[Flang] [OpenMP] Add semantic checks for OpenMP firstprivate , lastprivate and copyprivate clauses

[praveen]

Summary

Add semantic checks and the corresponding lit test cases for the following OpenMP 4.5 clauses

• 2.15.4.2 - Copyprivate clause

• 2.15.3.4 - Firstprivate clause

• 2.15.3.5 - Lastprivate clause

[kiranchandramohan]

Flang.Semantics::omp-clause-validity01.f90

Script:

: 'RUN: at line 1'; /mnt/disks/ssd0/agent/llvm-project/flang/test/Semantics/test_errors.sh /mnt/disks/ssd0/agent/llvm-project/flang/test/Semantics/omp-clause-validity01.f90 /mnt/disks/ssd0/agent/llvm- project/build/tools/flang/test/Semantics/Output/omp-clause-validity01.f90.tmp /mnt/disks/ssd0/agent/llvm-project/build/bin/f18 -intrinsic-module-directory /mnt/disks/ssd0/agent/llvm-project/build/tools/flang/include/flang -fopenmp

Can you check why a unit test is failing?

[praveen]

@kiranchandramohan In the test file omp-clause-validity01.f90 , the following no symbol found errors are not being thrown.

!ERROR: Internal: no symbol found for 'i'
!$omp taskwait depend(sink:i-1)

!ERROR: Internal: no symbol found for 'first'
!$omp critical (first)

Since the errors related to copyprivate clause in omp-clause-validity01.f90 (line numbers 324 and 340 above) being thrown as part of the checks for copyprivate in resolve-directives.cpp are marked as fatal errors , the flag errorOnUnresolvedName_ is being set to false and the no symbol found error is not being thrown.

 RewriteMutator(SemanticsContext &context)
 : errorOnUnresolvedName_{!context.AnyFatalError()},
    messages_{context.messages()} {}

void RewriteMutator::Post(parser::Name &name) {
   if (!name.symbol && errorOnUnresolvedName_) {
     messages_.Say(name.source, "Internal: no symbol found for '%s'"_err_en_US,
       name.source);
    }
 }

Is it not necessary to throw the "no symbol found" error if there is any error marked as fatal while resolving the directives ?

Should all the checks related to copyprivate be moved to check-omp-structure.cpp?

Can you please suggest the approach to follow for these changes?

Thanks!

[kiranchandramohan]

In D91920#2410202, @praveen wrote:

@kiranchandramohan In the test file omp-clause-validity01.f90 , the following no symbol found errors are not being thrown.

!ERROR: Internal: no symbol found for 'i'
!$omp taskwait depend(sink:i-1)

!ERROR: Internal: no symbol found for 'first'
!$omp critical (first)

Since the errors related to copyprivate clause in omp-clause-validity01.f90 (line numbers 324 and 340 above) being thrown as part of the checks for copyprivate in resolve-directives.cpp are marked as fatal errors , the flag errorOnUnresolvedName_ is being set to false and the no symbol found error is not being thrown.

 RewriteMutator(SemanticsContext &context)
 : errorOnUnresolvedName_{!context.AnyFatalError()},
    messages_{context.messages()} {}

void RewriteMutator::Post(parser::Name &name) {
   if (!name.symbol && errorOnUnresolvedName_) {
     messages_.Say(name.source, "Internal: no symbol found for '%s'"_err_en_US,
       name.source);
    }
 }

Is it not necessary to throw the "no symbol found" error if there is any error marked as fatal while resolving the directives ?

Should all the checks related to copyprivate be moved to check-omp-structure.cpp?

Can you please suggest the approach to follow for these changes?

Thanks!

I think these checks can stay in the current file since they need to check the data sharing attributes.
For the depend clause I think the test need to be changed to a valid test. Is that OK @kiranktp.
For critical I believe we may not have done the semantic check and hence the symbol is not available. For this I think it is OK to comment out the check.

Is it not necessary to throw the "no symbol found" error if there is any error marked as fatal while resolving the directives ?

Is this OK @clementval ?

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

806

Is this function part of this patch? If not can you either submit the other patch which introduces this code or make this patch a child of the other patch?

flang/lib/Semantics/resolve-directives.cpp

1337

Will this test erroneously include DSA other than private, firstprivate?

1340

Should there be a check for the SINGLE construct here?

flang/test/Semantics/omp-firstprivate01.f90

2

Can similar tests be all put in the same file? Like all tests which test for 2.15.3.4.

flang/test/Semantics/omp-lastprivate01.f90

5

Can tests be added for the following cases. If it is not covered elsewhere?
-> module variable which is protected.
-> subroutine argument which is intent(in)

flang/test/Semantics/omp-lastprivate02.f90

2

Combine this test and the next one if they are testing the same thing for different worksharing constructs.
Add a test for single as well if possible.

[praveen]

Addressed review comments and combined the test cases

[praveen]

In D91920#2410307, @kiranchandramohan wrote:

In D91920#2410202, @praveen wrote:

@kiranchandramohan In the test file omp-clause-validity01.f90 , the following no symbol found errors are not being thrown.

