[flang]Add Semantic Checks for OpenMP Allocate Clause

[kiranchandramohan]

You missed formatting with clang-format.
Please add additional tests to ensure allocate clause is accepted in all directives where it is allowed.
More comments inline.

flang/lib/Semantics/resolve-directives.cpp

887

There are two minor issues with putting this check here,

1. I think the location of the error should be the allocate clause and not the location of the occurrence of the variable inside the region.
2. If the variable does not occur inside the parallel region then I think it is still an error as per the standard. I believe the current implementation will miss this case.

889

This should be changed to something like "must appear in a private ...".

[kiranchandramohan]

flang/lib/Semantics/resolve-directives.cpp

94

Check whether this kind of clause is present in OpenACC also. If not then move it to the OmpDirectiveVisitor class.

105

Is this field required? Can we use an additional check (whether symbol is private) on the dataSharingAttributeObjects_ to obtain this set?

[Rin]

Update Allocate Clause

[Rin]

flang/lib/Semantics/resolve-directives.cpp

105

I should clear this set after each directive. I'll add this in the next update

889

Here it was meant to be "private data-sharing attribute clause". I'll change that in the next update. I'll fix formatting issue in that update also

[Rin]

Updates to Semantic Check

[Rin]

Remove unnecessary set

[kiranchandramohan]

Thanks Rin for making the changes. A few more comments.

flang/lib/Semantics/resolve-directives.cpp

232–246

Do all OpenMPBlockConstructs have allocate clause?
If not should we restrict this check to only those OpenMPBlockConstructs which have allocate clause?

307

Is a line required above?

336–338

Can this check be done before the find on dataSharingAttributeObjects?

flang/test/Semantics/omp-resolve06.f90

18

We need more error tests.
Use a mix of firstprivate, lastprivate + allocate and private with same and different variables.

[Rin]

Additional tests and fixed semantic checks.

[Rin]

Minor Changes

[Rin]

flang/lib/Semantics/resolve-directives.cpp

740

This will check for a data sharing attribute object, not a private data sharing attribute object. I will fix this.

[Rin]

Semantic Check fix

[kiranchandramohan]

flang/lib/Semantics/resolve-directives.cpp

680–781

Is the code in each of the above cases the same? Can we avoid the duplication by the following or making the common code a function?

case llvm::omp::Directive::OMPD_parallel:
case llvm::omp::Directive::OMPD_target:
case llvm::omp::Directive::OMPD_task:
case llvm::omp::Directive::OMPD_teams:
New Code
break;

[Rin]

Additional test and minor updates

[Rin]

Remove extra line

[clementval]

llvm/include/llvm/Frontend/OpenMP/OMP.td

211

This information is not used in Flang at the moment. Looks like couple of Clang OpenMP tests are failing with this patch and Clang is using this information. So I'm not sure it is worth adding this now or you would have to check the failure in Clang.

[Rin]

llvm/include/llvm/Frontend/OpenMP/OMP.td

211

I didn't notice that. I'll make sure to remove this.

[Rin]

Remove OpenMP 5.0 Version Check for Allocate Clause

[kiranchandramohan]

flang/lib/Semantics/resolve-directives.cpp

691

Can you use a range based for loop or std::for_each for this loop and the one below?

695

Can you use std::find_if or range based for.

698

Can break here.

[Rin]

Minor Updates

[Rin]

Change set of allocate symbols to set of allocate names

[kiranchandramohan]

Looking good. Few minor comments. Please wait for one more approval.

flang/lib/Semantics/resolve-directives.cpp

306

Thinking about this again, this set will probably not work since different occurrences of names have different pointers. And we probably want to flag error at each occurrence of the same variable. So we can very well have a vector/smallvector. Add a test

  integer ::  N = 2
  integer :: x

  !ERROR: The ALLOCATE clause requires that 'x' must be listed in a private data-sharing attribute clause on the same directive
  !$omp parallel allocate(omp_default_mem_space : x) allocate(omp_default_mem_space : x)
  do i = 1, N
     x = 2
  enddo
  !$omp end parallel

end

689

Check the clang-tidy warning here.

693

I think the ToString() is not required here.
if (allocName->source == symbolPrivate.name())

[Rin]

Change set of names to vector of names

[Rin]

Remove unnecessary print statement

[Rin]

Could someone outside of ARM review this patch as well please?

[clementval]

LGTM (just a small comment about an include)

flang/lib/Semantics/resolve-directives.cpp

21

Is this still needed?

[Rin]

flang/lib/Semantics/resolve-directives.cpp

21

No, it's not. I'll take it out before submitting the patch upstream.

[kiranktp]

LGTM

[Rin]

Remove unnecessary include

[raghavendhra]

LGTM

flang/include/flang/Semantics/symbol.h

ENUM_CLASS(Flag,
       // OpenACC miscellaneous flags
       AccCommonBlock, AccThreadPrivate, AccReduction, AccNone, AccPreDetermined,
 
       // OpenMP data-sharing attribute
       OmpShared, OmpPrivate, OmpLinear, OmpFirstPrivate, OmpLastPrivate,
       // OpenMP data-mapping attribute
       OmpMapTo, OmpMapFrom, OmpMapAlloc, OmpMapRelease, OmpMapDelete,
       // OpenMP miscellaneous flags
-      OmpCommonBlock, OmpReduction, OmpDeclareSimd, OmpDeclareTarget,
-      OmpThreadprivate, OmpDeclareReduction, OmpFlushed, OmpCriticalLock,
-      OmpIfSpecified, OmpNone, OmpPreDetermined);
+      OmpCommonBlock, OmpReduction, OmpAllocate, OmpDeclareSimd,
+      OmpDeclareTarget, OmpThreadprivate, OmpDeclareReduction, OmpFlushed,
+      OmpCriticalLock, OmpIfSpecified, OmpNone, OmpPreDetermined);
   using Flags = common::EnumSet<Flag, Flag_enumSize>;
 
   const Scope &owner() const { return *owner_; }
   const SourceName &name() const { return name_; }
   Attrs &attrs() { return attrs_; }
   const Attrs &attrs() const { return attrs_; }
   Flags &flags() { return flags_; }
   const Flags &flags() const { return flags_; }

flang/lib/Semantics/check-omp-structure.h

public:
   void Enter(const parser::OmpClause::ThreadLimit &);
   void Enter(const parser::OmpClause::To &);
   void Enter(const parser::OmpClause::Link &);
   void Enter(const parser::OmpClause::Uniform &);
   void Enter(const parser::OmpClause::UseDevicePtr &);
   void Enter(const parser::OmpClause::IsDevicePtr &);
 
   void Enter(const parser::OmpAlignedClause &);
+  void Enter(const parser::OmpAllocateClause &);
   void Enter(const parser::OmpDefaultClause &);
   void Enter(const parser::OmpDefaultmapClause &);
   void Enter(const parser::OmpDependClause &);
   void Enter(const parser::OmpIfClause &);
   void Enter(const parser::OmpLinearClause &);
   void Enter(const parser::OmpMapClause &);
   void Enter(const parser::OmpProcBindClause &);
   void Enter(const parser::OmpReductionClause &);

flang/lib/Semantics/check-omp-structure.cpp

if (const auto v{GetIntValue(*expr)}) {
         context_.Say(GetContext().clauseSource,
             "The ALIGNMENT parameter of the ALIGNED clause must be "
             "a constant positive integer expression"_err_en_US);
       }
     }
   }
   // 2.8.1 TODO: list-item attribute check
 }
+void OmpStructureChecker::Enter(const parser::OmpAllocateClause &) {
+  CheckAllowed(llvm::omp::Clause::OMPC_allocate);
+}
 void OmpStructureChecker::Enter(const parser::OmpDefaultClause &) {
   CheckAllowed(llvm::omp::Clause::OMPC_default);
 }
 void OmpStructureChecker::Enter(const parser::OmpDefaultmapClause &x) {
   CheckAllowed(llvm::omp::Clause::OMPC_defaultmap);
   using VariableCategory = parser::OmpDefaultmapClause::VariableCategory;
   if (!std::get<std::optional<VariableCategory>>(x.t)) {
     context_.Say(GetContext().clauseSource,

flang/lib/Semantics/resolve-directives.cpp

 //===----------------------------------------------------------------------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 
 #include "resolve-directives.h"
 
