NEXT: [[TMP6:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4 +// CHECK3-NEXT: [[TMP7:%.*]] = load i32, ptr [[TMP6]], align 4 +// CHECK3-NEXT: call void @__kmpc_for_static_init_4(ptr @[[GLOB3:[0-9]+]], i32 [[TMP7]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1) +// CHECK3-NEXT: [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4 +// CHECK3-NEXT: [[TMP9:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK3-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP8]], [[TMP9]] +// CHECK3-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] +// CHECK3: cond.true: +// CHECK3-NEXT: [[TMP10:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK3-NEXT: br label [[COND_END:%.*]] +// CHECK3: cond.false: +// CHECK3-NEXT: [[TMP11:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4 +// CHECK3-NEXT: br label [[COND_END]] +// CHECK3: cond.end: +// CHECK3-NEXT: [[COND:%.*]] = phi i32 [ [[TMP10]], [[COND_TRUE]] ], [ [[TMP11]], [[COND_FALSE]] ] +// CHECK3-NEXT: store i32 [[COND]], ptr [[DOTOMP_UB]], align 4 +// CHECK3-NEXT: [[TMP12:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4 +// CHECK3-NEXT: store i32 [[TMP12]], ptr [[DOTOMP_IV]], align 4 +// CHECK3-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] +// CHECK3: omp.inner.for.cond: +// CHECK3-NEXT: [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP22:![0-9]+]] +// CHECK3-NEXT: [[TMP14:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP22]] +// CHECK3-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]] +// CHECK3-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] +// CHECK3: omp.inner.for.body: +// CHECK3-NEXT: [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP22]] +// CHECK3-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1 +// CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]] +// CHECK3-NEXT: store i32 [[ADD]], ptr [[I3]], align 4, !llvm.access.group [[ACC_GRP22]] +// CHECK3-NEXT: br label [[OMP_BODY_CONTINUE:%.*]] +// CHECK3: omp.body.continue: +// CHECK3-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] +// CHECK3: omp.inner.for.inc: +// CHECK3-NEXT: [[TMP16:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP22]] +// CHECK3-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP16]], 1 +// CHECK3-NEXT: store i32 [[ADD6]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP22]] +// CHECK3-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP23:![0-9]+]] +// CHECK3: omp.inner.for.end: +// CHECK3-NEXT: br label [[OMP_LOOP_EXIT:%.*]] +// CHECK3: omp.loop.exit: +// CHECK3-NEXT: [[TMP17:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4 +// CHECK3-NEXT: [[TMP18:%.*]] = load i32, ptr [[TMP17]], align 4 +// CHECK3-NEXT: call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP18]]) +// CHECK3-NEXT: [[TMP19:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4 +// CHECK3-NEXT: [[TMP20:%.*]] = icmp ne i32 [[TMP19]], 0 +// CHECK3-NEXT: br i1 [[TMP20]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]] +// CHECK3: .omp.final.then: +// CHECK3-NEXT: [[TMP21:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4 +// CHECK3-NEXT: [[SUB7:%.*]] = sub nsw i32 [[TMP21]], 0 +// CHECK3-NEXT: [[DIV8:%.*]] = sdiv i32 [[SUB7]], 1 +// CHECK3-NEXT: [[MUL9:%.*]] = mul nsw i32 [[DIV8]], 1 +// CHECK3-NEXT: [[ADD10:%.*]] = add nsw i32 0, [[MUL9]] +// CHECK3-NEXT: store i32 [[ADD10]], ptr [[I3]], align 4 +// CHECK3-NEXT: br label [[DOTOMP_FINAL_DONE]] +// CHECK3: .omp.final.done: +// CHECK3-NEXT: br label [[OMP_PRECOND_END]] +// CHECK3: omp.precond.end: +// CHECK3-NEXT: ret void +// +// +// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l75 +// CHECK3-SAME: (i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] { +// CHECK3-NEXT: entry: +// CHECK3-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: store i32 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 4 +// CHECK3-NEXT: call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB1]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l75.omp_outlined) +// CHECK3-NEXT: ret void +// +// +// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l75.omp_outlined +// CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] { +// CHECK3-NEXT: entry: +// CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4 +// CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4 +// CHECK3-NEXT: store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4 +// CHECK3-NEXT: store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4 +// CHECK3-NEXT: ret void +// +// +// CHECK3-LABEL: define {{[^@]+}}@.omp_task_privates_map. +// CHECK3-SAME: (ptr noalias noundef [[TMP0:%.*]], ptr noalias noundef [[TMP1:%.*]], ptr noalias noundef [[TMP2:%.*]], ptr noalias noundef [[TMP3:%.*]]) #[[ATTR3:[0-9]+]] { +// CHECK3-NEXT: entry: +// CHECK3-NEXT: [[DOTADDR:%.*]] = alloca ptr, align 4 +// CHECK3-NEXT: [[DOTADDR1:%.*]] = alloca ptr, align 4 +// CHECK3-NEXT: [[DOTADDR2:%.*]] = alloca ptr, align 4 +// CHECK3-NEXT: [[DOTADDR3:%.*]] = alloca ptr, align 4 +// CHECK3-NEXT: store ptr [[TMP0]], ptr [[DOTADDR]], align 4 +// CHECK3-NEXT: store ptr [[TMP1]], ptr [[DOTADDR1]], align 4 +// CHECK3-NEXT: store ptr [[TMP2]], ptr [[DOTADDR2]], align 4 +// CHECK3-NEXT: store ptr [[TMP3]], ptr [[DOTADDR3]], align 4 +// CHECK3-NEXT: [[TMP4:%.*]] = load ptr, ptr [[DOTADDR]], align 4 +// CHECK3-NEXT: [[TMP5:%.*]] = getelementptr inbounds nuw [[STRUCT__KMP_PRIVATES_T:%.*]], ptr [[TMP4]], i32 0, i32 0 +// CHECK3-NEXT: [[TMP6:%.*]] = load ptr, ptr [[DOTADDR3]], align 4 +// CHECK3-NEXT: store ptr [[TMP5]], ptr [[TMP6]], align 4 +// CHECK3-NEXT: [[TMP7:%.*]] = getelementptr inbounds nuw [[STRUCT__KMP_PRIVATES_T]], ptr [[TMP4]], i32 0, i32 1 +// CHECK3-NEXT: [[TMP8:%.*]] = load ptr, ptr [[DOTADDR1]], align 4 +// CHECK3-NEXT: store ptr [[TMP7]], ptr [[TMP8]], align 4 +// CHECK3-NEXT: [[TMP9:%.*]] = getelementptr inbounds nuw [[STRUCT__KMP_PRIVATES_T]], ptr [[TMP4]], i32 0, i32 2 +// CHECK3-NEXT: [[TMP10:%.*]] = load ptr, ptr [[DOTADDR2]], align 4 +// CHECK3-NEXT: store ptr [[TMP9]], ptr [[TMP10]], align 4 +// CHECK3-NEXT: ret void +// +// +// CHECK3-LABEL: define {{[^@]+}}@.omp_task_entry. +// CHECK3-SAME: (i32 noundef [[TMP0:%.*]], ptr noalias noundef [[TMP1:%.*]]) #[[ATTR4:[0-9]+]] { +// CHECK3-NEXT: entry: +// CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: [[DOTPART_ID__ADDR_I:%.*]] = alloca ptr, align 4 +// CHECK3-NEXT: [[DOTPRIVATES__ADDR_I:%.*]] = alloca ptr, align 4 +// CHECK3-NEXT: [[DOTCOPY_FN__ADDR_I:%.*]] = alloca ptr, align 4 +// CHECK3-NEXT: [[DOTTASK_T__ADDR_I:%.*]] = alloca ptr, align 4 +// CHECK3-NEXT: [[__CONTEXT_ADDR_I:%.*]] = alloca ptr, align 4 +// CHECK3-NEXT: [[DOTFIRSTPRIV_PTR_ADDR_I:%.*]] = alloca ptr, align 4 +// CHECK3-NEXT: [[DOTFIRSTPRIV_PTR_ADDR1_I:%.*]] = alloca ptr, align 4 +// CHECK3-NEXT: [[DOTFIRSTPRIV_PTR_ADDR2_I:%.*]] = alloca ptr, align 4 +// CHECK3-NEXT: [[KERNEL_ARGS_I:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8 +// CHECK3-NEXT: [[DOTCAPTURE_EXPR__CASTED_I:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: [[DOTADDR:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: [[DOTADDR1:%.*]] = alloca ptr, align 4 +// CHECK3-NEXT: store i32 [[TMP0]], ptr [[DOTADDR]], align 4 +// CHECK3-NEXT: store ptr [[TMP1]], ptr [[DOTADDR1]], align 4 +// CHECK3-NEXT: [[TMP2:%.*]] = load i32, ptr [[DOTADDR]], align 4 +// CHECK3-NEXT: [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 4 +// CHECK3-NEXT: [[TMP4:%.*]] = getelementptr inbounds nuw [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], ptr [[TMP3]], i32 0, i32 0 +// CHECK3-NEXT: [[TMP5:%.*]] = getelementptr inbounds nuw [[STRUCT_KMP_TASK_T:%.*]], ptr [[TMP4]], i32 0, i32 2 +// CHECK3-NEXT: [[TMP6:%.*]] = getelementptr inbounds nuw [[STRUCT_KMP_TASK_T]], ptr [[TMP4]], i32 0, i32 0 +// CHECK3-NEXT: [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 4 +// CHECK3-NEXT: [[TMP8:%.*]] = getelementptr inbounds nuw [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], ptr [[TMP3]], i32 0, i32 1 +// CHECK3-NEXT: call void @llvm.experimental.noalias.scope.decl(metadata [[META25:![0-9]+]]) +// CHECK3-NEXT: call void @llvm.experimental.noalias.scope.decl(metadata [[META28:![0-9]+]]) +// CHECK3-NEXT: call void @llvm.experimental.noalias.scope.decl(metadata [[META30:![0-9]+]]) +// CHECK3-NEXT: call void @llvm.experimental.noalias.scope.decl(metadata [[META32:![0-9]+]]) +// CHECK3-NEXT: store i32 [[TMP2]], ptr [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias [[META34:![0-9]+]] +// CHECK3-NEXT: store ptr [[TMP5]], ptr [[DOTPART_ID__ADDR_I]], align 4, !noalias [[META34]] +// CHECK3-NEXT: store ptr [[TMP8]], ptr [[DOTPRIVATES__ADDR_I]], align 4, !noalias [[META34]] +// CHECK3-NEXT: store ptr @.omp_task_privates_map., ptr [[DOTCOPY_FN__ADDR_I]], align 4, !noalias [[META34]] +// CHECK3-NEXT: store ptr [[TMP3]], ptr [[DOTTASK_T__ADDR_I]], align 4, !noalias [[META34]] +// CHECK3-NEXT: store ptr [[TMP7]], ptr [[__CONTEXT_ADDR_I]], align 4, !noalias [[META34]] +// CHECK3-NEXT: [[TMP9:%.