!ERROR: Internal: no symbol found for 'i'
!$omp taskwait depend(sink:i-1)

!ERROR: Internal: no symbol found for 'first'
!$omp critical (first)

Since the errors related to copyprivate clause in omp-clause-validity01.f90 (line numbers 324 and 340 above) being thrown as part of the checks for copyprivate in resolve-directives.cpp are marked as fatal errors , the flag errorOnUnresolvedName_ is being set to false and the no symbol found error is not being thrown.

 RewriteMutator(SemanticsContext &context)
 : errorOnUnresolvedName_{!context.AnyFatalError()},
    messages_{context.messages()} {}

void RewriteMutator::Post(parser::Name &name) {
   if (!name.symbol && errorOnUnresolvedName_) {
     messages_.Say(name.source, "Internal: no symbol found for '%s'"_err_en_US,
       name.source);
    }
 }

Is it not necessary to throw the "no symbol found" error if there is any error marked as fatal while resolving the directives ?

Should all the checks related to copyprivate be moved to check-omp-structure.cpp?

Can you please suggest the approach to follow for these changes?

Thanks!

I think these checks can stay in the current file since they need to check the data sharing attributes.
For the depend clause I think the test need to be changed to a valid test. Is that OK @kiranktp.
For critical I believe we may not have done the semantic check and hence the symbol is not available. For this I think it is OK to comment out the check.

Is it not necessary to throw the "no symbol found" error if there is any error marked as fatal while resolving the directives ?

Is this OK @clementval ?

@kiranchandramohan @kiranktp @clementval I have commented the test cases which were failing with "no symbol found" error.
The symbol in the sink of the depend clause (OmpDependSinkVec) is not resolved . Hence we are running into "no symbol found" error . Do we resolve it as part of a separate patch or can I include it as part of this patch?

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

806

@kiranchandramohan This function was originally introduced in the patch for private clause https://reviews.llvm.org/D90210 .

Since it is not yet checked in and the same check is required for copyprivate clause, i have included this as part of this patch as well.

806

This check is now merged to master .

flang/lib/Semantics/resolve-directives.cpp

1337

@kiranchandramohan Thanks for pointing this out . Added the checks for private and firstprivate.

1340

@kiranchandramohan Added the SINGLE construct check . Thanks

flang/test/Semantics/omp-firstprivate01.f90

2

@kiranchandramohan Combined all the similar test cases into a single file.

flang/test/Semantics/omp-lastprivate01.f90

5

@kiranchandramohan Added test cases for the module variable which is protected and subroutine argument which is intent(in).

flang/test/Semantics/omp-lastprivate02.f90

2

@kiranchandramohan Combined the test cases as suggested .
Lastprivate clause is not allowed on the single clause.

[kiranchandramohan]

A few comments.

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

763

If this section is common code with firstprivate then sharing code would be good.

766

"A list item that is private within a parallel region, or that appears in the reduction clause of a parallel construct, must not appear in a lastprivate clause on a worksharing construct if any of the corresponding worksharing regions ever binds to any of the corresponding parallel regions."

It seems you are checking for reduction only here. From the standard (quoted above) it appears that we have to check for all kind of privates and reduction.

906

Can this function go to the base class?

[kiranchandramohan]

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

906

Could evaluate::CollectSymbols have been used instead of this since a Designator can also be an expression?

[praveen]

Handle all the private symbols in enclosing context for firstprivate and lastprivate clauses

[praveen]

@kiranchandramohan @sameeranjoshi Will the patch https://reviews.llvm.org/D93105 be committed to the master again?
If so , the function void GetSymbolsInDesignatorList(const std::list<parser::Designator> &, SymbolSourceMap &); would not be required.

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

763

Moved the common code to separate function. Thanks!

766

@kiranchandramohan Handled all the privates and reduction for firstprivate and lastprivate clauses. Thanks!

906

Moved this function to base class

906

@kiranchandramohan I wasnt able to pass the designator as an argument to evaluate::CollectSymbols() . Do we have to convert the designator to some form of expression so that it can be processed as an expression?

[clementval]

In D91920#2475125, @praveen wrote:

@kiranchandramohan @sameeranjoshi Will the patch https://reviews.llvm.org/D93105 be committed to the master again?
If so , the function void GetSymbolsInDesignatorList(const std::list<parser::Designator> &, SymbolSourceMap &); would not be required.

D93105 will be landed again. Waiting for @sameeranjoshi to do it.

[praveen]

Update to latest syntax of Omp Reduction clause

[kiranchandramohan]

I believe the checks have to be made more precise. Please have a look at the comments.

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

857

Can you check why we cannot catch the error in the following test. gfortran seems able to.

program mn
integer :: a(10), b(10), c(10)
!$omp parallel private(a,b)
call tmp()
!$omp end parallel
contains
  subroutine tmp()
  !$omp do firstprivate(b)
  do i = 1, 10
    c(i) = a(i) + b(i) + i
  end do
  !$omp end do
  end subroutine
end program

864

Is there an assumption here that the enclosing directive is the parallel/teams region?