 #include "check-acc-structure.h"
 #include "check-omp-structure.h"
 #include "resolve-names-utils.h"
 #include "flang/Common/idioms.h"
 #include "flang/Evaluate/fold.h"
+#include "flang/Evaluate/type.h"
 #include "flang/Parser/parse-tree-visitor.h"
 #include "flang/Parser/parse-tree.h"
 #include "flang/Parser/tools.h"
 #include "flang/Semantics/expression.h"
 #include <list>
 #include <map>
 
 namespace Fortran::semantics {
 
 template <typename T> class DirectiveAttributeVisitor {
 public:
   explicit DirectiveAttributeVisitor(SemanticsContext &context)
       : context_{context} {}
static const parser::Name *GetDesignatorNameIfDataRef(
     return dataRef ? std::get_if<parser::Name>(&dataRef->u) : nullptr;
   }
   void AddDataSharingAttributeObject(SymbolRef object) {
     dataSharingAttributeObjects_.insert(object);
   }
   void ClearDataSharingAttributeObjects() {
     dataSharingAttributeObjects_.clear();
   }
   bool HasDataSharingAttributeObject(const Symbol &);
+  bool HasPrivateDataSharingAttributeObject(const Symbol &);
   const parser::Name &GetLoopIndex(const parser::DoConstruct &);
   const parser::DoConstruct *GetDoConstructIf(
       const parser::ExecutionPartConstruct &);
   Symbol *DeclarePrivateAccessEntity(
       const parser::Name &, Symbol::Flag, Scope &);
   Symbol *DeclarePrivateAccessEntity(Symbol &, Symbol::Flag, Scope &);
   Symbol *DeclareOrMarkOtherAccessEntity(const parser::Name &, Symbol::Flag);
 
   SymbolSet dataSharingAttributeObjects_; // on one directive
+  SymbolSet privateDataSharingAttributeObjects_;
   SemanticsContext &context_;
   std::vector<DirContext> dirContext_; // used as a stack
 };
 
 class AccAttributeVisitor : DirectiveAttributeVisitor<llvm::acc::Directive> {
 public:
   explicit AccAttributeVisitor(SemanticsContext &context)
       : DirectiveAttributeVisitor(context) {}
public:
   template <typename A> void Post(const A &) {}
 
   bool Pre(const parser::SpecificationPart &x) {
     Walk(std::get<std::list<parser::OpenMPDeclarativeConstruct>>(x.t));
     return false;
   }
 
   bool Pre(const parser::OpenMPBlockConstruct &);
-  void Post(const parser::OpenMPBlockConstruct &) { PopContext(); }
+  void Post(const parser::OpenMPBlockConstruct &) {
+    std::set<SymbolRef>::iterator it = allocateObjects_.begin();
+    while (it != allocateObjects_.end()) {
+      const Symbol &symbol{*it};
+      if (!HasPrivateDataSharingAttributeObject(symbol)) {
+        context_.Say(symbol.name(),
+            "The ALLOCATE clause requires that '%s' must be listed in a "
+            "private "
+            "data-sharing attribute clause on the same directive"_err_en_US,
+            symbol.name().ToString());
+      }
+      it++;
+    }
+    PopContext();
+  }
   void Post(const parser::OmpBeginBlockDirective &) {
     GetContext().withinConstruct = true;
   }
 
   bool Pre(const parser::OpenMPLoopConstruct &);
   void Post(const parser::OpenMPLoopConstruct &) { PopContext(); }
   void Post(const parser::OmpBeginLoopDirective &) {
     GetContext().withinConstruct = true;
public:
   bool Pre(const parser::OmpClause::Shared &x) {
     ResolveOmpObjectList(x.v, Symbol::Flag::OmpShared);
     return false;
   }
   bool Pre(const parser::OmpClause::Private &x) {
     ResolveOmpObjectList(x.v, Symbol::Flag::OmpPrivate);
     return false;
   }
+  bool Pre(const parser::OmpAllocateClause &x) {
+    const auto &objectList{std::get<parser::OmpObjectList>(x.t)};
+    ResolveOmpObjectList(objectList, Symbol::Flag::OmpAllocate);
+    return false;
+  }
   bool Pre(const parser::OmpClause::Firstprivate &x) {
     ResolveOmpObjectList(x.v, Symbol::Flag::OmpFirstPrivate);
     return false;
   }
   bool Pre(const parser::OmpClause::Lastprivate &x) {
     ResolveOmpObjectList(x.v, Symbol::Flag::OmpLastPrivate);
     return false;
   }
private:
   static constexpr Symbol::Flags ompFlagsRequireNewSymbol{
       Symbol::Flag::OmpPrivate, Symbol::Flag::OmpLinear,
       Symbol::Flag::OmpFirstPrivate, Symbol::Flag::OmpLastPrivate,
       Symbol::Flag::OmpReduction};
 
   static constexpr Symbol::Flags ompFlagsRequireMark{
       Symbol::Flag::OmpThreadprivate};
 
+  SymbolSet allocateObjects_;
+  void AddAllocateObject(SymbolRef object) { allocateObjects_.insert(object); }
+
   // Predetermined DSA rules
   void PrivatizeAssociatedLoopIndex(const parser::OpenMPLoopConstruct &);
   void ResolveSeqLoopIndexInParallelOrTaskConstruct(const parser::Name &);
 
   void ResolveOmpObjectList(const parser::OmpObjectList &, Symbol::Flag);
   void ResolveOmpObject(const parser::OmpObject &, Symbol::Flag);
   Symbol *ResolveOmp(const parser::Name &, Symbol::Flag, Scope &);
   Symbol *ResolveOmp(Symbol &, Symbol::Flag, Scope &);
   Symbol *ResolveOmpCommonBlockName(const parser::Name *);
   Symbol *DeclareOrMarkOtherAccessEntity(const parser::Name &, Symbol::Flag);
   Symbol *DeclareOrMarkOtherAccessEntity(Symbol &, Symbol::Flag);
   void CheckMultipleAppearances(
       const parser::Name &, const Symbol &, Symbol::Flag);
 };
 
 template <typename T>
 bool DirectiveAttributeVisitor<T>::HasDataSharingAttributeObject(
     const Symbol &object) {
   auto it{dataSharingAttributeObjects_.find(object)};
   return it != dataSharingAttributeObjects_.end();
 }
 