*]] = load ptr, ptr [[__CONTEXT_ADDR_I]], align 4, !noalias [[META34]] +// CHECK3-NEXT: [[TMP10:%.*]] = load ptr, ptr [[DOTCOPY_FN__ADDR_I]], align 4, !noalias [[META34]] +// CHECK3-NEXT: [[TMP11:%.*]] = load ptr, ptr [[DOTPRIVATES__ADDR_I]], align 4, !noalias [[META34]] +// CHECK3-NEXT: call void [[TMP10]](ptr [[TMP11]], ptr [[DOTFIRSTPRIV_PTR_ADDR_I]], ptr [[DOTFIRSTPRIV_PTR_ADDR1_I]], ptr [[DOTFIRSTPRIV_PTR_ADDR2_I]]) #[[ATTR2]] +// CHECK3-NEXT: [[TMP12:%.*]] = load ptr, ptr [[DOTFIRSTPRIV_PTR_ADDR_I]], align 4, !noalias [[META34]] +// CHECK3-NEXT: [[TMP13:%.*]] = load ptr, ptr [[DOTFIRSTPRIV_PTR_ADDR1_I]], align 4, !noalias [[META34]] +// CHECK3-NEXT: [[TMP14:%.*]] = load ptr, ptr [[DOTFIRSTPRIV_PTR_ADDR2_I]], align 4, !noalias [[META34]] +// CHECK3-NEXT: [[TMP15:%.*]] = load i32, ptr [[TMP9]], align 4 +// CHECK3-NEXT: store i32 3, ptr [[KERNEL_ARGS_I]], align 4, !noalias [[META34]] +// CHECK3-NEXT: [[TMP16:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS_I]], i32 0, i32 1 +// CHECK3-NEXT: store i32 1, ptr [[TMP16]], align 4, !noalias [[META34]] +// CHECK3-NEXT: [[TMP17:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS_I]], i32 0, i32 2 +// CHECK3-NEXT: store ptr [[TMP12]], ptr [[TMP17]], align 4, !noalias [[META34]] +// CHECK3-NEXT: [[TMP18:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS_I]], i32 0, i32 3 +// CHECK3-NEXT: store ptr [[TMP13]], ptr [[TMP18]], align 4, !noalias [[META34]] +// CHECK3-NEXT: [[TMP19:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS_I]], i32 0, i32 4 +// CHECK3-NEXT: store ptr [[TMP14]], ptr [[TMP19]], align 4, !noalias [[META34]] +// CHECK3-NEXT: [[TMP20:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS_I]], i32 0, i32 5 +// CHECK3-NEXT: store ptr @.offload_maptypes.6, ptr [[TMP20]], align 4, !noalias [[META34]] +// CHECK3-NEXT: [[TMP21:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS_I]], i32 0, i32 6 +// CHECK3-NEXT: store ptr null, ptr [[TMP21]], align 4, !noalias [[META34]] +// CHECK3-NEXT: [[TMP22:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS_I]], i32 0, i32 7 +// CHECK3-NEXT: store ptr null, ptr [[TMP22]], align 4, !noalias [[META34]] +// CHECK3-NEXT: [[TMP23:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS_I]], i32 0, i32 8 +// CHECK3-NEXT: store i64 0, ptr [[TMP23]], align 8, !noalias [[META34]] +// CHECK3-NEXT: [[TMP24:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS_I]], i32 0, i32 9 +// CHECK3-NEXT: store i64 9, ptr [[TMP24]], align 8, !noalias [[META34]] +// CHECK3-NEXT: [[TMP25:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS_I]], i32 0, i32 10 +// CHECK3-NEXT: store [3 x i32] zeroinitializer, ptr [[TMP25]], align 4, !noalias [[META34]] +// CHECK3-NEXT: [[TMP26:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS_I]], i32 0, i32 11 +// CHECK3-NEXT: store [3 x i32] zeroinitializer, ptr [[TMP26]], align 4, !noalias [[META34]] +// CHECK3-NEXT: [[TMP27:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS_I]], i32 0, i32 12 +// CHECK3-NEXT: store i32 [[TMP15]], ptr [[TMP27]], align 4, !noalias [[META34]] +// CHECK3-NEXT: [[TMP28:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB1]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l75.region_id, ptr [[KERNEL_ARGS_I]]) +// CHECK3-NEXT: [[TMP29:%.*]] = icmp ne i32 [[TMP28]], 0 +// CHECK3-NEXT: br i1 [[TMP29]], label [[OMP_OFFLOAD_FAILED_I:%.*]], label [[DOTOMP_OUTLINED__EXIT:%.*]] +// CHECK3: omp_offload.failed.i: +// CHECK3-NEXT: [[TMP30:%.*]] = load i32, ptr [[TMP9]], align 4 +// CHECK3-NEXT: store i32 [[TMP30]], ptr [[DOTCAPTURE_EXPR__CASTED_I]], align 4, !noalias [[META34]] +// CHECK3-NEXT: [[TMP31:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__CASTED_I]], align 4, !noalias [[META34]] +// CHECK3-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l75(i32 [[TMP31]]) #[[ATTR2]] +// CHECK3-NEXT: br label [[DOTOMP_OUTLINED__EXIT]] +// CHECK3: .omp_outlined..exit: +// CHECK3-NEXT: ret i32 0 +// +// +// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l55 +// CHECK3-SAME: () #[[ATTR1]] { +// CHECK3-NEXT: entry: +// CHECK3-NEXT: call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB1]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l55.omp_outlined) +// CHECK3-NEXT: ret void +// +// +// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l55.omp_outlined +// CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] { +// CHECK3-NEXT: entry: +// CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4 +// CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4 +// CHECK3-NEXT: store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4 +// CHECK3-NEXT: store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4 +// CHECK3-NEXT: ret void +// +// +// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l60 +// CHECK3-SAME: (i32 noundef [[A:%.*]], i32 noundef [[B:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] { +// CHECK3-NEXT: entry: +// CHECK3-NEXT: [[A_ADDR:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: [[B_ADDR:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: [[A_CASTED:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: [[B_CASTED:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]]) +// CHECK3-NEXT: store i32 [[A]], ptr [[A_ADDR]], align 4 +// CHECK3-NEXT: store i32 [[B]], ptr [[B_ADDR]], align 4 +// CHECK3-NEXT: store i32 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 4 +// CHECK3-NEXT: [[TMP1:%.*]] = load i16, ptr [[DOTCAPTURE_EXPR__ADDR]], align 2 +// CHECK3-NEXT: [[TMP2:%.*]] = sext i16 [[TMP1]] to i32 +// CHECK3-NEXT: call void @__kmpc_push_num_teams(ptr @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]], i32 0) +// CHECK3-NEXT: [[TMP3:%.*]] = load i32, ptr [[A_ADDR]], align 4 +// CHECK3-NEXT: store i32 [[TMP3]], ptr [[A_CASTED]], align 4 +// CHECK3-NEXT: [[TMP4:%.*]] = load i32, ptr [[A_CASTED]], align 4 +// CHECK3-NEXT: [[TMP5:%.*]] = load i16, ptr [[B_ADDR]], align 2 +// CHECK3-NEXT: store i16 [[TMP5]], ptr [[B_CASTED]], align 2 +// CHECK3-NEXT: [[TMP6:%.*]] = load i32, ptr [[B_CASTED]], align 4 +// CHECK3-NEXT: call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB1]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l60.omp_outlined, i32 [[TMP4]], i32 [[TMP6]]) +// CHECK3-NEXT: ret void +// +// +// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l60.omp_outlined +// CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]], i32 noundef [[B:%.*]]) #[[ATTR1]] { +// CHECK3-NEXT: entry: +// CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4 +// CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4 +// CHECK3-NEXT: [[A_ADDR:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: [[B_ADDR:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4 +// CHECK3-NEXT: store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4 +// CHECK3-NEXT: store i32 [[A]], ptr [[A_ADDR]], align 4 +// CHECK3-NEXT: store i32 [[B]], ptr [[B_ADDR]], align 4 +// CHECK3-NEXT: [[TMP0:%.*]] = load i16, ptr [[B_ADDR]], align 2 +// CHECK3-NEXT: [[CONV:%.*]] = sext i16 [[TMP0]] to i32 +// CHECK3-NEXT: [[TMP1:%.*]] = load i32, ptr [[A_ADDR]], align 4 +// CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CONV]] +// CHECK3-NEXT: store i32 [[ADD]], ptr [[A_ADDR]], align 4 +// CHECK3-NEXT: ret void +// +// +// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l71 +// CHECK9-SAME: (ptr noalias noundef [[DYN_PTR:%.*]], i64 noundef [[N:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]], i64 noundef [[DOTCAPTURE_EXPR_1:%.*]]) #[[ATTR0:[0-9]+]] { +// CHECK9-NEXT: entry: +// CHECK9-NEXT: [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8 +// CHECK9-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8 +// CHECK9-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8 +// CHECK9-NEXT: [[DOTCAPTURE_EXPR__ADDR2:%.*]] = alloca i64, align 8 +// CHECK9-NEXT: [[N_CASTED:%.*]] = alloca i64, align 8 +// CHECK9-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB3:[0-9]+]]) +// CHECK9-NEXT: store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8 +// CHECK9-NEXT: store i64 [[N]], ptr [[N_ADDR]], align 8 +// CHECK9-NEXT: store i64 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 8 +// CHECK9-NEXT: store i64 [[DOTCAPTURE_EXPR_1]], ptr [[DOTCAPTURE_EXPR__ADDR2]], align 8 +// CHECK9-NEXT: [[TMP1:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ADDR]], align 4 +// CHECK9-NEXT: call void @__kmpc_push_num_teams(ptr @[[GLOB3]], i32 [[TMP0]], i32 [[TMP1]], i32 0) +// CHECK9-NEXT: [[TMP2:%.