Can it happen that the enclosing immediate directive is not the parallel/teams region to which the worksharing construct binds? Something like the following?

parallel private(x)
  some_omp_construct
    worksharing_construct firstprivate(x)

865

I think whether this check should be performed is dependent on the triple enclosing_dir, enclosing_clause, cur_dir.

So there should be no error for the following code.

program omp_firstprivate

  integer :: i, a(10), b(10), c(10)

  a = 10
  b = 20

  !$omp parallel private(a, b)
  !$omp task firstprivate(a)
  do i = 1, 10
    a(i) = a(i) + b(i) - i
  end do
  !$omp end task
  !$omp end parallel
end program

But there should be an error for the following code.

program omp_firstprivate
  integer :: i, a
  a = 10
  i = 2
  !$omp parallel reduction(+:a)
  !$omp task firstprivate(a)
    a = a + i
  !$omp end task
  !$omp end parallel
end program

flang/test/Semantics/omp-copyprivate03.f90

27

A threadprivate variable cannot appear in a private clause.

[praveen]

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

857

@kiranchandramohan We are not able to catch the error in the above case since the context information is cleared in the ResetPartialContext() after the OpenMP End construct is encountered.

Hence, when checking the enclosing directives for firstprivate(b), it returns null.
I tried to keep track of the context information past the execution of the construct , but it ran into other issues for cases which expects the context information to be cleared.

864

@kiranchandramohan I am not sure if there can be any immediate enclosing directives for workshare constructs .

Since we already check 'HasInvalidWorksharingNesting' , do we need to check for the 'firstprivate/lastprivate variables that are private in enclosing context' for the variables on the worksharing constructs?

The task construct can contain directives other than parallel such as the following.

!$omp parallel reduction (+:a)
  !$omp single
    !$omp task firstprivate (a)
      a = a + b
    !$omp end task
  !$omp end single
!$omp end parallel

Should we thrown an error for the cases such as the one above ?

865

@kiranchandramohan Thanks . Will handle this check

906

Removed the function as OmpObjectList is being used for Reduction clause instead of a list of Designators.

[praveen]

Address review comments and rebase

[kiranchandramohan]

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

857

What did you finally do for this case? If it is not handled can you add an entry into the spreadsheet for this case?

864

A list item that appears in a reduction clause of a parallel construct must not appear in a
firstprivate clause on a worksharing, task, or taskloop construct if any of the
worksharing or task regions arising from the worksharing, task, or taskloop construct ever
bind to any of the parallel regions arising from the parallel construct.

As per the above statement from the standard, yes it has to be flagged. If there are issues in handling this then please let me know.

[praveen]

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

857

@kiranchandramohan I will add an entry to the spreadsheet for this case.
I guess we may have to explore to check if a separate parse-tree walk would help for such cases .

I am not sure if I am missing something or gfortran (gcc 9.3.0) does not catch the above mentioned error as expected .

firstprivate01.f90

program mn
integer :: a(10), b(10), c(10)
!$omp parallel shared(a,b)
call tmp()
!$omp end parallel
contains
  subroutine tmp()
  !$omp do firstprivate(b)
  do i = 1, 10
    c(i) = a(i) + b(i) + i
  end do
  !$omp end do
  end subroutine
end program

firstprivate02.f90

program mn
integer :: a(10), b(10), c(10)
!$omp parallel
call tmp()
!$omp end parallel
contains
  subroutine tmp()
  !$omp do firstprivate(a)
  do i = 1, 10
    c(i) = a(i) + b(i) + i
  end do
  !$omp end do
  end subroutine
end program

firstprivate03.f90

program mn
integer :: a(10), b(10), c(10), i
!$omp do firstprivate(a)
do i = 1, 10
  c(i) = a(i) + b(i) + i
end do
!$omp end do
end program

In the above test cases firstprivate01.f90, firstprivate02.f90 and firstprivate03.f90 , the firstprivate symbol is not added as private in the enclosing context.

But gfortran identifies the above cases as errors. In gfortran , it seems that if the variable is not explicitly specified in any of the clauses , then it is assumed as private implicity.

Please let me know if it is the expected behaviour?

nesting.f90

program mn
integer :: a(10), b(10), c(10)
!$omp parallel
!$omp single
call tmp()
!$omp end single
!$omp end parallel
contains
  subroutine tmp()
  !$omp do
  do i = 1, 10
    c(i) = a(i) + b(i) + i
  end do
  !$omp end do
  end subroutine
end program

In the above program nesting.f90, should there be an "invalid nesting of worksharing regions" error ?

copyprivate.f90

program mn
integer :: a(10), b(10), c(10)
!$omp parallel
call tmp()
!$omp end parallel
contains
  subroutine tmp()
  !$omp single
   c = a + b
  !$omp end single copyprivate(a)
  end subroutine
end program

For the above case copyprivate.f90, we are identifying this as an error since "copyprivate variable is not private or thread-private in outer context" , but gfortran does not throw the error similar to the above cases.