 template <typename T>
+bool DirectiveAttributeVisitor<T>::HasPrivateDataSharingAttributeObject(
+    const Symbol &object) {
+  auto it{dataSharingAttributeObjects_.find(object)};
+  if (it != dataSharingAttributeObjects_.end() &&
+      (object.test(Symbol::Flag::OmpPrivate) ||
+          object.test(Symbol::Flag::OmpFirstPrivate) ||
+          object.test(Symbol::Flag::OmpLastPrivate))) {
+    return true;
+  }
+  return false;
+}
+
+template <typename T>
 const parser::Name &DirectiveAttributeVisitor<T>::GetLoopIndex(
     const parser::DoConstruct &x) {
   using Bounds = parser::LoopControl::Bounds;
   return std::get<Bounds>(x.GetLoopControl()->u).name.thing;
 }
 
 template <typename T>
 const parser::DoConstruct *DirectiveAttributeVisitor<T>::GetDoConstructIf(
bool OmpAttributeVisitor::Pre(const parser::OpenMPBlockConstruct &x) {
   ClearDataSharingAttributeObjects();
   return true;
 }
 
 bool OmpAttributeVisitor::Pre(const parser::OpenMPLoopConstruct &x) {
   const auto &beginLoopDir{std::get<parser::OmpBeginLoopDirective>(x.t)};
   const auto &beginDir{std::get<parser::OmpLoopDirective>(beginLoopDir.t)};
   const auto &clauseList{std::get<parser::OmpClauseList>(beginLoopDir.t)};
   switch (beginDir.v) {
   case llvm::omp::Directive::OMPD_distribute:
   case llvm::omp::Directive::OMPD_distribute_parallel_do:
   case llvm::omp::Directive::OMPD_distribute_parallel_do_simd:
   case llvm::omp::Directive::OMPD_distribute_simd:
   case llvm::omp::Directive::OMPD_do:
   case llvm::omp::Directive::OMPD_do_simd:
   case llvm::omp::Directive::OMPD_parallel_do:
   case llvm::omp::Directive::OMPD_parallel_do_simd:
   case llvm::omp::Directive::OMPD_simd:
   case llvm::omp::Directive::OMPD_target_parallel_do:
   case llvm::omp::Directive::OMPD_target_parallel_do_simd:
   case llvm::omp::Directive::OMPD_target_teams_distribute:
   case llvm::omp::Directive::OMPD_target_teams_distribute_parallel_do:
   case llvm::omp::Directive::OMPD_target_teams_distribute_parallel_do_simd:
   case llvm::omp::Directive::OMPD_target_teams_distribute_simd:
   case llvm::omp::Directive::OMPD_target_simd:
   case llvm::omp::Directive::OMPD_taskloop:
   case llvm::omp::Directive::OMPD_taskloop_simd:
   case llvm::omp::Directive::OMPD_teams_distribute:
   case llvm::omp::Directive::OMPD_teams_distribute_parallel_do:
   case llvm::omp::Directive::OMPD_teams_distribute_parallel_do_simd:
   case llvm::omp::Directive::OMPD_teams_distribute_simd:
     PushContext(beginDir.source, beginDir.v);
     break;
   default:
     break;
   }
   ClearDataSharingAttributeObjects();
   SetContextAssociatedLoopLevel(GetAssociatedLoopLevelFromClauses(clauseList));
   PrivatizeAssociatedLoopIndex(x);
   return true;
 }
 
 void OmpAttributeVisitor::ResolveSeqLoopIndexInParallelOrTaskConstruct(
     const parser::Name &iv) {
   auto targetIt{dirContext_.rbegin()};
   for (;; ++targetIt) {
     if (targetIt == dirContext_.rend()) {
       return;
     }
     if (llvm::omp::parallelSet.test(targetIt->directive) ||
         llvm::omp::taskGeneratingSet.test(targetIt->directive)) {
       break;
     }
   }
   if (auto *symbol{ResolveOmp(iv, Symbol::Flag::OmpPrivate, targetIt->scope)}) {
     targetIt++;
     symbol->set(Symbol::Flag::OmpPreDetermined);
     iv.symbol = symbol; // adjust the symbol within region
     for (auto it{dirContext_.rbegin()}; it != targetIt; ++it) {
       AddToContextObjectWithDSA(*symbol, Symbol::Flag::OmpPrivate, *it);
     }
   }
 }
 
 // 2.15.1.1 Data-sharing Attribute Rules - Predetermined
 //   - A loop iteration variable for a sequential loop in a parallel
 //     or task generating construct is private in the innermost such
 //     construct that encloses the loop
 bool OmpAttributeVisitor::Pre(const parser::DoConstruct &x) {
   if (!dirContext_.empty() && GetContext().withinConstruct) {
     if (const auto &iv{GetLoopIndex(x)}; iv.symbol) {
       if (!iv.symbol->test(Symbol::Flag::OmpPreDetermined)) {
         ResolveSeqLoopIndexInParallelOrTaskConstruct(iv);
       } else {
         // TODO: conflict checks with explicitly determined DSA
       }
     }
   }
   return true;
 }
 
 std::int64_t OmpAttributeVisitor::GetAssociatedLoopLevelFromClauses(
     const parser::OmpClauseList &x) {
   std::int64_t orderedLevel{0};
   std::int64_t collapseLevel{0};
   for (const auto &clause : x.v) {
     if (const auto *orderedClause{
             std::get_if<parser::OmpClause::Ordered>(&clause.u)}) {
       if (const auto v{EvaluateInt64(context_, orderedClause->v)}) {
         orderedLevel = *v;
       }
     }
     if (const auto *collapseClause{
             std::get_if<parser::OmpClause::Collapse>(&clause.u)}) {
       if (const auto v{EvaluateInt64(context_, collapseClause->v)}) {
         collapseLevel = *v;
       }
     }
   }
 
   if (orderedLevel && (!collapseLevel || orderedLevel >= collapseLevel)) {
     return orderedLevel;
   } else if (!orderedLevel && collapseLevel) {
     return collapseLevel;
   } // orderedLevel < collapseLevel is an error handled in structural checks
   return 1; // default is outermost loop
 }
 
 // 2.15.1.1 Data-sharing Attribute Rules - Predetermined
 //   - The loop iteration variable(s) in the associated do-loop(s) of a do,
 //     parallel do, taskloop, or distribute construct is (are) private.
 //   - The loop iteration variable in the associated do-loop of a simd construct
 //     with just one associated do-loop is linear with a linear-step that is the
 //     increment of the associated do-loop.
 //   - The loop iteration variables in the associated do-loops of a simd
 //     construct with multiple associated do-loops are lastprivate.
 //
 // TODO: revisit after semantics checks are completed for do-loop association of
if (!symbol->owner().IsDerivedType() && !symbol->has<ProcEntityDetails>() &&
         } else if (GetContext().defaultDSA == Symbol::Flag::OmpNone) {
           context_.Say(name.source,
               "The DEFAULT(NONE) clause requires that '%s' must be listed in "
               "a data-sharing attribute clause"_err_en_US,
               symbol->name());
         }
       }
     }
   } // within OpenMP construct
 }
 