*]] = load i32, ptr [[N_ADDR]], align 4 +// CHECK9-NEXT: store i32 [[TMP2]], ptr [[N_CASTED]], align 4 +// CHECK9-NEXT: [[TMP3:%.*]] = load i64, ptr [[N_CASTED]], align 8 +// CHECK9-NEXT: call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l71.omp_outlined, i64 [[TMP3]]) +// CHECK9-NEXT: ret void +// +// +// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l71.omp_outlined +// CHECK9-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[N:%.*]]) #[[ATTR0]] { +// CHECK9-NEXT: entry: +// CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8 +// CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8 +// CHECK9-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8 +// CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 +// CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4 +// CHECK9-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4 +// CHECK9-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4 +// CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4 +// CHECK9-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4 +// CHECK9-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4 +// CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 +// CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 +// CHECK9-NEXT: [[I3:%.*]] = alloca i32, align 4 +// CHECK9-NEXT: [[N_CASTED:%.*]] = alloca i64, align 8 +// CHECK9-NEXT: store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8 +// CHECK9-NEXT: store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8 +// CHECK9-NEXT: store i64 [[N]], ptr [[N_ADDR]], align 8 +// CHECK9-NEXT: [[TMP0:%.*]] = load i32, ptr [[N_ADDR]], align 4 +// CHECK9-NEXT: store i32 [[TMP0]], ptr [[DOTCAPTURE_EXPR_]], align 4 +// CHECK9-NEXT: [[TMP1:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4 +// CHECK9-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP1]], 0 +// CHECK9-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1 +// CHECK9-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1 +// CHECK9-NEXT: store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK9-NEXT: store i32 0, ptr [[I]], align 4 +// CHECK9-NEXT: [[TMP2:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4 +// CHECK9-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP2]] +// CHECK9-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]] +// CHECK9: omp.precond.then: +// CHECK9-NEXT: store i32 0, ptr [[DOTOMP_COMB_LB]], align 4 +// CHECK9-NEXT: [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK9-NEXT: store i32 [[TMP3]], ptr [[DOTOMP_COMB_UB]], align 4 +// CHECK9-NEXT: store i32 1, ptr [[DOTOMP_STRIDE]], align 4 +// CHECK9-NEXT: store i32 0, ptr [[DOTOMP_IS_LAST]], align 4 +// CHECK9-NEXT: [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8 +// CHECK9-NEXT: [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4 +// CHECK9-NEXT: call void @__kmpc_for_static_init_4(ptr @[[GLOB1:[0-9]+]], i32 [[TMP5]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1) +// CHECK9-NEXT: [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4 +// CHECK9-NEXT: [[TMP7:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK9-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP6]], [[TMP7]] +// CHECK9-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] +// CHECK9: cond.true: +// CHECK9-NEXT: [[TMP8:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK9-NEXT: br label [[COND_END:%.*]] +// CHECK9: cond.false: +// CHECK9-NEXT: [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4 +// CHECK9-NEXT: br label [[COND_END]] +// CHECK9: cond.end: +// CHECK9-NEXT: [[COND:%.*]] = phi i32 [ [[TMP8]], [[COND_TRUE]] ], [ [[TMP9]], [[COND_FALSE]] ] +// CHECK9-NEXT: store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4 +// CHECK9-NEXT: [[TMP10:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4 +// CHECK9-NEXT: store i32 [[TMP10]], ptr [[DOTOMP_IV]], align 4 +// CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] +// CHECK9: omp.inner.for.cond: +// CHECK9-NEXT: [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP16:![0-9]+]] +// CHECK9-NEXT: [[TMP12:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP16]] +// CHECK9-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]] +// CHECK9-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] +// CHECK9: omp.inner.for.body: +// CHECK9-NEXT: [[TMP13:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP16]] +// CHECK9-NEXT: [[TMP14:%.*]] = zext i32 [[TMP13]] to i64 +// CHECK9-NEXT: [[TMP15:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP16]] +// CHECK9-NEXT: [[TMP16:%.*]] = zext i32 [[TMP15]] to i64 +// CHECK9-NEXT: [[TMP17:%.*]] = load i32, ptr [[N_ADDR]], align 4, !llvm.access.group [[ACC_GRP16]] +// CHECK9-NEXT: store i32 [[TMP17]], ptr [[N_CASTED]], align 4, !llvm.access.group [[ACC_GRP16]] +// CHECK9-NEXT: [[TMP18:%.*]] = load i64, ptr [[N_CASTED]], align 8, !llvm.access.group [[ACC_GRP16]] +// CHECK9-NEXT: call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 3, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l71.omp_outlined.omp_outlined, i64 [[TMP14]], i64 [[TMP16]], i64 [[TMP18]]), !llvm.access.group [[ACC_GRP16]] +// CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] +// CHECK9: omp.inner.for.inc: +// CHECK9-NEXT: [[TMP19:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP16]] +// CHECK9-NEXT: [[TMP20:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP16]] +// CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP19]], [[TMP20]] +// CHECK9-NEXT: store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP16]] +// CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP17:![0-9]+]] +// CHECK9: omp.inner.for.end: +// CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]] +// CHECK9: omp.loop.exit: +// CHECK9-NEXT: [[TMP21:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8 +// CHECK9-NEXT: [[TMP22:%.*]] = load i32, ptr [[TMP21]], align 4 +// CHECK9-NEXT: call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP22]]) +// CHECK9-NEXT: [[TMP23:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4 +// CHECK9-NEXT: [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0 +// CHECK9-NEXT: br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]] +// CHECK9: .omp.final.then: +// CHECK9-NEXT: [[TMP25:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4 +// CHECK9-NEXT: [[SUB6:%.*]] = sub nsw i32 [[TMP25]], 0 +// CHECK9-NEXT: [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1 +// CHECK9-NEXT: [[MUL:%.*]] = mul nsw i32 [[DIV7]], 1 +// CHECK9-NEXT: [[ADD8:%.*]] = add nsw i32 0, [[MUL]] +// CHECK9-NEXT: store i32 [[ADD8]], ptr [[I3]], align 4 +// CHECK9-NEXT: br label [[DOTOMP_FINAL_DONE]] +// CHECK9: .omp.final.done: +// CHECK9-NEXT: br label [[OMP_PRECOND_END]] +// CHECK9: omp.precond.end: +// CHECK9-NEXT: ret void +// +// +// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l71.omp_outlined.omp_outlined +// CHECK9-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i64 noundef [[N:%.*]]) #[[ATTR0]] { +// CHECK9-NEXT: entry: +// CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8 +// CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8 +// CHECK9-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8 +// CHECK9-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8 +// CHECK9-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8 +// CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 +// CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4 +// CHECK9-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4 +// CHECK9-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4 +// CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4 +// CHECK9-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4 +// CHECK9-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4 +// CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 +// CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 +// CHECK9-NEXT: [[I4:%.*]] = alloca i32, align 4 +// CHECK9-NEXT: store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8 +// CHECK9-NEXT: store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8 +// CHECK9-NEXT: store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8 +// CHECK9-NEXT: store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8 +// CHECK9-NEXT: store i64 [[N]], ptr [[N_ADDR]], align 8 +// CHECK9-NEXT: [[TMP0:%.