864

@kiranchandramohan Have updated the code to identify the enclosing constructs that are not immediate enclosing directives and binds to the current directive.

864

@kiranchandramohan I have updated the code to handle cases such as the above example where the binding region may not be the immediate enclosing directive.

865

Updated the logic to check based on the triple.

flang/test/Semantics/omp-copyprivate03.f90

27

A threadprivate variable cannot appear in a private clause.

27

Done

[kiranchandramohan]

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

857

Thanks @praveen for your questions.
I believe we are talking about the following restrictions for,

1. firstprivate: "A list item that is private within a parallel region must not appear in a firstprivate clause on a worksharing construct if any of the worksharing regions arising from the worksharing construct ever bind to any of the parallel regions arising from the parallel construct."

2. nesting: "A worksharing region may not be closely nested inside a worksharing, explicit task, taskloop, critical, ordered, atomic, or master region"

3. copyprivate: "All list items that appear in the copyprivate clause must be either threadprivate or private in the enclosing context"

I think gfortran is giving an incorrect messages here since for (1), (2), (3).
But for (3) we must be careful since it talks about "enclosing context" and not regions. Enclosing context in Fortran is defined as the innermost scoping unit enclosing an OpenMP directive.

[praveen]

Check threadprivate variables in symbols with HostAssocDetails

[praveen]

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

857

@kiranchandramohan Thanks for the explanation . I have added an entry in the spreadsheet for the above cases as part of Nesting of Regions part.

For the restrictions on the copyprivate variables , we are checking in the enclosing scope only.

[kiranchandramohan]

Quick comments/questions before I have a final look.

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

524

Nit: Add an empty line above.

887–894

Will this clash with the reduction work that @yhegde is doing in https://reviews.llvm.org/D90697?

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

81–83

Can you add comments on why these Data Structures are necessary either here or where they are used?

[kiranchandramohan]

A few more questions.

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

864

Could you point me to where and how this was handled?

869

Having to hard-code this looks unfortunate. Can you add a TODO to replace this hardcoding? Try one of the following,

1. autogen checks
2. or an autogenerated function which maps parser::OmpClause::<name> to llvm::omp::Clause::OMPC_<name>

flang/lib/Semantics/resolve-directives.cpp

1334

Nit: would it be better to add a check for Symbol::Flag::OmpCopyPrivate here and then do the remaining checks inside?

Otherwise add a check or assert that it is CopyPrivate.

[praveen]

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

864

@kiranchandramohan The function GetEnclosingContextWithDir returns the matching directive in the dirContext and it is called in check-omp-structure.cpp:1011

DirectiveContext *GetEnclosingContextWithDir(D dir) {
  CHECK(!dirContext_.empty());
  auto it{dirContext_.rbegin()};
  while (++it != dirContext_.rend()) {
    if (it->directive == dir) {
      return &(*it);
    }
  }
  return nullptr;
}

869

@kiranchandramohan Will try to add autogenerated function / autogen checks for the same

887–894

@kiranchandramohan In this patch , only the variables in the firstprivate and lastprivate clauses on the worksharing constructs are being checked while in https://reviews.llvm.org/D90697 added by @yhegde , variables in the reduction clause is checked . So , I dont think there would be a clash regarding the checks.

But the checks for privates and reduction in the enclosing region is common for both.

[kiranchandramohan]

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

869

@clementval Is it currently possible to get the llvm::omp::Clause corresponding to a parser::OmpClause in a function?
From the top of your head is there a different way to write this std::visit?

[clementval]

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

869

We do not have such a function generated by TableGen right now. That would be a nice addition and useful in many places.

[kiranchandramohan]

If you can add the autogen portion that will be great. It can be added either as part of this patch or a separate patch. But it need not be a blocker for this. I am approving this patch.

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

864

OK.

869

Thanks valentine.

869

OK @praveen, see whether you can add them but that is not a blocker for this patch.

[praveen]

Add TODO and comments for the data structures

[praveen]

@kiranchandramohan The test case flang/test/Semantics/resolve102.f90 is failing with the error messages possibly due to the commit

https://github.com/llvm/llvm-project/commit/07de0846a5055015b55dc2b8faa2143f9902e549#diff-bc82e30d81709b26f2c8030300784e95003f1b968e1ed7842fa717858e497336

actual at 13: Procedure 'p' is recursively defined. Procedures in the cycle: 'p2', 'sub', 'p'
expect at 13: Procedure 'p' is recursively defined. Procedures in the cycle: 'p2', 'p', 'sub'
actual at 25: Procedure 'p' is recursively defined. Procedures in the cycle: 'p2', 'sub', 'p'
expect at 25: Procedure 'p' is recursively defined. Procedures in the cycle: 'p2', 'p', 'sub'

Can I push this patch now or only after the above failed case is fixed ?