 Symbol *OmpAttributeVisitor::ResolveOmpCommonBlockName(
     const parser::Name *name) {
   if (auto *prev{name
               ? GetContext().scope.parent().FindCommonBlock(name->source)
               : nullptr}) {
     name->symbol = prev;
     return prev;
   } else {
std::visit(
       common::visitors{
           [&](const parser::Designator &designator) {
             if (const auto *name{GetDesignatorNameIfDataRef(designator)}) {
               if (auto *symbol{ResolveOmp(*name, ompFlag, currScope())}) {
                 AddToContextObjectWithDSA(*symbol, ompFlag);
                 if (dataSharingAttributeFlags.test(ompFlag)) {
                   CheckMultipleAppearances(*name, *symbol, ompFlag);
                 }
+                if (ompFlag == Symbol::Flag::OmpAllocate) {
+                  AddAllocateObject(*symbol);
+                }
               }
             } else {
               // Array sections to be changed to substrings as needed
               if (AnalyzeExpr(context_, designator)) {
                 if (std::holds_alternative<parser::Substring>(designator.u)) {
                   context_.Say(designator.source,
                       "Substrings are not allowed on OpenMP "
                       "directives or clauses"_err_en_US);

flang/test/Semantics/omp-clause-validity01.f90

 ! RUN: %S/test_errors.sh %s %t %f18 -fopenmp
 use omp_lib
 ! Check OpenMP clause validity for the following directives:
 !
 !    2.5 PARALLEL construct
 !    2.7.1 Loop construct
 !    ...
 
 ! TODO: all the internal errors
 
   integer :: b = 128
-  integer :: c = 32
+  integer :: z, c = 32
   integer, parameter :: num = 16
   real(8) :: arrayA(256), arrayB(512)
 
   integer(omp_memspace_handle_kind) :: xy_memspace = omp_default_mem_space
   type(omp_alloctrait) :: xy_traits(1) = [omp_alloctrait(omp_atk_alignment,64)]
   integer(omp_allocator_handle_kind) :: xy_alloc
   xy_alloc = omp_init_allocator(xy_memspace, 1, xy_traits)
 

 !                        allocate-clause
 
   !$omp parallel
   do i = 1, N
      a = 3.14
   enddo
   !$omp end parallel
 
-  !$omp parallel allocate(b)
+  !$omp parallel private(b) allocate(b)
   do i = 1, N
      a = 3.14
   enddo
   !$omp end parallel
 
-  !$omp parallel allocate(omp_default_mem_space : b, c)
+  !$omp parallel private(c, b) allocate(omp_default_mem_space : b, c)
   do i = 1, N
      a = 3.14
   enddo
   !$omp end parallel
 
-  !$omp parallel allocate(b) allocate(c)
+  !$omp parallel allocate(b) allocate(c) private(b, c)
   do i = 1, N
      a = 3.14
   enddo
   !$omp end parallel
 
-  !$omp parallel allocate(xy_alloc :b) 
+  !$omp parallel allocate(xy_alloc :b) private(b)
   do i = 1, N
      a = 3.14
   enddo
   !$omp end parallel
   
+  !$omp task private(b) allocate(b)
+  do i = 1, N
+     z = 2
+  end do
+  !$omp end task
+
+  !$omp teams private(b) allocate(b)
+  do i = 1, N
+     z = 2
+  end do
+  !$omp end teams
+
+  !$omp target private(b) allocate(b)
+  do i = 1, N
+     z = 2
+  end do
+  !$omp end target
+ 
+  !ERROR: ALLOCATE clause is not allowed on the TARGET DATA directive
+  !$omp target data map(from: b) allocate(b)
+  do i = 1, N
+     z = 2
+  enddo
+   !$omp end target data
+
   !ERROR: SCHEDULE clause is not allowed on the PARALLEL directive
   !$omp parallel schedule(static)
   do i = 1, N
      a = 3.14
   enddo
   !$omp end parallel
 
   !ERROR: COLLAPSE clause is not allowed on the PARALLEL directive

flang/test/Semantics/omp-resolve06.f90

+! RUN: %S/test_errors.sh %s %t %f18 -fopenmp
+use omp_lib
+!2.11.4 Allocate Clause
+!For any list item that is specified in the allocate
+!clause on a directive, a data-sharing attribute clause
+!that may create a private copy of that list item must be
+!specified on the same directive.
+
+  integer ::  N = 2
+
+  !ERROR: The ALLOCATE clause requires that 'x' must be listed in a private data-sharing attribute clause on the same directive
+  !$omp parallel allocate(omp_default_mem_space : x)
+  do i = 1, N
+     x = 2
+  enddo
+  !$omp end parallel
+end

llvm/include/llvm/Frontend/OpenMP/OMP.td

 def OMP_Parallel : Directive<"parallel"> {
   let allowedClauses = [
     VersionedClause<OMPC_Default>,
     VersionedClause<OMPC_Private>,
     VersionedClause<OMPC_FirstPrivate>,
     VersionedClause<OMPC_Shared>,
     VersionedClause<OMPC_Reduction>,
     VersionedClause<OMPC_Copyin>,
-    VersionedClause<OMPC_Allocate>
+    VersionedClause<OMPC_Allocate, 50>
   ];
   let allowedOnceClauses = [
     VersionedClause<OMPC_If>,
     VersionedClause<OMPC_NumThreads>,
     VersionedClause<OMPC_ProcBind>,
   ];
 }
 def OMP_Task : Directive<"task"> {
   let allowedClauses = [
     VersionedClause<OMPC_Default>,
     VersionedClause<OMPC_Private>,
     VersionedClause<OMPC_FirstPrivate>,
     VersionedClause<OMPC_Shared>,
     VersionedClause<OMPC_Untied>,
     VersionedClause<OMPC_Mergeable>,
     VersionedClause<OMPC_Depend>,
     VersionedClause<OMPC_InReduction>,
-    VersionedClause<OMPC_Allocate>,
+    VersionedClause<OMPC_Allocate, 50>,
     VersionedClause<OMPC_Detach, 50>,
     VersionedClause<OMPC_Affinity, 50>
   ];
   let allowedOnceClauses = [
     VersionedClause<OMPC_If>,
     VersionedClause<OMPC_Final>,
     VersionedClause<OMPC_Priority>
   ];
 }
 def OMP_Simd : Directive<"simd"> {
   let allowedClauses = [
     VersionedClause<OMPC_Private>,
     VersionedClause<OMPC_LastPrivate>,
     VersionedClause<OMPC_Linear>,
     VersionedClause<OMPC_Aligned>,
     VersionedClause<OMPC_Reduction>,
-    VersionedClause<OMPC_Allocate>,
+    VersionedClause<OMPC_Allocate, 50>,
     VersionedClause<OMPC_NonTemporal, 50>,
     VersionedClause<OMPC_Order, 50>
   ];
   let allowedOnceClauses = [
     VersionedClause<OMPC_Collapse>,
     VersionedClause<OMPC_SafeLen>,
     VersionedClause<OMPC_SimdLen>,
     VersionedClause<OMPC_If, 50>,
   ];
 }
 def OMP_For : Directive<"for"> {
   let allowedClauses = [
     VersionedClause<OMPC_Private>,
     VersionedClause<OMPC_LastPrivate>,
     VersionedClause<OMPC_FirstPrivate>,
     VersionedClause<OMPC_Reduction>,
     VersionedClause<OMPC_Collapse>,
     VersionedClause<OMPC_Schedule>,
     VersionedClause<OMPC_Ordered>,
     VersionedClause<OMPC_NoWait>,
     VersionedClause<OMPC_Linear>,
-    VersionedClause<OMPC_Allocate>,
+    VersionedClause<OMPC_Allocate, 50>,
     VersionedClause<OMPC_Order, 50>
   ];
 }
 def OMP_Do : Directive<"do"> {
   let allowedClauses = [
     VersionedClause<OMPC_Private>,
     VersionedClause<OMPC_FirstPrivate>,
     VersionedClause<OMPC_LastPrivate>,