*]] = load i32, ptr [[N_ADDR]], align 4 +// CHECK9-NEXT: store i32 [[TMP0]], ptr [[DOTCAPTURE_EXPR_]], align 4 +// CHECK9-NEXT: [[TMP1:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4 +// CHECK9-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP1]], 0 +// CHECK9-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1 +// CHECK9-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1 +// CHECK9-NEXT: store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK9-NEXT: store i32 0, ptr [[I]], align 4 +// CHECK9-NEXT: [[TMP2:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4 +// CHECK9-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP2]] +// CHECK9-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]] +// CHECK9: omp.precond.then: +// CHECK9-NEXT: store i32 0, ptr [[DOTOMP_LB]], align 4 +// CHECK9-NEXT: [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK9-NEXT: store i32 [[TMP3]], ptr [[DOTOMP_UB]], align 4 +// CHECK9-NEXT: [[TMP4:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8 +// CHECK9-NEXT: [[CONV:%.*]] = trunc i64 [[TMP4]] to i32 +// CHECK9-NEXT: [[TMP5:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8 +// CHECK9-NEXT: [[CONV3:%.*]] = trunc i64 [[TMP5]] to i32 +// CHECK9-NEXT: store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4 +// CHECK9-NEXT: store i32 [[CONV3]], ptr [[DOTOMP_UB]], align 4 +// CHECK9-NEXT: store i32 1, ptr [[DOTOMP_STRIDE]], align 4 +// CHECK9-NEXT: store i32 0, ptr [[DOTOMP_IS_LAST]], align 4 +// CHECK9-NEXT: [[TMP6:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8 +// CHECK9-NEXT: [[TMP7:%.*]] = load i32, ptr [[TMP6]], align 4 +// CHECK9-NEXT: call void @__kmpc_for_static_init_4(ptr @[[GLOB2:[0-9]+]], i32 [[TMP7]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1) +// CHECK9-NEXT: [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4 +// CHECK9-NEXT: [[TMP9:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK9-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP8]], [[TMP9]] +// CHECK9-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] +// CHECK9: cond.true: +// CHECK9-NEXT: [[TMP10:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK9-NEXT: br label [[COND_END:%.*]] +// CHECK9: cond.false: +// CHECK9-NEXT: [[TMP11:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4 +// CHECK9-NEXT: br label [[COND_END]] +// CHECK9: cond.end: +// CHECK9-NEXT: [[COND:%.*]] = phi i32 [ [[TMP10]], [[COND_TRUE]] ], [ [[TMP11]], [[COND_FALSE]] ] +// CHECK9-NEXT: store i32 [[COND]], ptr [[DOTOMP_UB]], align 4 +// CHECK9-NEXT: [[TMP12:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4 +// CHECK9-NEXT: store i32 [[TMP12]], ptr [[DOTOMP_IV]], align 4 +// CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] +// CHECK9: omp.inner.for.cond: +// CHECK9-NEXT: [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP20:![0-9]+]] +// CHECK9-NEXT: [[TMP14:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP20]] +// CHECK9-NEXT: [[CMP6:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]] +// CHECK9-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] +// CHECK9: omp.inner.for.body: +// CHECK9-NEXT: [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP20]] +// CHECK9-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1 +// CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]] +// CHECK9-NEXT: store i32 [[ADD]], ptr [[I4]], align 4, !llvm.access.group [[ACC_GRP20]] +// CHECK9-NEXT: br label [[OMP_BODY_CONTINUE:%.*]] +// CHECK9: omp.body.continue: +// CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] +// CHECK9: omp.inner.for.inc: +// CHECK9-NEXT: [[TMP16:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP20]] +// CHECK9-NEXT: [[ADD7:%.*]] = add nsw i32 [[TMP16]], 1 +// CHECK9-NEXT: store i32 [[ADD7]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP20]] +// CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP21:![0-9]+]] +// CHECK9: omp.inner.for.end: +// CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]] +// CHECK9: omp.loop.exit: +// CHECK9-NEXT: [[TMP17:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8 +// CHECK9-NEXT: [[TMP18:%.*]] = load i32, ptr [[TMP17]], align 4 +// CHECK9-NEXT: call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP18]]) +// CHECK9-NEXT: [[TMP19:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4 +// CHECK9-NEXT: [[TMP20:%.*]] = icmp ne i32 [[TMP19]], 0 +// CHECK9-NEXT: br i1 [[TMP20]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]] +// CHECK9: .omp.final.then: +// CHECK9-NEXT: [[TMP21:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4 +// CHECK9-NEXT: [[SUB8:%.*]] = sub nsw i32 [[TMP21]], 0 +// CHECK9-NEXT: [[DIV9:%.*]] = sdiv i32 [[SUB8]], 1 +// CHECK9-NEXT: [[MUL10:%.*]] = mul nsw i32 [[DIV9]], 1 +// CHECK9-NEXT: [[ADD11:%.*]] = add nsw i32 0, [[MUL10]] +// CHECK9-NEXT: store i32 [[ADD11]], ptr [[I4]], align 4 +// CHECK9-NEXT: br label [[DOTOMP_FINAL_DONE]] +// CHECK9: .omp.final.done: +// CHECK9-NEXT: br label [[OMP_PRECOND_END]] +// CHECK9: omp.precond.end: +// CHECK9-NEXT: ret void +// +// +// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l75 +// CHECK9-SAME: (ptr noalias noundef [[DYN_PTR:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] { +// CHECK9-NEXT: entry: +// CHECK9-NEXT: [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8 +// CHECK9-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8 +// CHECK9-NEXT: store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8 +// CHECK9-NEXT: store i64 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 8 +// CHECK9-NEXT: call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l75.omp_outlined) +// CHECK9-NEXT: ret void +// +// +// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l75.omp_outlined +// CHECK9-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] { +// CHECK9-NEXT: entry: +// CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8 +// CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8 +// CHECK9-NEXT: store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8 +// CHECK9-NEXT: store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8 +// CHECK9-NEXT: ret void +// +// +// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l88 +// CHECK9-SAME: (ptr noalias noundef [[DYN_PTR:%.*]], ptr noundef [[THIS:%.*]], i64 noundef [[B:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] { +// CHECK9-NEXT: entry: +// CHECK9-NEXT: [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8 +// CHECK9-NEXT: [[THIS_ADDR:%.*]] = alloca ptr, align 8 +// CHECK9-NEXT: [[B_ADDR:%.*]] = alloca i64, align 8 +// CHECK9-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8 +// CHECK9-NEXT: [[B_CASTED:%.*]] = alloca i64, align 8 +// CHECK9-NEXT: store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8 +// CHECK9-NEXT: store ptr [[THIS]], ptr [[THIS_ADDR]], align 8 +// CHECK9-NEXT: store i64 [[B]], ptr [[B_ADDR]], align 8 +// CHECK9-NEXT: store i64 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 8 +// CHECK9-NEXT: [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8 +// CHECK9-NEXT: [[TMP1:%.*]] = load i32, ptr [[B_ADDR]], align 4 +// CHECK9-NEXT: store i32 [[TMP1]], ptr [[B_CASTED]], align 4 +// CHECK9-NEXT: [[TMP2:%.*]] = load i64, ptr [[B_CASTED]], align 8 +// CHECK9-NEXT: call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l88.omp_outlined, ptr [[TMP0]], i64 [[TMP2]]) +// CHECK9-NEXT: ret void +// +// +// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l88.omp_outlined +// CHECK9-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef [[THIS:%.*]], i64 noundef [[B:%.*]]) #[[ATTR0]] { +// CHECK9-NEXT: entry: +// CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8 +// CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8 +// CHECK9-NEXT: [[THIS_ADDR:%.*]] = alloca ptr, align 8 +// CHECK9-NEXT: [[B_ADDR:%.*]] = alloca i64, align 8 +// CHECK9-NEXT: store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8 +// CHECK9-NEXT: store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8 +// CHECK9-NEXT: store ptr [[THIS]], ptr [[THIS_ADDR]], align 8 +// CHECK9-NEXT: store i64 [[B]], ptr [[B_ADDR]], align 8 +// CHECK9-NEXT: [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8 +// CHECK9-NEXT: [[TMP1:%.*]] = load i32, ptr [[B_ADDR]], align 4 +// CHECK9-NEXT: [[CONV:%.*]] = sitofp i32 [[TMP1]] to double +// CHECK9-NEXT: [[ADD:%.*]] = fadd double [[CONV]], 1.500000e+00 +// CHECK9-NEXT: [[A:%.*]] = getelementptr inbounds nuw [[STRUCT_S1:%.*]], ptr [[TMP0]], i32 0, i32 0 +// CHECK9-NEXT: store double [[ADD]], ptr [[A]], align 8 +// CHECK9-NEXT: ret void +// +// +// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l93 +// CHECK9-SAME: (ptr noalias noundef [[DYN_PTR:%.*]], ptr noundef [[THIS:%.*]]) #[[ATTR0]] { +// CHECK9-NEXT: entry: +// CHECK9-NEXT: [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8 +// CHECK9-NEXT: [[THIS_ADDR:%.