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

869

@kiranchandramohan @clementval Thanks . I have added a TODO to replace the hard-coded names as part of this patch.

Will create a separate patch if it is possible to add an auto generated function to maintain the map.

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

81–83

Added the comments.

flang/lib/Semantics/resolve-directives.cpp

1334

Added a check for Symbol::Flag::OmpCopyPrivate

[kiranchandramohan]

In D91920#2593420, @praveen wrote:

@kiranchandramohan The test case flang/test/Semantics/resolve102.f90 is failing with the error messages possibly due to the commit

https://github.com/llvm/llvm-project/commit/07de0846a5055015b55dc2b8faa2143f9902e549#diff-bc82e30d81709b26f2c8030300784e95003f1b968e1ed7842fa717858e497336

actual at 13: Procedure 'p' is recursively defined. Procedures in the cycle: 'p2', 'sub', 'p'
expect at 13: Procedure 'p' is recursively defined. Procedures in the cycle: 'p2', 'p', 'sub'
actual at 25: Procedure 'p' is recursively defined. Procedures in the cycle: 'p2', 'sub', 'p'
expect at 25: Procedure 'p' is recursively defined. Procedures in the cycle: 'p2', 'p', 'sub'

Can I push this patch now or only after the above failed case is fixed ?

You can push it now since the failure is unrelated to your patch.

flang/include/flang/Semantics/symbol.h

ENUM_CLASS(Flag,
       AccCopyIn, AccCopyOut, AccCreate, AccDelete, AccPresent,
       // 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 data-copying attribute
-      OmpCopyIn,
+      OmpCopyIn, OmpCopyPrivate,
       // OpenMP miscellaneous flags
       OmpCommonBlock, OmpReduction, OmpAllocate, OmpDeclareSimd,
       OmpDeclareTarget, OmpThreadprivate, OmpDeclareReduction, OmpFlushed,
       OmpCriticalLock, OmpIfSpecified, OmpNone, OmpPreDetermined, OmpAligned);
   using Flags = common::EnumSet<Flag, Flag_enumSize>;
 
   const Scope &owner() const { return *owner_; }
   const SourceName &name() const { return name_; }

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

const PC *FindClause(C type) {
     }
     return nullptr;
   }
 
   void PushContext(const parser::CharBlock &source, D dir) {
     dirContext_.emplace_back(source, dir);
   }
 
+  DirectiveContext *GetEnclosingDirContext() {
+    CHECK(!dirContext_.empty());
+    auto it{dirContext_.rbegin()};
+    if (++it != dirContext_.rend()) {
+      return &(*it);
+    }
+    return nullptr;
+  }
+
   bool CurrentDirectiveIsNested() { return dirContext_.size() > 0; };
 
   void SetClauseSets(D dir) {
     dirContext_.back().allowedClauses = directiveClausesMap_[dir].allowed;
     dirContext_.back().allowedOnceClauses =
         directiveClausesMap_[dir].allowedOnce;
     dirContext_.back().allowedExclusiveClauses =
         directiveClausesMap_[dir].allowedExclusive;

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

static OmpDirectiveSet simdSet{Directive::OMPD_distribute_parallel_do_simd,
     Directive::OMPD_teams_distribute_simd};
 static OmpDirectiveSet taskGeneratingSet{
     OmpDirectiveSet{Directive::OMPD_task} | taskloopSet};
 } // namespace omp
 } // namespace llvm
 
 namespace Fortran::semantics {
 
+using SymbolSourceMap = std::multimap<const Symbol *, parser::CharBlock>;
+
 class OmpStructureChecker
     : public DirectiveStructureChecker<llvm::omp::Directive, llvm::omp::Clause,
           parser::OmpClause, llvm::omp::Clause_enumSize> {
 public:
   OmpStructureChecker(SemanticsContext &context)
       : DirectiveStructureChecker(context,
 #define GEN_FLANG_DIRECTIVE_CLAUSE_MAP
 #include "llvm/Frontend/OpenMP/OMP.inc"
         ) {
private:
   llvm::StringRef getDirectiveName(llvm::omp::Directive directive) override;
 
   void CheckDependList(const parser::DataRef &);
   void CheckDependArraySection(
       const common::Indirection<parser::ArrayElement> &, const parser::Name &);
   void CheckIsVarPartOfAnotherVar(const parser::OmpObjectList &objList);
   void CheckIntentInPointer(
       const parser::OmpObjectList &, const llvm::omp::Clause);
-  void GetSymbolsInObjectList(
-      const parser::OmpObjectList &, std::vector<const Symbol *> &);
+  void GetSymbolsInObjectList(const parser::OmpObjectList &, SymbolSourceMap &);
+  void CheckDefinableObjects(SymbolSourceMap &, const llvm::omp::Clause);
+  void CheckPrivateSymbolsInOuterCxt(SymbolSourceMap &, const OmpDirectiveSet &,
+      const OmpDirectiveSet &, const llvm::omp::Clause);
 };
 } // namespace Fortran::semantics
 #endif // FORTRAN_SEMANTICS_CHECK_OMP_STRUCTURE_H_