 }
 def OMP_Sections : Directive<"sections"> {
   let allowedClauses = [
     VersionedClause<OMPC_Private>,
     VersionedClause<OMPC_LastPrivate>,
     VersionedClause<OMPC_FirstPrivate>,
     VersionedClause<OMPC_Reduction>,
     VersionedClause<OMPC_NoWait>,
-    VersionedClause<OMPC_Allocate>
+    VersionedClause<OMPC_Allocate, 50>
   ];
 }
 def OMP_Section : Directive<"section"> {}
 def OMP_Single : Directive<"single"> {
   let allowedClauses = [
     VersionedClause<OMPC_Private>,
     VersionedClause<OMPC_FirstPrivate>,
     VersionedClause<OMPC_CopyPrivate>,
     VersionedClause<OMPC_NoWait>,
-    VersionedClause<OMPC_Allocate>
+    VersionedClause<OMPC_Allocate, 50>
   ];
 }
 def OMP_Master : Directive<"master"> {}
 def OMP_Critical : Directive<"critical"> {
   let allowedClauses = [
     VersionedClause<OMPC_Hint>
   ];
 }
def OMP_Target : Directive<"target"> {
   let allowedClauses = [
     VersionedClause<OMPC_If>,
     VersionedClause<OMPC_Map>,
     VersionedClause<OMPC_Private>,
     VersionedClause<OMPC_Depend>,
     VersionedClause<OMPC_FirstPrivate>,
     VersionedClause<OMPC_IsDevicePtr>,
     VersionedClause<OMPC_Reduction>,
-    VersionedClause<OMPC_Allocate>,
+    VersionedClause<OMPC_Allocate, 50>,
     VersionedClause<OMPC_UsesAllocators, 50>
   ];
   let allowedOnceClauses = [
     VersionedClause<OMPC_Device>,
     VersionedClause<OMPC_DefaultMap>,
     VersionedClause<OMPC_NoWait>
   ];
 }
 def OMP_Teams : Directive<"teams"> {
   let allowedClauses = [
     VersionedClause<OMPC_Private>,
     VersionedClause<OMPC_FirstPrivate>,
     VersionedClause<OMPC_Shared>,
     VersionedClause<OMPC_Reduction>,
-    VersionedClause<OMPC_Allocate>
+    VersionedClause<OMPC_Allocate, 50>
   ];
   let allowedOnceClauses = [
     VersionedClause<OMPC_Default>,
     VersionedClause<OMPC_NumTeams>,
     VersionedClause<OMPC_ThreadLimit>
   ];
 }
 def OMP_Cancel : Directive<"cancel"> {
let allowedClauses = [
     VersionedClause<OMPC_NoWait>,
     VersionedClause<OMPC_Depend>,
     VersionedClause<OMPC_Private>,
     VersionedClause<OMPC_FirstPrivate>,
     VersionedClause<OMPC_Default>,
     VersionedClause<OMPC_Shared>,
     VersionedClause<OMPC_Reduction>,
     VersionedClause<OMPC_IsDevicePtr>,
-    VersionedClause<OMPC_Allocate>,
+    VersionedClause<OMPC_Allocate, 50>,
     VersionedClause<OMPC_UsesAllocators, 50>
   ];
   let allowedOnceClauses = [
     VersionedClause<OMPC_DefaultMap>,
     VersionedClause<OMPC_Device>,
     VersionedClause<OMPC_If>,
     VersionedClause<OMPC_NumThreads>,
     VersionedClause<OMPC_ProcBind>
let allowedClauses = [
     VersionedClause<OMPC_ProcBind>,
     VersionedClause<OMPC_Shared>,
     VersionedClause<OMPC_Reduction>,
     VersionedClause<OMPC_Collapse>,
     VersionedClause<OMPC_Schedule>,
     VersionedClause<OMPC_Ordered>,
     VersionedClause<OMPC_Linear>,
     VersionedClause<OMPC_IsDevicePtr>,
-    VersionedClause<OMPC_Allocate>,
+    VersionedClause<OMPC_Allocate, 50>,
     VersionedClause<OMPC_Order, 50>,
     VersionedClause<OMPC_UsesAllocators, 50>
   ];
 }
 def OMP_TargetParallelDo : Directive<"target parallel do"> {
   let allowedClauses = [
     VersionedClause<OMPC_Map>,
     VersionedClause<OMPC_Private>,
let allowedClauses = [
     VersionedClause<OMPC_Shared>,
     VersionedClause<OMPC_Reduction>,
     VersionedClause<OMPC_Copyin>,
     VersionedClause<OMPC_LastPrivate>,
     VersionedClause<OMPC_Collapse>,
     VersionedClause<OMPC_Schedule>,
     VersionedClause<OMPC_Ordered>,
     VersionedClause<OMPC_Linear>,
-    VersionedClause<OMPC_Allocate>,
+    VersionedClause<OMPC_Allocate, 50>,
     VersionedClause<OMPC_Order, 50>
   ];
 }
 def OMP_ParallelDo : Directive<"parallel do"> {
   let allowedClauses = [
     VersionedClause<OMPC_Default>,
     VersionedClause<OMPC_Private>,
     VersionedClause<OMPC_FirstPrivate>,
let allowedClauses = [
     VersionedClause<OMPC_LastPrivate>,
     VersionedClause<OMPC_Collapse>,
     VersionedClause<OMPC_Schedule>,
     VersionedClause<OMPC_SafeLen>,
     VersionedClause<OMPC_SimdLen>,
     VersionedClause<OMPC_Linear>,
     VersionedClause<OMPC_Aligned>,
     VersionedClause<OMPC_Ordered>,
-    VersionedClause<OMPC_Allocate>,
+    VersionedClause<OMPC_Allocate, 50>,
     VersionedClause<OMPC_NonTemporal, 50>,
     VersionedClause<OMPC_Order, 50>
   ];
 }
 def OMP_ParallelDoSimd : Directive<"parallel do simd"> {
   let allowedClauses = [
     VersionedClause<OMPC_Default>,
     VersionedClause<OMPC_Private>,
let allowedClauses = [
     VersionedClause<OMPC_NumThreads>,
     VersionedClause<OMPC_Default>,
     VersionedClause<OMPC_Private>,
     VersionedClause<OMPC_FirstPrivate>,
     VersionedClause<OMPC_Shared>,
     VersionedClause<OMPC_Copyin>,
     VersionedClause<OMPC_Reduction>,
     VersionedClause<OMPC_ProcBind>,
-    VersionedClause<OMPC_Allocate>
+    VersionedClause<OMPC_Allocate, 50>
   ];
 }