*]] = alloca ptr, align 8 +// CHECK9-NEXT: store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8 +// CHECK9-NEXT: store ptr [[THIS]], ptr [[THIS_ADDR]], align 8 +// CHECK9-NEXT: [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8 +// CHECK9-NEXT: [[A:%.*]] = getelementptr inbounds nuw [[STRUCT_S1:%.*]], ptr [[TMP0]], i32 0, i32 0 +// CHECK9-NEXT: store double 2.500000e+00, ptr [[A]], align 8 +// CHECK9-NEXT: ret void +// +// +// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l55 +// CHECK9-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] { +// CHECK9-NEXT: entry: +// CHECK9-NEXT: [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8 +// CHECK9-NEXT: store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8 +// CHECK9-NEXT: call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l55.omp_outlined) +// CHECK9-NEXT: ret void +// +// +// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l55.omp_outlined +// CHECK9-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] { +// CHECK9-NEXT: entry: +// CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8 +// CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8 +// CHECK9-NEXT: store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8 +// CHECK9-NEXT: store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8 +// CHECK9-NEXT: ret void +// +// +// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l60 +// CHECK9-SAME: (ptr noalias noundef [[DYN_PTR:%.*]], i64 noundef [[A:%.*]], i64 noundef [[B:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] { +// CHECK9-NEXT: entry: +// CHECK9-NEXT: [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8 +// CHECK9-NEXT: [[A_ADDR:%.*]] = alloca i64, align 8 +// CHECK9-NEXT: [[B_ADDR:%.*]] = alloca i64, align 8 +// CHECK9-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8 +// CHECK9-NEXT: [[A_CASTED:%.*]] = alloca i64, align 8 +// CHECK9-NEXT: [[B_CASTED:%.*]] = alloca i64, align 8 +// CHECK9-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB3]]) +// CHECK9-NEXT: store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8 +// CHECK9-NEXT: store i64 [[A]], ptr [[A_ADDR]], align 8 +// CHECK9-NEXT: store i64 [[B]], ptr [[B_ADDR]], align 8 +// CHECK9-NEXT: store i64 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 8 +// CHECK9-NEXT: [[TMP1:%.*]] = load i16, ptr [[DOTCAPTURE_EXPR__ADDR]], align 2 +// CHECK9-NEXT: [[TMP2:%.*]] = sext i16 [[TMP1]] to i32 +// CHECK9-NEXT: call void @__kmpc_push_num_teams(ptr @[[GLOB3]], i32 [[TMP0]], i32 [[TMP2]], i32 0) +// CHECK9-NEXT: [[TMP3:%.*]] = load i32, ptr [[A_ADDR]], align 4 +// CHECK9-NEXT: store i32 [[TMP3]], ptr [[A_CASTED]], align 4 +// CHECK9-NEXT: [[TMP4:%.*]] = load i64, ptr [[A_CASTED]], align 8 +// CHECK9-NEXT: [[TMP5:%.*]] = load i16, ptr [[B_ADDR]], align 2 +// CHECK9-NEXT: store i16 [[TMP5]], ptr [[B_CASTED]], align 2 +// CHECK9-NEXT: [[TMP6:%.*]] = load i64, ptr [[B_CASTED]], align 8 +// CHECK9-NEXT: call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l60.omp_outlined, i64 [[TMP4]], i64 [[TMP6]]) +// CHECK9-NEXT: ret void +// +// +// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l60.omp_outlined +// CHECK9-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]], i64 noundef [[B:%.*]]) #[[ATTR0]] { +// CHECK9-NEXT: entry: +// CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8 +// CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8 +// CHECK9-NEXT: [[A_ADDR:%.*]] = alloca i64, align 8 +// CHECK9-NEXT: [[B_ADDR:%.*]] = alloca i64, align 8 +// CHECK9-NEXT: store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8 +// CHECK9-NEXT: store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8 +// CHECK9-NEXT: store i64 [[A]], ptr [[A_ADDR]], align 8 +// CHECK9-NEXT: store i64 [[B]], ptr [[B_ADDR]], align 8 +// CHECK9-NEXT: [[TMP0:%.*]] = load i16, ptr [[B_ADDR]], align 2 +// CHECK9-NEXT: [[CONV:%.*]] = sext i16 [[TMP0]] to i32 +// CHECK9-NEXT: [[TMP1:%.*]] = load i32, ptr [[A_ADDR]], align 4 +// CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CONV]] +// CHECK9-NEXT: store i32 [[ADD]], ptr [[A_ADDR]], align 4 +// CHECK9-NEXT: ret void +// +// +// CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l71 +// CHECK11-SAME: (ptr noalias noundef [[DYN_PTR:%.*]], i32 noundef [[N:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]], i32 noundef [[DOTCAPTURE_EXPR_1:%.*]]) #[[ATTR0:[0-9]+]] { +// CHECK11-NEXT: entry: +// CHECK11-NEXT: [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4 +// CHECK11-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4 +// CHECK11-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4 +// CHECK11-NEXT: [[DOTCAPTURE_EXPR__ADDR2:%.*]] = alloca i32, align 4 +// CHECK11-NEXT: [[N_CASTED:%.*]] = alloca i32, align 4 +// CHECK11-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB3:[0-9]+]]) +// CHECK11-NEXT: store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4 +// CHECK11-NEXT: store i32 [[N]], ptr [[N_ADDR]], align 4 +// CHECK11-NEXT: store i32 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 4 +// CHECK11-NEXT: store i32 [[DOTCAPTURE_EXPR_1]], ptr [[DOTCAPTURE_EXPR__ADDR2]], align 4 +// CHECK11-NEXT: [[TMP1:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ADDR]], align 4 +// CHECK11-NEXT: call void @__kmpc_push_num_teams(ptr @[[GLOB3]], i32 [[TMP0]], i32 [[TMP1]], i32 0) +// CHECK11-NEXT: [[TMP2:%.*]] = load i32, ptr [[N_ADDR]], align 4 +// CHECK11-NEXT: store i32 [[TMP2]], ptr [[N_CASTED]], align 4 +// CHECK11-NEXT: [[TMP3:%.*]] = load i32, ptr [[N_CASTED]], align 4 +// CHECK11-NEXT: call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l71.omp_outlined, i32 [[TMP3]]) +// CHECK11-NEXT: ret void +// +// +// CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l71.omp_outlined +// CHECK11-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[N:%.*]]) #[[ATTR0]] { +// CHECK11-NEXT: entry: +// CHECK11-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4 +// CHECK11-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4 +// CHECK11-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4 +// CHECK11-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 +// CHECK11-NEXT: [[TMP:%.*]] = alloca i32, align 4 +// CHECK11-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4 +// CHECK11-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4 +// CHECK11-NEXT: [[I:%.*]] = alloca i32, align 4 +// CHECK11-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4 +// CHECK11-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4 +// CHECK11-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 +// CHECK11-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 +// CHECK11-NEXT: [[I3:%.*]] = alloca i32, align 4 +// CHECK11-NEXT: [[N_CASTED:%.*]] = alloca i32, align 4 +// CHECK11-NEXT: store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4 +// CHECK11-NEXT: store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4 +// CHECK11-NEXT: store i32 [[N]], ptr [[N_ADDR]], align 4 +// CHECK11-NEXT: [[TMP0:%.*]] = load i32, ptr [[N_ADDR]], align 4 +// CHECK11-NEXT: store i32 [[TMP0]], ptr [[DOTCAPTURE_EXPR_]], align 4 +// CHECK11-NEXT: [[TMP1:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4 +// CHECK11-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP1]], 0 +// CHECK11-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1 +// CHECK11-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1 +// CHECK11-NEXT: store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK11-NEXT: store i32 0, ptr [[I]], align 4 +// CHECK11-NEXT: [[TMP2:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4 +// CHECK11-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP2]] +// CHECK11-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]] +// CHECK11: omp.precond.then: +// CHECK11-NEXT: store i32 0, ptr [[DOTOMP_COMB_LB]], align 4 +// CHECK11-NEXT: [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK11-NEXT: store i32 [[TMP3]], ptr [[DOTOMP_COMB_UB]], align 4 +// CHECK11-NEXT: store i32 1, ptr [[DOTOMP_STRIDE]], align 4 +// CHECK11-NEXT: store i32 0, ptr [[DOTOMP_IS_LAST]], align 4 +// CHECK11-NEXT: [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4 +// CHECK11-NEXT: [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4 +// CHECK11-NEXT: call void @__kmpc_for_static_init_4(ptr @[[GLOB1:[0-9]+]], i32 [[TMP5]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1) +// CHECK11-NEXT: [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4 +// CHECK11-NEXT: [[TMP7:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK11-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP6]], [[TMP7]] +// CHECK11-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] +// CHECK11: cond.true: +// CHECK11-NEXT: [[TMP8:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK11-NEXT: br label [[COND_END:%.