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

 void OmpStructureChecker::Enter(const parser::OmpClause &x) {
   SetContextClause(x);
 }
 
 // Following clauses do not have a seperate node in parse-tree.h.
 // They fall under 'struct OmpClause' in parse-tree.h.
 CHECK_SIMPLE_CLAUSE(Allocate, OMPC_allocate)
 CHECK_SIMPLE_CLAUSE(Copyin, OMPC_copyin)
-CHECK_SIMPLE_CLAUSE(Copyprivate, OMPC_copyprivate)
 CHECK_SIMPLE_CLAUSE(Default, OMPC_default)
 CHECK_SIMPLE_CLAUSE(Device, OMPC_device)
 CHECK_SIMPLE_CLAUSE(Final, OMPC_final)
-CHECK_SIMPLE_CLAUSE(Firstprivate, OMPC_firstprivate)
 CHECK_SIMPLE_CLAUSE(From, OMPC_from)
 CHECK_SIMPLE_CLAUSE(Inbranch, OMPC_inbranch)
 CHECK_SIMPLE_CLAUSE(IsDevicePtr, OMPC_is_device_ptr)
-CHECK_SIMPLE_CLAUSE(Lastprivate, OMPC_lastprivate)
 CHECK_SIMPLE_CLAUSE(Link, OMPC_link)
 CHECK_SIMPLE_CLAUSE(Mergeable, OMPC_mergeable)
 CHECK_SIMPLE_CLAUSE(Nogroup, OMPC_nogroup)
 CHECK_SIMPLE_CLAUSE(Notinbranch, OMPC_notinbranch)
 CHECK_SIMPLE_CLAUSE(Nowait, OMPC_nowait)
 CHECK_SIMPLE_CLAUSE(Reduction, OMPC_reduction)
 CHECK_SIMPLE_CLAUSE(TaskReduction, OMPC_task_reduction)
 CHECK_SIMPLE_CLAUSE(To, OMPC_to)

 void OmpStructureChecker::Leave(const parser::OmpAtomic &) {
   if (const auto *clause{FindClause(llvm::omp::Clause::OMPC_acquire)}) {
     context_.Say(clause->source,
         "Clause ACQUIRE is not allowed on the ATOMIC directive"_err_en_US);
   }
   if (const auto *clause{FindClause(llvm::omp::Clause::OMPC_acq_rel)}) {
     context_.Say(clause->source,
         "Clause ACQ_REL is not allowed on the ATOMIC directive"_err_en_US);
   }
 }
 // Restrictions specific to each clause are implemented apart from the
 // generalized restrictions.
 void OmpStructureChecker::Enter(const parser::OmpAlignedClause &x) {
   CheckAllowed(llvm::omp::Clause::OMPC_aligned);
 
   if (const auto &expr{
           std::get<std::optional<parser::ScalarIntConstantExpr>>(x.t)}) {
for (const auto &ele : designators) {
                     &dataRef->u)}) {
           CheckDependArraySection(*arr, GetLastName(*dataRef));
         }
       }
     }
   }
 }
 
+void OmpStructureChecker::Enter(const parser::OmpClause::Copyprivate &x) {
+  CheckAllowed(llvm::omp::Clause::OMPC_copyprivate);
+  CheckIntentInPointer(x.v, llvm::omp::Clause::OMPC_copyprivate);
+}
+
+void OmpStructureChecker::Enter(const parser::OmpClause::Firstprivate &x) {
+  CheckAllowed(llvm::omp::Clause::OMPC_firstprivate);
+
+  SymbolSourceMap currSymbols, enclosingSymbols;
+  GetSymbolsInObjectList(x.v, currSymbols);
+
+  const OmpDirectiveSet currDirSet{llvm::omp::Directive::OMPD_do,
+      llvm::omp::Directive::OMPD_sections,
+      llvm::omp::Directive::OMPD_distribute, llvm::omp::Directive::OMPD_task,
+      llvm::omp::Directive::OMPD_taskloop};
+  const OmpDirectiveSet enclosingDirSet{llvm::omp::Directive::OMPD_parallel,
+      llvm::omp::Directive::OMPD_teams,
+      llvm::omp::Directive::OMPD_target_teams};
+  CheckPrivateSymbolsInOuterCxt(currSymbols, currDirSet, enclosingDirSet,
+      llvm::omp::Clause::OMPC_firstprivate);
+}
+
+void OmpStructureChecker::Enter(const parser::OmpClause::Lastprivate &x) {
+  CheckAllowed(llvm::omp::Clause::OMPC_lastprivate);
+
+  SymbolSourceMap currSymbols, enclosingSymbols;
+  GetSymbolsInObjectList(x.v, currSymbols);
+  CheckDefinableObjects(currSymbols, llvm::omp::Clause::OMPC_lastprivate);
+
+  const OmpDirectiveSet currDirSet{
+      llvm::omp::Directive::OMPD_do, llvm::omp::Directive::OMPD_sections};
+  const OmpDirectiveSet enclosingDirSet{llvm::omp::Directive::OMPD_parallel};
+  CheckPrivateSymbolsInOuterCxt(currSymbols, currDirSet, enclosingDirSet,
+      llvm::omp::Clause::OMPC_lastprivate);
+}
+
 llvm::StringRef OmpStructureChecker::getClauseName(llvm::omp::Clause clause) {
   return llvm::omp::getOpenMPClauseName(clause);
 }
 