 def OMP_ParallelSections : Directive<"parallel sections"> {
   let allowedClauses = [
     VersionedClause<OMPC_If>,
     VersionedClause<OMPC_Default>,
     VersionedClause<OMPC_ProcBind>,
     VersionedClause<OMPC_Private>,
     VersionedClause<OMPC_FirstPrivate>,
     VersionedClause<OMPC_Shared>,
     VersionedClause<OMPC_Reduction>,
     VersionedClause<OMPC_Copyin>,
     VersionedClause<OMPC_LastPrivate>,
-    VersionedClause<OMPC_Allocate>
+    VersionedClause<OMPC_Allocate, 50>
   ];
   let allowedOnceClauses = [
     VersionedClause<OMPC_NumThreads>
   ];
 }
 def OMP_ForSimd : Directive<"for simd"> {
   let allowedClauses = [
     VersionedClause<OMPC_Private>,
     VersionedClause<OMPC_FirstPrivate>,
     VersionedClause<OMPC_LastPrivate>,
     VersionedClause<OMPC_Reduction>,
     VersionedClause<OMPC_Schedule>,
     VersionedClause<OMPC_Collapse>,
     VersionedClause<OMPC_NoWait>,
     VersionedClause<OMPC_SafeLen>,
     VersionedClause<OMPC_SimdLen>,
     VersionedClause<OMPC_Linear>,
     VersionedClause<OMPC_Aligned>,
     VersionedClause<OMPC_Ordered>,
-    VersionedClause<OMPC_Allocate>,
+    VersionedClause<OMPC_Allocate, 50>,
     VersionedClause<OMPC_If, 50>,
     VersionedClause<OMPC_NonTemporal, 50>,
     VersionedClause<OMPC_Order, 50>,
   ];
 }
 def OMP_DoSimd : Directive<"do simd"> {
   let allowedClauses = [
     VersionedClause<OMPC_Aligned>,
let allowedClauses = [
     VersionedClause<OMPC_FirstPrivate>,
     VersionedClause<OMPC_LastPrivate>,
     VersionedClause<OMPC_Default>,
     VersionedClause<OMPC_Untied>,
     VersionedClause<OMPC_Mergeable>,
     VersionedClause<OMPC_NoGroup>,
     VersionedClause<OMPC_Reduction>,
     VersionedClause<OMPC_InReduction>,
-    VersionedClause<OMPC_Allocate>
+    VersionedClause<OMPC_Allocate, 50>
   ];
   let allowedOnceClauses = [
     VersionedClause<OMPC_If>,
     VersionedClause<OMPC_Collapse>,
     VersionedClause<OMPC_Final>,
     VersionedClause<OMPC_Priority>,
   ];
   let allowedExclusiveClauses = [
     VersionedClause<OMPC_GrainSize>,
     VersionedClause<OMPC_NumTasks>
   ];
 }
 def OMP_TaskLoopSimd : Directive<"taskloop simd"> {
   let allowedClauses = [
     VersionedClause<OMPC_Aligned>,
-    VersionedClause<OMPC_Allocate>,
+    VersionedClause<OMPC_Allocate, 50>,
     VersionedClause<OMPC_Default>,
     VersionedClause<OMPC_FirstPrivate>,
     VersionedClause<OMPC_InReduction>,
     VersionedClause<OMPC_LastPrivate>,
     VersionedClause<OMPC_Linear>,
     VersionedClause<OMPC_Mergeable>,
     VersionedClause<OMPC_NoGroup>,
     VersionedClause<OMPC_NonTemporal, 50>,
let allowedExclusiveClauses = [
     VersionedClause<OMPC_NumTasks>
   ];
 }
 def OMP_Distribute : Directive<"distribute"> {
   let allowedClauses = [
     VersionedClause<OMPC_Private>,
     VersionedClause<OMPC_FirstPrivate>,
     VersionedClause<OMPC_LastPrivate>,
-    VersionedClause<OMPC_Allocate>
+    VersionedClause<OMPC_Allocate, 50>
   ];
   let allowedOnceClauses = [
     VersionedClause<OMPC_Collapse>,
     VersionedClause<OMPC_DistSchedule>
   ];
 }
 def OMP_DeclareTarget : Directive<"declare target"> {}
 def OMP_EndDeclareTarget : Directive<"end declare target"> {}
 def OMP_DistributeParallelFor : Directive<"distribute parallel for"> {
   let allowedClauses = [
     VersionedClause<OMPC_FirstPrivate>,
     VersionedClause<OMPC_LastPrivate>,
     VersionedClause<OMPC_Collapse>,
     VersionedClause<OMPC_DistSchedule>,
     VersionedClause<OMPC_If>,
     VersionedClause<OMPC_NumThreads>,
     VersionedClause<OMPC_Default>,
     VersionedClause<OMPC_ProcBind>,
     VersionedClause<OMPC_Private>,
     VersionedClause<OMPC_Shared>,
     VersionedClause<OMPC_Reduction>,
     VersionedClause<OMPC_Copyin>,
     VersionedClause<OMPC_Schedule>,
-    VersionedClause<OMPC_Allocate>,
+    VersionedClause<OMPC_Allocate, 50>,
     VersionedClause<OMPC_Order, 50>
   ];
 }
 def OMP_DistributeParallelDo : Directive<"distribute parallel do"> {
   let allowedClauses = [
     VersionedClause<OMPC_Private>,
     VersionedClause<OMPC_FirstPrivate>,
     VersionedClause<OMPC_LastPrivate>,
let allowedClauses = [
     VersionedClause<OMPC_Shared>,
     VersionedClause<OMPC_Reduction>,
     VersionedClause<OMPC_Copyin>,
     VersionedClause<OMPC_Schedule>,
     VersionedClause<OMPC_Linear>,
     VersionedClause<OMPC_Aligned>,
     VersionedClause<OMPC_SafeLen>,
     VersionedClause<OMPC_SimdLen>,
-    VersionedClause<OMPC_Allocate>,
+    VersionedClause<OMPC_Allocate, 50>,
     VersionedClause<OMPC_NonTemporal, 50>,
     VersionedClause<OMPC_Order, 50>
   ];
 }
 def OMP_DistributeParallelDoSimd : Directive<"distribute parallel do simd"> {
   let allowedClauses = [
     VersionedClause<OMPC_FirstPrivate>,
     VersionedClause<OMPC_LastPrivate>,
let allowedClauses = [
     VersionedClause<OMPC_Allocate>,
     VersionedClause<OMPC_NonTemporal>,
     VersionedClause<OMPC_Order>
   ];
 }