*]] +// CHECK11: cond.false: +// CHECK11-NEXT: [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4 +// CHECK11-NEXT: br label [[COND_END]] +// CHECK11: cond.end: +// CHECK11-NEXT: [[COND:%.*]] = phi i32 [ [[TMP8]], [[COND_TRUE]] ], [ [[TMP9]], [[COND_FALSE]] ] +// CHECK11-NEXT: store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4 +// CHECK11-NEXT: [[TMP10:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4 +// CHECK11-NEXT: store i32 [[TMP10]], ptr [[DOTOMP_IV]], align 4 +// CHECK11-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] +// CHECK11: omp.inner.for.cond: +// CHECK11-NEXT: [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP17:![0-9]+]] +// CHECK11-NEXT: [[TMP12:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP17]] +// CHECK11-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]] +// CHECK11-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] +// CHECK11: omp.inner.for.body: +// CHECK11-NEXT: [[TMP13:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP17]] +// CHECK11-NEXT: [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP17]] +// CHECK11-NEXT: [[TMP15:%.*]] = load i32, ptr [[N_ADDR]], align 4, !llvm.access.group [[ACC_GRP17]] +// CHECK11-NEXT: store i32 [[TMP15]], ptr [[N_CASTED]], align 4, !llvm.access.group [[ACC_GRP17]] +// CHECK11-NEXT: [[TMP16:%.*]] = load i32, ptr [[N_CASTED]], align 4, !llvm.access.group [[ACC_GRP17]] +// CHECK11-NEXT: call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 3, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l71.omp_outlined.omp_outlined, i32 [[TMP13]], i32 [[TMP14]], i32 [[TMP16]]), !llvm.access.group [[ACC_GRP17]] +// CHECK11-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] +// CHECK11: omp.inner.for.inc: +// CHECK11-NEXT: [[TMP17:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP17]] +// CHECK11-NEXT: [[TMP18:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP17]] +// CHECK11-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP17]], [[TMP18]] +// CHECK11-NEXT: store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP17]] +// CHECK11-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP18:![0-9]+]] +// CHECK11: omp.inner.for.end: +// CHECK11-NEXT: br label [[OMP_LOOP_EXIT:%.*]] +// CHECK11: omp.loop.exit: +// CHECK11-NEXT: [[TMP19:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4 +// CHECK11-NEXT: [[TMP20:%.*]] = load i32, ptr [[TMP19]], align 4 +// CHECK11-NEXT: call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP20]]) +// CHECK11-NEXT: [[TMP21:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4 +// CHECK11-NEXT: [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0 +// CHECK11-NEXT: br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]] +// CHECK11: .omp.final.then: +// CHECK11-NEXT: [[TMP23:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4 +// CHECK11-NEXT: [[SUB6:%.*]] = sub nsw i32 [[TMP23]], 0 +// CHECK11-NEXT: [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1 +// CHECK11-NEXT: [[MUL:%.*]] = mul nsw i32 [[DIV7]], 1 +// CHECK11-NEXT: [[ADD8:%.*]] = add nsw i32 0, [[MUL]] +// CHECK11-NEXT: store i32 [[ADD8]], ptr [[I3]], align 4 +// CHECK11-NEXT: br label [[DOTOMP_FINAL_DONE]] +// CHECK11: .omp.final.done: +// CHECK11-NEXT: br label [[OMP_PRECOND_END]] +// CHECK11: omp.precond.end: +// CHECK11-NEXT: ret void +// +// +// CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l71.omp_outlined.omp_outlined +// CHECK11-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], i32 noundef [[N:%.*]]) #[[ATTR0]] { +// CHECK11-NEXT: entry: +// CHECK11-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4 +// CHECK11-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4 +// CHECK11-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4 +// CHECK11-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4 +// CHECK11-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4 +// CHECK11-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 +// CHECK11-NEXT: [[TMP:%.*]] = alloca i32, align 4 +// CHECK11-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4 +// CHECK11-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4 +// CHECK11-NEXT: [[I:%.*]] = alloca i32, align 4 +// CHECK11-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4 +// CHECK11-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4 +// CHECK11-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 +// CHECK11-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 +// CHECK11-NEXT: [[I3:%.*]] = alloca i32, align 4 +// CHECK11-NEXT: store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4 +// CHECK11-NEXT: store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4 +// CHECK11-NEXT: store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4 +// CHECK11-NEXT: store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4 +// CHECK11-NEXT: store i32 [[N]], ptr [[N_ADDR]], align 4 +// CHECK11-NEXT: [[TMP0:%.*]] = load i32, ptr [[N_ADDR]], align 4 +// CHECK11-NEXT: store i32 [[TMP0]], ptr [[DOTCAPTURE_EXPR_]], align 4 +// CHECK11-NEXT: [[TMP1:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4 +// CHECK11-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP1]], 0 +// CHECK11-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1 +// CHECK11-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1 +// CHECK11-NEXT: store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK11-NEXT: store i32 0, ptr [[I]], align 4 +// CHECK11-NEXT: [[TMP2:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4 +// CHECK11-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP2]] +// CHECK11-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]] +// CHECK11: omp.precond.then: +// CHECK11-NEXT: store i32 0, ptr [[DOTOMP_LB]], align 4 +// CHECK11-NEXT: [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK11-NEXT: store i32 [[TMP3]], ptr [[DOTOMP_UB]], align 4 +// CHECK11-NEXT: [[TMP4:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4 +// CHECK11-NEXT: [[TMP5:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4 +// CHECK11-NEXT: store i32 [[TMP4]], ptr [[DOTOMP_LB]], align 4 +// CHECK11-NEXT: store i32 [[TMP5]], ptr [[DOTOMP_UB]], align 4 +// CHECK11-NEXT: store i32 1, ptr [[DOTOMP_STRIDE]], align 4 +// CHECK11-NEXT: store i32 0, ptr [[DOTOMP_IS_LAST]], align 4 +// CHECK11-NEXT: [[TMP6:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4 +// CHECK11-NEXT: [[TMP7:%.*]] = load i32, ptr [[TMP6]], align 4 +// CHECK11-NEXT: call void @__kmpc_for_static_init_4(ptr @[[GLOB2:[0-9]+]], i32 [[TMP7]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1) +// CHECK11-NEXT: [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4 +// CHECK11-NEXT: [[TMP9:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK11-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP8]], [[TMP9]] +// CHECK11-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] +// CHECK11: cond.true: +// CHECK11-NEXT: [[TMP10:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK11-NEXT: br label [[COND_END:%.*]] +// CHECK11: cond.false: +// CHECK11-NEXT: [[TMP11:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4 +// CHECK11-NEXT: br label [[COND_END]] +// CHECK11: cond.end: +// CHECK11-NEXT: [[COND:%.*]] = phi i32 [ [[TMP10]], [[COND_TRUE]] ], [ [[TMP11]], [[COND_FALSE]] ] +// CHECK11-NEXT: store i32 [[COND]], ptr [[DOTOMP_UB]], align 4 +// CHECK11-NEXT: [[TMP12:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4 +// CHECK11-NEXT: store i32 [[TMP12]], ptr [[DOTOMP_IV]], align 4 +// CHECK11-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] +// CHECK11: omp.inner.for.cond: +// CHECK11-NEXT: [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP21:![0-9]+]] +// CHECK11-NEXT: [[TMP14:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP21]] +// CHECK11-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]] +// CHECK11-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] +// CHECK11: omp.inner.for.body: +// CHECK11-NEXT: [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP21]] +// CHECK11-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1 +// CHECK11-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]] +// CHECK11-NEXT: store i32 [[ADD]], ptr [[I3]], align 4, !llvm.access.group [[ACC_GRP21]] +// CHECK11-NEXT: br label [[OMP_BODY_CONTINUE:%.