 llvm::StringRef OmpStructureChecker::getDirectiveName(
     llvm::omp::Directive directive) {
   return llvm::omp::getOpenMPDirectiveName(directive);
 }
 
 void OmpStructureChecker::CheckDependList(const parser::DataRef &d) {
   std::visit(
       common::visitors{
           [&](const common::Indirection<parser::ArrayElement> &elem) {
             // Check if the base element is valid on Depend Clause
             CheckDependList(elem.value().base);
           },
           [&](const common::Indirection<parser::StructureComponent> &) {
             context_.Say(GetContext().clauseSource,
                 "A variable that is part of another variable "
                 "(such as an element of a structure) but is not an array "
                 "element or an array section cannot appear in a DEPEND "
                 "clause"_err_en_US);
           },
           [&](const common::Indirection<parser::CoindexedNamedObject> &) {
             context_.Say(GetContext().clauseSource,
                 "Coarrays are not supported in DEPEND clause"_err_en_US);
           },
           [&](const parser::Name &) { return; },
if (const auto *triplet{
               name.ToString());
           break;
         }
       }
     }
   }
 }
 
 void OmpStructureChecker::CheckIntentInPointer(
     const parser::OmpObjectList &objectList, const llvm::omp::Clause clause) {
-  std::vector<const Symbol *> symbols;
+  SymbolSourceMap symbols;
   GetSymbolsInObjectList(objectList, symbols);
-  for (const auto *symbol : symbols) {
+  for (auto it{symbols.begin()}; it != symbols.end(); ++it) {
+    const auto *symbol{it->first};
+    const auto source{it->second};
     if (IsPointer(*symbol) && IsIntentIn(*symbol)) {
-      context_.Say(GetContext().clauseSource,
+      context_.Say(source,
           "Pointer '%s' with the INTENT(IN) attribute may not appear "
           "in a %s clause"_err_en_US,
           symbol->name(),
           parser::ToUpperCaseLetters(getClauseName(clause).str()));
     }
   }
 }
 
 void OmpStructureChecker::GetSymbolsInObjectList(
-    const parser::OmpObjectList &objectList,
-    std::vector<const Symbol *> &symbols) {
+    const parser::OmpObjectList &objectList, SymbolSourceMap &symbols) {
   for (const auto &ompObject : objectList.v) {
     if (const auto *name{parser::Unwrap<parser::Name>(ompObject)}) {
       if (const auto *symbol{name->symbol}) {
         if (const auto *commonBlockDetails{
                 symbol->detailsIf<CommonBlockDetails>()}) {
           for (const auto &object : commonBlockDetails->objects()) {
-            symbols.emplace_back(&object->GetUltimate());
+            symbols.emplace(&object->GetUltimate(), name->source);
           }
         } else {
-          symbols.emplace_back(&symbol->GetUltimate());
+          symbols.emplace(&symbol->GetUltimate(), name->source);
+        }
+      }
+    }
+  }
+}
+
+void OmpStructureChecker::CheckDefinableObjects(
+    SymbolSourceMap &symbols, const llvm::omp::Clause clause) {
+  for (auto it{symbols.begin()}; it != symbols.end(); ++it) {
+    const auto *symbol{it->first};
+    const auto source{it->second};
+    if (auto msg{WhyNotModifiable(*symbol, context_.FindScope(source))}) {
+      context_
+          .Say(source,
+              "Variable '%s' on the %s clause is not definable"_err_en_US,
+              symbol->name(),
+              parser::ToUpperCaseLetters(getClauseName(clause).str()))
+          .Attach(source, std::move(*msg), symbol->name());
+    }
+  }
+}
+
+void OmpStructureChecker::CheckPrivateSymbolsInOuterCxt(
+    SymbolSourceMap &currSymbols, const OmpDirectiveSet &currDirSet,
+    const OmpDirectiveSet &enclosingDirSet,
+    const llvm::omp::Clause currClause) {
+  SymbolSourceMap enclosingSymbols;
+
+  if (currDirSet.test(GetContext().directive)) {
+    if (auto *enclosingContext{GetEnclosingDirContext()}) {
+      if (enclosingDirSet.test(enclosingContext->directive)) {
+        for (auto it{enclosingContext->clauseInfo.begin()};
+             it != enclosingContext->clauseInfo.end(); ++it) {
+          std::visit(common::visitors{
+                         [&](const parser::OmpClause::Private &x) {
+                           GetSymbolsInObjectList(x.v, enclosingSymbols);
+                         },
+                         [&](const parser::OmpClause::Firstprivate &x) {
+                           GetSymbolsInObjectList(x.v, enclosingSymbols);
+                         },
+                         [&](const parser::OmpClause::Lastprivate &x) {
+                           GetSymbolsInObjectList(x.v, enclosingSymbols);
+                         },
+                         [&](const parser::OmpClause::Reduction &x) {
+                           const auto &ompObjectList{
+                               std::get<parser::OmpObjectList>(x.v.t)};
+                           GetSymbolsInObjectList(
+                               ompObjectList, enclosingSymbols);
+                         },
+                         [&](const auto &) {},
+                     },
+              it->second->u);
+        }
+
+        // Check if the symbols in current context are private in outer context
+        for (auto iter{currSymbols.begin()}; iter != currSymbols.end();
+             ++iter) {
+          const auto *symbol{iter->first};
+          const auto source{iter->second};
+          if (enclosingSymbols.find(symbol) != enclosingSymbols.end()) {
+            context_.Say(source,
+                "%s variable '%s' is PRIVATE in outer context"_err_en_US,
+                parser::ToUpperCaseLetters(getClauseName(currClause).str()),
+                symbol->name());
+          }
         }
       }
     }
   }
 }
 