 def OMP_DistributeSimd : Directive<"distribute simd"> {
   let allowedClauses = [
     VersionedClause<OMPC_Aligned>,
-    VersionedClause<OMPC_Allocate>,
+    VersionedClause<OMPC_Allocate, 50>,
     VersionedClause<OMPC_Copyin>,
     VersionedClause<OMPC_Default>,
     VersionedClause<OMPC_Linear>,
     VersionedClause<OMPC_FirstPrivate>,
     VersionedClause<OMPC_LastPrivate>,
     VersionedClause<OMPC_NonTemporal, 50>,
     VersionedClause<OMPC_Order, 50>,
     VersionedClause<OMPC_Private>,
let allowedClauses = [
     VersionedClause<OMPC_Collapse>,
     VersionedClause<OMPC_Schedule>,
     VersionedClause<OMPC_Ordered>,
     VersionedClause<OMPC_Linear>,
     VersionedClause<OMPC_SafeLen>,
     VersionedClause<OMPC_SimdLen>,
     VersionedClause<OMPC_Aligned>,
     VersionedClause<OMPC_IsDevicePtr>,
-    VersionedClause<OMPC_Allocate>,
+    VersionedClause<OMPC_Allocate, 50>,
     VersionedClause<OMPC_NonTemporal, 50>,
     VersionedClause<OMPC_Order, 50>,
     VersionedClause<OMPC_UsesAllocators, 50>
   ];
 }
 def OMP_TargetParallelDoSimd : Directive<"target parallel do simd"> {
   let allowedClauses = [
     VersionedClause<OMPC_If>,
let allowedClauses = [
     VersionedClause<OMPC_NonTemporal>,
     VersionedClause<OMPC_Order>,
     VersionedClause<OMPC_UsesAllocators>
   ];
 }
 def OMP_TargetSimd : Directive<"target simd"> {
   let allowedClauses = [
     VersionedClause<OMPC_Aligned>,
-    VersionedClause<OMPC_Allocate>,
+    VersionedClause<OMPC_Allocate, 50>,
     VersionedClause<OMPC_Depend>,
     VersionedClause<OMPC_FirstPrivate>,
     VersionedClause<OMPC_IsDevicePtr>,
     VersionedClause<OMPC_LastPrivate>,
     VersionedClause<OMPC_Linear>,
     VersionedClause<OMPC_Map>,
     VersionedClause<OMPC_NonTemporal, 50>,
     VersionedClause<OMPC_NoWait>,
let allowedClauses = [
     VersionedClause<OMPC_FirstPrivate>,
     VersionedClause<OMPC_Shared>,
     VersionedClause<OMPC_Reduction>,
     VersionedClause<OMPC_NumTeams>,
     VersionedClause<OMPC_ThreadLimit>,
     VersionedClause<OMPC_LastPrivate>,
     VersionedClause<OMPC_Collapse>,
     VersionedClause<OMPC_DistSchedule>,
-    VersionedClause<OMPC_Allocate>
+    VersionedClause<OMPC_Allocate, 50>
   ];
 }
 def OMP_TeamsDistributeSimd : Directive<"teams distribute simd"> {
   let allowedClauses = [
     VersionedClause<OMPC_Aligned>,
-    VersionedClause<OMPC_Allocate>,
+    VersionedClause<OMPC_Allocate, 50>,
     VersionedClause<OMPC_FirstPrivate>,
     VersionedClause<OMPC_LastPrivate>,
     VersionedClause<OMPC_Linear>,
     VersionedClause<OMPC_NonTemporal, 50>,
     VersionedClause<OMPC_Order, 50>,
     VersionedClause<OMPC_Private>,
     VersionedClause<OMPC_Reduction>,
     VersionedClause<OMPC_Shared>
let allowedClauses = [
     VersionedClause<OMPC_Reduction>,
     VersionedClause<OMPC_Schedule>,
     VersionedClause<OMPC_Linear>,
     VersionedClause<OMPC_Aligned>,
     VersionedClause<OMPC_SafeLen>,
     VersionedClause<OMPC_SimdLen>,
     VersionedClause<OMPC_NumTeams>,
     VersionedClause<OMPC_ThreadLimit>,
-    VersionedClause<OMPC_Allocate>,
+    VersionedClause<OMPC_Allocate, 50>,
     VersionedClause<OMPC_NonTemporal, 50>,
     VersionedClause<OMPC_Order, 50>
   ];
 }
 def OMP_TeamsDistributeParallelDoSimd :
     Directive<"teams distribute parallel do simd"> {
   let allowedClauses = [
     VersionedClause<OMPC_Private>,
let allowedClauses = [
     VersionedClause<OMPC_ProcBind>,
     VersionedClause<OMPC_Private>,
     VersionedClause<OMPC_Shared>,
     VersionedClause<OMPC_Reduction>,
     VersionedClause<OMPC_Schedule>,
     VersionedClause<OMPC_NumTeams>,
     VersionedClause<OMPC_ThreadLimit>,
     VersionedClause<OMPC_Copyin>,
-    VersionedClause<OMPC_Allocate>,
+    VersionedClause<OMPC_Allocate, 50>,
     VersionedClause<OMPC_Order, 50>
   ];
 }
 def OMP_TeamsDistributeParallelDo :
     Directive<"teams distribute parallel do"> {
   let allowedClauses = [
     VersionedClause<OMPC_Private>,
     VersionedClause<OMPC_FirstPrivate>,
     VersionedClause<OMPC_LastPrivate>,
     VersionedClause<OMPC_Shared>,
     VersionedClause<OMPC_Reduction>,
-    VersionedClause<OMPC_Allocate>,
+    VersionedClause<OMPC_Allocate, 50>,
     VersionedClause<OMPC_Copyin>,
     VersionedClause<OMPC_Linear>
   ];
 let allowedOnceClauses = [
     VersionedClause<OMPC_NumTeams>,
     VersionedClause<OMPC_ThreadLimit>,
     VersionedClause<OMPC_Default>,
     VersionedClause<OMPC_Collapse>,
def OMP_TargetTeams : Directive<"target teams"> {
   let allowedClauses = [
     VersionedClause<OMPC_If>,
     VersionedClause<OMPC_Map>,
     VersionedClause<OMPC_Private>,
     VersionedClause<OMPC_Depend>,
     VersionedClause<OMPC_FirstPrivate>,
     VersionedClause<OMPC_IsDevicePtr>,
     VersionedClause<OMPC_Reduction>,
-    VersionedClause<OMPC_Allocate>,
+    VersionedClause<OMPC_Allocate, 50>,
     VersionedClause<OMPC_UsesAllocators, 50>,
     VersionedClause<OMPC_Shared>
   ];
 