*]] +// CHECK11: omp.body.continue: +// CHECK11-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] +// CHECK11: omp.inner.for.inc: +// CHECK11-NEXT: [[TMP16:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP21]] +// CHECK11-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP16]], 1 +// CHECK11-NEXT: store i32 [[ADD6]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP21]] +// CHECK11-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP22:![0-9]+]] +// CHECK11: omp.inner.for.end: +// CHECK11-NEXT: br label [[OMP_LOOP_EXIT:%.*]] +// CHECK11: omp.loop.exit: +// CHECK11-NEXT: [[TMP17:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4 +// CHECK11-NEXT: [[TMP18:%.*]] = load i32, ptr [[TMP17]], align 4 +// CHECK11-NEXT: call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP18]]) +// CHECK11-NEXT: [[TMP19:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4 +// CHECK11-NEXT: [[TMP20:%.*]] = icmp ne i32 [[TMP19]], 0 +// CHECK11-NEXT: br i1 [[TMP20]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]] +// CHECK11: .omp.final.then: +// CHECK11-NEXT: [[TMP21:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4 +// CHECK11-NEXT: [[SUB7:%.*]] = sub nsw i32 [[TMP21]], 0 +// CHECK11-NEXT: [[DIV8:%.*]] = sdiv i32 [[SUB7]], 1 +// CHECK11-NEXT: [[MUL9:%.*]] = mul nsw i32 [[DIV8]], 1 +// CHECK11-NEXT: [[ADD10:%.*]] = add nsw i32 0, [[MUL9]] +// CHECK11-NEXT: store i32 [[ADD10]], ptr [[I3]], align 4 +// CHECK11-NEXT: br label [[DOTOMP_FINAL_DONE]] +// CHECK11: .omp.final.done: +// CHECK11-NEXT: br label [[OMP_PRECOND_END]] +// CHECK11: omp.precond.end: +// CHECK11-NEXT: ret void +// +// +// CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l75 +// CHECK11-SAME: (ptr noalias noundef [[DYN_PTR:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] { +// CHECK11-NEXT: entry: +// CHECK11-NEXT: [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4 +// CHECK11-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4 +// CHECK11-NEXT: store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4 +// CHECK11-NEXT: store i32 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 4 +// CHECK11-NEXT: call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l75.omp_outlined) +// CHECK11-NEXT: ret void +// +// +// CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l75.omp_outlined +// CHECK11-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] { +// CHECK11-NEXT: entry: +// CHECK11-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4 +// CHECK11-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4 +// CHECK11-NEXT: store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4 +// CHECK11-NEXT: store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4 +// CHECK11-NEXT: ret void +// +// +// CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l88 +// CHECK11-SAME: (ptr noalias noundef [[DYN_PTR:%.*]], ptr noundef [[THIS:%.*]], i32 noundef [[B:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] { +// CHECK11-NEXT: entry: +// CHECK11-NEXT: [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4 +// CHECK11-NEXT: [[THIS_ADDR:%.*]] = alloca ptr, align 4 +// CHECK11-NEXT: [[B_ADDR:%.*]] = alloca i32, align 4 +// CHECK11-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4 +// CHECK11-NEXT: [[B_CASTED:%.*]] = alloca i32, align 4 +// CHECK11-NEXT: store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4 +// CHECK11-NEXT: store ptr [[THIS]], ptr [[THIS_ADDR]], align 4 +// CHECK11-NEXT: store i32 [[B]], ptr [[B_ADDR]], align 4 +// CHECK11-NEXT: store i32 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 4 +// CHECK11-NEXT: [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4 +// CHECK11-NEXT: [[TMP1:%.*]] = load i32, ptr [[B_ADDR]], align 4 +// CHECK11-NEXT: store i32 [[TMP1]], ptr [[B_CASTED]], align 4 +// CHECK11-NEXT: [[TMP2:%.*]] = load i32, ptr [[B_CASTED]], align 4 +// CHECK11-NEXT: call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l88.omp_outlined, ptr [[TMP0]], i32 [[TMP2]]) +// CHECK11-NEXT: ret void +// +// +// CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l88.omp_outlined +// CHECK11-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef [[THIS:%.*]], i32 noundef [[B:%.*]]) #[[ATTR0]] { +// CHECK11-NEXT: entry: +// CHECK11-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4 +// CHECK11-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4 +// CHECK11-NEXT: [[THIS_ADDR:%.*]] = alloca ptr, align 4 +// CHECK11-NEXT: [[B_ADDR:%.*]] = alloca i32, align 4 +// CHECK11-NEXT: store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4 +// CHECK11-NEXT: store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4 +// CHECK11-NEXT: store ptr [[THIS]], ptr [[THIS_ADDR]], align 4 +// CHECK11-NEXT: store i32 [[B]], ptr [[B_ADDR]], align 4 +// CHECK11-NEXT: [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4 +// CHECK11-NEXT: [[TMP1:%.*]] = load i32, ptr [[B_ADDR]], align 4 +// CHECK11-NEXT: [[CONV:%.*]] = sitofp i32 [[TMP1]] to double +// CHECK11-NEXT: [[ADD:%.*]] = fadd double [[CONV]], 1.500000e+00 +// CHECK11-NEXT: [[A:%.*]] = getelementptr inbounds nuw [[STRUCT_S1:%.*]], ptr [[TMP0]], i32 0, i32 0 +// CHECK11-NEXT: store double [[ADD]], ptr [[A]], align 4 +// CHECK11-NEXT: ret void +// +// +// CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l93 +// CHECK11-SAME: (ptr noalias noundef [[DYN_PTR:%.*]], ptr noundef [[THIS:%.*]]) #[[ATTR0]] { +// CHECK11-NEXT: entry: +// CHECK11-NEXT: [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4 +// CHECK11-NEXT: [[THIS_ADDR:%.*]] = alloca ptr, align 4 +// CHECK11-NEXT: store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4 +// CHECK11-NEXT: store ptr [[THIS]], ptr [[THIS_ADDR]], align 4 +// CHECK11-NEXT: [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4 +// CHECK11-NEXT: [[A:%.*]] = getelementptr inbounds nuw [[STRUCT_S1:%.*]], ptr [[TMP0]], i32 0, i32 0 +// CHECK11-NEXT: store double 2.500000e+00, ptr [[A]], align 4 +// CHECK11-NEXT: ret void +// +// +// CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l55 +// CHECK11-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] { +// CHECK11-NEXT: entry: +// CHECK11-NEXT: [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4 +// CHECK11-NEXT: store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4 +// CHECK11-NEXT: call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l55.omp_outlined) +// CHECK11-NEXT: ret void +// +// +// CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l55.omp_outlined +// CHECK11-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] { +// CHECK11-NEXT: entry: +// CHECK11-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4 +// CHECK11-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4 +// CHECK11-NEXT: store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4 +// CHECK11-NEXT: store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4 +// CHECK11-NEXT: ret void +// +// +// CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l60 +// CHECK11-SAME: (ptr noalias noundef [[DYN_PTR:%.*]], i32 noundef [[A:%.*]], i32 noundef [[B:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] { +// CHECK11-NEXT: entry: +// CHECK11-NEXT: [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4 +// CHECK11-NEXT: [[A_ADDR:%.*]] = alloca i32, align 4 +// CHECK11-NEXT: [[B_ADDR:%.*]] = alloca i32, align 4 +// CHECK11-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4 +// CHECK11-NEXT: [[A_CASTED:%.*]] = alloca i32, align 4 +// CHECK11-NEXT: [[B_CASTED:%.*]] = alloca i32, align 4 +// CHECK11-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB3]]) +// CHECK11-NEXT: store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4 +// CHECK11-NEXT: store i32 [[A]], ptr [[A_ADDR]], align 4 +// CHECK11-NEXT: store i32 [[B]], ptr [[B_ADDR]], align 4 +// CHECK11-NEXT: store i32 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 4 +// CHECK11-NEXT: [[TMP1:%.*]] = load i16, ptr [[DOTCAPTURE_EXPR__ADDR]], align 2 +// CHECK11-NEXT: [[TMP2:%.*]] = sext i16 [[TMP1]] to i32 +// CHECK11-NEXT: call void @__kmpc_push_num_teams(ptr @[[GLOB3]], i32 [[TMP0]], i32 [[TMP2]], i32 0) +// CHECK11-NEXT: [[TMP3:%.*]] = load i32, ptr [[A_ADDR]], align 4 +// CHECK11-NEXT: store i32 [[TMP3]], ptr [[A_CASTED]], align 4 +// CHECK11-NEXT: [[TMP4:%.*]] = load i32, ptr [[A_CASTED]], align 4 +// CHECK11-NEXT: [[TMP5:%.*]] = load i16, ptr [[B_ADDR]], align 2 +// CHECK11-NEXT: store i16 [[TMP5]], ptr [[B_CASTED]], align 2 +// CHECK11-NEXT: [[TMP6:%.*]] = load i32, ptr [[B_CASTED]], align 4 +// CHECK11-NEXT: call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l60.omp_outlined, i32 [[TMP4]], i32 [[TMP6]]) +// CHECK11-NEXT: ret void +// +// +// CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l60.omp_outlined +// CHECK11-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]], i32 noundef [[B:%.*]]) #[[ATTR0]] { +// CHECK11-NEXT: entry: +// CHECK11-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4 +// CHECK11-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4 +// CHECK11-NEXT: [[A_ADDR:%.*]] = alloca i32, align 4 +// CHECK11-NEXT: [[B_ADDR:%.*]] = alloca i32, align 4 +// CHECK11-NEXT: store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4 +// CHECK11-NEXT: store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4 +// CHECK11-NEXT: store i32 [[A]], ptr [[A_ADDR]], align 4 +// CHECK11-NEXT: store i32 [[B]], ptr [[B_ADDR]], align 4 +// CHECK11-NEXT: [[TMP0:%.*]] = load i16, ptr [[B_ADDR]], align 2 +// CHECK11-NEXT: [[CONV:%.*]] = sext i16 [[TMP0]] to i32 +// CHECK11-NEXT: [[TMP1:%.*]] = load i32, ptr [[A_ADDR]], align 4 +// CHECK11-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CONV]] +// CHECK11-NEXT: store i32 [[ADD]], ptr [[A_ADDR]], align 4 +// CHECK11-NEXT: ret void +// From b76e32c4b08fcd3285cb74271c7dae9495207263 Mon Sep 17 00:00:00 2001 From: Kevin Sala Date: Sat, 1 Nov 2025 20:51:11 -0700 Subject: [PATCH 16/17] Fix review comments --- clang/lib/Sema/SemaOpenMP.cpp | 10 +++++++--- clang/lib/Serialization/ASTReader.cpp | 5 ++--- 2 files changed, 9 insertions(+), 6 deletions(-) diff --git a/clang/lib/Sema/SemaOpenMP.cpp b/clang/lib/Sema/SemaOpenMP.cpp index e8855103b061f..15d9f711a46e8 100644 --- a/clang/lib/Sema/SemaOpenMP.cpp +++ b/clang/lib/Sema/SemaOpenMP.cpp @@ -17697,7 +17697,8 @@ OMPClause *SemaOpenMP::ActOnOpenMPSingleExprWithArgClause( static_cast(Argument[ScheduleKind]), Expr, StartLoc, LParenLoc, ArgumentLoc[Modifier1], ArgumentLoc[Modifier2], ArgumentLoc[ScheduleKind], DelimLoc, EndLoc); - } break; + break; + } case OMPC_if: assert(Argument.size() == 1 && ArgumentLoc.size() == 1); Res = ActOnOpenMPIfClause(static_cast(Argument.back()), @@ -17756,14 +17757,17 @@ OMPClause *SemaOpenMP::ActOnOpenMPSingleExprWithArgClause( case OMPC_dyn_groupprivate: { enum { Modifier1, Modifier2, NumberOfElements }; assert(Argument.size() == NumberOfElements && - ArgumentLoc.size() == NumberOfElements); + ArgumentLoc.size() == NumberOfElements && + "Modifiers for dyn_groupprivate clause and their locations are " + "expected."); Res = ActOnOpenMPDynGroupprivateClause( static_cast(Argument[Modifier1]), static_cast( Argument[Modifier2]), Expr, StartLoc, LParenLoc, ArgumentLoc[Modifier1], ArgumentLoc[Modifier2], EndLoc); - } break; + break; + } case OMPC_num_threads: assert(Argument.size() == 1 && ArgumentLoc.size() == 1 && "Modifier for num_threads clause and its location are expected."); diff --git a/clang/lib/Serialization/ASTReader.cpp b/clang/lib/Serialization/ASTReader.cpp index 46f10998ca76b..ac72cbb8ae459 100644 --- a/clang/lib/Serialization/ASTReader.cpp +++ b/clang/lib/Serialization/ASTReader.cpp @@ -12729,10 +12729,9 @@ void OMPClauseReader::VisitOMPDynGroupprivateClause( OMPDynGroupprivateClause *C) { VisitOMPClauseWithPreInit(C); C->setDynGroupprivateModifier( - static_cast(Record.readInt())); + Record.readEnum()); C->setDynGroupprivateFallbackModifier( - static_cast( - Record.readInt())); + Record.readEnum()); C->setSize(Record.readSubExpr()); C->setLParenLoc(Record.readSourceLocation()); C->setDynGroupprivateModifierLoc(Record.readSourceLocation()); From a3cd7ef4260b992ea5026a372c7484e2b28a34ba Mon Sep 17 00:00:00 2001 From: Kevin Sala Date: Tue, 4 Nov 2025 12:20:02 -0800 Subject: [PATCH 17/17] Add cgroup mem parameters in createTarget --- .../llvm/Frontend/OpenMP/OMPIRBuilder.h | 15 ++++++++++---- llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp | 20 +++++++++++-------- 2 files changed, 23 insertions(+), 12 deletions(-) diff --git a/llvm/include/llvm/Frontend/OpenMP/OMPIRBuilder.h b/llvm/include/llvm/Frontend/OpenMP/OMPIRBuilder.h index aa370606c6539..daa0c53157dec 100644 --- a/llvm/include/llvm/Frontend/OpenMP/OMPIRBuilder.h +++ b/llvm/include/llvm/Frontend/OpenMP/OMPIRBuilder.h @@ -2446,7 +2446,7 @@ class OpenMPIRBuilder { /// The number of threads. ArrayRef NumThreads; /// The size of the dynamic shared memory. - Value *DynCGGroupMem = nullptr; + Value *DynCGroupMem = nullptr; /// True if the kernel has 'no wait' clause. bool HasNoWait = false; /// The fallback mechanism for the shared memory. @@ -2457,12 +2457,12 @@ class OpenMPIRBuilder { TargetKernelArgs() {} TargetKernelArgs(unsigned NumTargetItems, TargetDataRTArgs RTArgs, Value *NumIterations, ArrayRef NumTeams, - ArrayRef NumThreads, Value *DynCGGroupMem, + ArrayRef NumThreads, Value *DynCGroupMem, bool HasNoWait, omp::OMPDynGroupprivateFallbackType DynCGroupMemFallback) : NumTargetItems(NumTargetItems), RTArgs(RTArgs), NumIterations(NumIterations), NumTeams(NumTeams), - NumThreads(NumThreads), DynCGGroupMem(DynCGGroupMem), + NumThreads(NumThreads), DynCGroupMem(DynCGroupMem), HasNoWait(HasNoWait), DynCGroupMemFallback(DynCGroupMemFallback) {} }; @@ -3248,6 +3248,10 @@ class OpenMPIRBuilder { /// dependency information as passed in the depend clause /// \param HasNowait Whether the target construct has a `nowait` clause or /// not. + /// \param DynCGroupMem The size of the dynamic groupprivate memory for each + /// cgroup. + /// \param DynCGroupMem The fallback mechanism to execute if the requested + /// cgroup memory cannot be provided. LLVM_ABI InsertPointOrErrorTy createTarget( const LocationDescription &Loc, bool IsOffloadEntry, OpenMPIRBuilder::InsertPointTy AllocaIP, @@ -3259,7 +3263,10 @@ class OpenMPIRBuilder { TargetBodyGenCallbackTy BodyGenCB, TargetGenArgAccessorsCallbackTy ArgAccessorFuncCB, CustomMapperCallbackTy CustomMapperCB, - const SmallVector &Dependencies, bool HasNowait = false); + const SmallVector &Dependencies, bool HasNowait = false, + Value *DynCGroupMem = nullptr, + omp::OMPDynGroupprivateFallbackType DynCGroupMemFallback = + omp::OMPDynGroupprivateFallbackType::Abort); /// Returns __kmpc_for_static_init_* runtime function for the specified /// size \a IVSize and sign \a IVSigned. Will create a distribute call diff --git a/llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp b/llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp index 9f80e3eb61bfc..5db9dbf8097e0 100644 --- a/llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp +++ b/llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp @@ -565,7 +565,7 @@ void OpenMPIRBuilder::getKernelArgsVector(TargetKernelArgs &KernelArgs, Flags, NumTeams3D, NumThreads3D, - KernelArgs.DynCGGroupMem}; + KernelArgs.DynCGroupMem}; } void OpenMPIRBuilder::addAttributes(omp::RuntimeFunction FnID, Function &Fn) { @@ -8229,7 +8229,8 @@ static void emitTargetCall( OpenMPIRBuilder::GenMapInfoCallbackTy GenMapInfoCB, OpenMPIRBuilder::CustomMapperCallbackTy CustomMapperCB, const SmallVector &Dependencies, - bool HasNoWait) { + bool HasNoWait, Value *DynCGroupMem, + OMPDynGroupprivateFallbackType DynCGroupMemFallback) { // Generate a function call to the host fallback implementation of the target // region. This is called by the host when no offload entry was generated for // the target region and when the offloading call fails at runtime. @@ -8365,12 +8366,13 @@ static void emitTargetCall( /*isSigned=*/false) : Builder.getInt64(0); - // TODO: Use correct DynCGGroupMem - Value *DynCGGroupMem = Builder.getInt32(0); + // Request zero groupprivate bytes by default. + if (!DynCGroupMem) + DynCGroupMem = Builder.getInt32(0); KArgs = OpenMPIRBuilder::TargetKernelArgs( - NumTargetItems, RTArgs, TripCount, NumTeamsC, NumThreadsC, - DynCGGroupMem, HasNoWait, OMPDynGroupprivateFallbackType::Abort); + NumTargetItems, RTArgs, TripCount, NumTeamsC, NumThreadsC, DynCGroupMem, + HasNoWait, DynCGroupMemFallback); // Assume no error was returned because TaskBodyCB and // EmitTargetCallFallbackCB don't produce any. @@ -8419,7 +8421,8 @@ OpenMPIRBuilder::InsertPointOrErrorTy OpenMPIRBuilder::createTarget( OpenMPIRBuilder::TargetBodyGenCallbackTy CBFunc, OpenMPIRBuilder::TargetGenArgAccessorsCallbackTy ArgAccessorFuncCB, CustomMapperCallbackTy CustomMapperCB, - const SmallVector &Dependencies, bool HasNowait) { + const SmallVector &Dependencies, bool HasNowait, + Value *DynCGroupMem, OMPDynGroupprivateFallbackType DynCGroupMemFallback) { if (!updateToLocation(Loc)) return InsertPointTy(); @@ -8442,7 +8445,8 @@ OpenMPIRBuilder::InsertPointOrErrorTy OpenMPIRBuilder::createTarget( if (!Config.isTargetDevice()) emitTargetCall(*this, Builder, AllocaIP, Info, DefaultAttrs, RuntimeAttrs, IfCond, OutlinedFn, OutlinedFnID, Inputs, GenMapInfoCB, - CustomMapperCB, Dependencies, HasNowait); + CustomMapperCB, Dependencies, HasNowait, DynCGroupMem, + DynCGroupMemFallback); return Builder.saveIP(); }