   let allowedOnceClauses = [
     VersionedClause<OMPC_Device>,
     VersionedClause<OMPC_NoWait>,
     VersionedClause<OMPC_DefaultMap>,
     VersionedClause<OMPC_Default>,
     VersionedClause<OMPC_NumTeams>,
     VersionedClause<OMPC_ThreadLimit>
   ];
 }
 def OMP_TargetTeamsDistribute : Directive<"target teams distribute"> {
   let allowedClauses = [
     VersionedClause<OMPC_If>,
     VersionedClause<OMPC_Map>,
     VersionedClause<OMPC_Private>,
     VersionedClause<OMPC_Depend>,
     VersionedClause<OMPC_FirstPrivate>,
     VersionedClause<OMPC_IsDevicePtr>,
     VersionedClause<OMPC_Reduction>,
-    VersionedClause<OMPC_Allocate>,
+    VersionedClause<OMPC_Allocate, 50>,
     VersionedClause<OMPC_UsesAllocators, 50>,
     VersionedClause<OMPC_Shared>,
     VersionedClause<OMPC_LastPrivate>
   ];
   let allowedOnceClauses = [
     VersionedClause<OMPC_Device>,
     VersionedClause<OMPC_NoWait>,
     VersionedClause<OMPC_DefaultMap>,
let allowedClauses = [
     VersionedClause<OMPC_NumTeams>,
     VersionedClause<OMPC_ThreadLimit>,
     VersionedClause<OMPC_LastPrivate>,
     VersionedClause<OMPC_Collapse>,
     VersionedClause<OMPC_DistSchedule>,
     VersionedClause<OMPC_NumThreads>,
     VersionedClause<OMPC_ProcBind>,
     VersionedClause<OMPC_Schedule>,
-    VersionedClause<OMPC_Allocate>,
+    VersionedClause<OMPC_Allocate, 50>,
     VersionedClause<OMPC_Order, 50>,
     VersionedClause<OMPC_UsesAllocators, 50>
   ];
 }
 def OMP_TargetTeamsDistributeParallelDo :
     Directive<"target teams distribute parallel do"> {
   let allowedClauses = [
     VersionedClause<OMPC_If>,
let allowedClauses = [
     VersionedClause<OMPC_DistSchedule>,
     VersionedClause<OMPC_NumThreads>,
     VersionedClause<OMPC_ProcBind>,
     VersionedClause<OMPC_Schedule>,
     VersionedClause<OMPC_Linear>,
     VersionedClause<OMPC_Aligned>,
     VersionedClause<OMPC_SafeLen>,
     VersionedClause<OMPC_SimdLen>,
-    VersionedClause<OMPC_Allocate>,
+    VersionedClause<OMPC_Allocate, 50>,
     VersionedClause<OMPC_NonTemporal, 50>,
     VersionedClause<OMPC_Order, 50>,
     VersionedClause<OMPC_UsesAllocators, 50>
   ];
 }
 def OMP_TargetTeamsDistributeParallelDoSimd :
     Directive<"target teams distribute parallel do simd"> {
   let allowedClauses = [
let allowedOnceClauses = [
     VersionedClause<OMPC_SafeLen>,
     VersionedClause<OMPC_SimdLen>
   ];
 }
 def OMP_TargetTeamsDistributeSimd :
     Directive<"target teams distribute simd"> {
   let allowedClauses = [
     VersionedClause<OMPC_Aligned>,
-    VersionedClause<OMPC_Allocate>,
+    VersionedClause<OMPC_Allocate, 50>,
     VersionedClause<OMPC_Depend>,
     VersionedClause<OMPC_FirstPrivate>,
     VersionedClause<OMPC_If>,
     VersionedClause<OMPC_IsDevicePtr>,
     VersionedClause<OMPC_LastPrivate>,
     VersionedClause<OMPC_Linear>,
     VersionedClause<OMPC_Map>,
     VersionedClause<OMPC_NonTemporal, 50>,
let allowedClauses = [
     VersionedClause<OMPC_Untied>,
     VersionedClause<OMPC_Mergeable>,
     VersionedClause<OMPC_Priority>,
     VersionedClause<OMPC_GrainSize>,
     VersionedClause<OMPC_NoGroup>,
     VersionedClause<OMPC_NumTasks>,
     VersionedClause<OMPC_Reduction>,
     VersionedClause<OMPC_InReduction>,
-    VersionedClause<OMPC_Allocate>
+    VersionedClause<OMPC_Allocate, 50>
   ];
 }
 def OMP_ParallelMasterTaskloop :
     Directive<"parallel master taskloop"> {
   let allowedClauses = [
     VersionedClause<OMPC_If>,
     VersionedClause<OMPC_Shared>,
     VersionedClause<OMPC_Private>,
     VersionedClause<OMPC_FirstPrivate>,
     VersionedClause<OMPC_LastPrivate>,
     VersionedClause<OMPC_Default>,
     VersionedClause<OMPC_Collapse>,
     VersionedClause<OMPC_Final>,
     VersionedClause<OMPC_Untied>,
     VersionedClause<OMPC_Mergeable>,
     VersionedClause<OMPC_Priority>,
     VersionedClause<OMPC_GrainSize>,
     VersionedClause<OMPC_NoGroup>,
     VersionedClause<OMPC_NumTasks>,
     VersionedClause<OMPC_Reduction>,
-    VersionedClause<OMPC_Allocate>,
+    VersionedClause<OMPC_Allocate, 50>,
     VersionedClause<OMPC_NumThreads>,
     VersionedClause<OMPC_ProcBind>,
     VersionedClause<OMPC_Copyin>
   ];
 }
 def OMP_MasterTaskloopSimd : Directive<"master taskloop simd"> {
   let allowedClauses = [
     VersionedClause<OMPC_If>,
let allowedClauses = [
     VersionedClause<OMPC_Aligned>,
     VersionedClause<OMPC_SafeLen>,
     VersionedClause<OMPC_SimdLen>,
     VersionedClause<OMPC_GrainSize>,
     VersionedClause<OMPC_NoGroup>,
     VersionedClause<OMPC_NumTasks>,
     VersionedClause<OMPC_Reduction>,
     VersionedClause<OMPC_InReduction>,
-    VersionedClause<OMPC_Allocate>,
+    VersionedClause<OMPC_Allocate, 50>,
     VersionedClause<OMPC_NonTemporal, 50>,
     VersionedClause<OMPC_Order, 50>
   ];
 }
 def OMP_ParallelMasterTaskloopSimd :
     Directive<"parallel master taskloop simd"> {
   let allowedClauses = [
     VersionedClause<OMPC_If>,
     VersionedClause<OMPC_Shared>,
     VersionedClause<OMPC_Private>,
     VersionedClause<OMPC_FirstPrivate>,
     VersionedClause<OMPC_LastPrivate>,
     VersionedClause<OMPC_Default>,
     VersionedClause<OMPC_Collapse>,
     VersionedClause<OMPC_Final>,
     VersionedClause<OMPC_Untied>,
     VersionedClause<OMPC_Mergeable>,
     VersionedClause<OMPC_Priority>,
     VersionedClause<OMPC_GrainSize>,
     VersionedClause<OMPC_NoGroup>,
     VersionedClause<OMPC_NumTasks>,
     VersionedClause<OMPC_Reduction>,
-    VersionedClause<OMPC_Allocate>,
+    VersionedClause<OMPC_Allocate, 50>,
     VersionedClause<OMPC_NumThreads>,
     VersionedClause<OMPC_ProcBind>,
     VersionedClause<OMPC_Copyin>,
     VersionedClause<OMPC_Linear>,
     VersionedClause<OMPC_Aligned>,
     VersionedClause<OMPC_SafeLen>,
     VersionedClause<OMPC_SimdLen>,
     VersionedClause<OMPC_NonTemporal, 50>,
let allowedClauses = [
     VersionedClause<OMPC_Inclusive, 50>,
     VersionedClause<OMPC_Exclusive, 50>
   ];
 }
 def OMP_BeginDeclareVariant : Directive<"begin declare variant"> {}
 def OMP_EndDeclareVariant : Directive<"end declare variant"> {}
 def OMP_ParallelWorkshare : Directive<"parallel workshare"> {
   let allowedClauses = [
-    VersionedClause<OMPC_Allocate>,
+    VersionedClause<OMPC_Allocate, 50>,
     VersionedClause<OMPC_Copyin>,
     VersionedClause<OMPC_Default>,
     VersionedClause<OMPC_FirstPrivate>,
     VersionedClause<OMPC_Private>,
     VersionedClause<OMPC_Reduction>,
     VersionedClause<OMPC_Shared>
   ];
   let allowedOnceClauses = [