 } // namespace Fortran::semantics

flang/lib/Semantics/resolve-directives.cpp

public:
   bool Pre(const parser::OmpClause::Lastprivate &x) {
     ResolveOmpObjectList(x.v, Symbol::Flag::OmpLastPrivate);
     return false;
   }
   bool Pre(const parser::OmpClause::Copyin &x) {
     ResolveOmpObjectList(x.v, Symbol::Flag::OmpCopyIn);
     return false;
   }
+  bool Pre(const parser::OmpClause::Copyprivate &x) {
+    ResolveOmpObjectList(x.v, Symbol::Flag::OmpCopyPrivate);
+    return false;
+  }
   bool Pre(const parser::OmpLinearClause &x) {
     std::visit(common::visitors{
                    [&](const parser::OmpLinearClause::WithoutModifier
                            &linearWithoutModifier) {
                      ResolveOmpNameList(
                          linearWithoutModifier.names, Symbol::Flag::OmpLinear);
                    },
                    [&](const parser::OmpLinearClause::WithModifier
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};
 
   static constexpr Symbol::Flags dataCopyingAttributeFlags{
-      Symbol::Flag::OmpCopyIn};
+      Symbol::Flag::OmpCopyIn, Symbol::Flag::OmpCopyPrivate};
 
   std::vector<const parser::Name *> allocateNames_; // on one directive
   SymbolSet privateDataSharingAttributeObjects_; // on one directive
   SymbolSet stmtFunctionExprSymbols_;
 
   void AddAllocateName(const parser::Name *&object) {
     allocateNames_.push_back(object);
   }
private:
   Symbol *ResolveName(const parser::Name *);
   Symbol *DeclareOrMarkOtherAccessEntity(const parser::Name &, Symbol::Flag);
   Symbol *DeclareOrMarkOtherAccessEntity(Symbol &, Symbol::Flag);
   void CheckMultipleAppearances(
       const parser::Name &, const Symbol &, Symbol::Flag);
 
   void CheckDataCopyingClause(
       const parser::Name &, const Symbol &, Symbol::Flag);
-
   void CheckAssocLoopLevel(std::int64_t level, const parser::OmpClause *clause);
   void CheckPrivateDSAObject(
       const parser::Name &, const Symbol &, Symbol::Flag);
+  bool HasSymbolInEnclosingScope(const Symbol &, Scope &);
 };
 
 template <typename T>
 bool DirectiveAttributeVisitor<T>::HasDataSharingAttributeObject(
     const Symbol &object) {
   auto it{dataSharingAttributeObjects_.find(object)};
   return it != dataSharingAttributeObjects_.end();
 }
if (const auto *details{symbol.detailsIf<HostAssocDetails>()})
       checkSymbol = &details->symbol();
 
     // List of items/objects that can appear in a 'copyin' clause must be
     // 'threadprivate'
     if (!checkSymbol->test(Symbol::Flag::OmpThreadprivate))
       context_.Say(name.source,
           "Non-THREADPRIVATE object '%s' in COPYIN clause"_err_en_US,
           checkSymbol->name());
+  } else if (GetContext().directive == llvm::omp::Directive::OMPD_single) {
+    // A list item that appears in a 'copyprivate' clause may not appear on a
+    // 'private' or 'firstprivate' clause on a single construct
+    if (IsObjectWithDSA(symbol) &&
+        (symbol.test(Symbol::Flag::OmpPrivate) ||
+            symbol.test(Symbol::Flag::OmpFirstPrivate))) {
+      context_.Say(name.source,
+          "COPYPRIVATE variable '%s' may not appear on a PRIVATE or "
+          "FIRSTPRIVATE clause on a SINGLE construct"_err_en_US,
+          symbol.name());
+    } else {
+      // List of items/objects that can appear in a 'copyprivate' clause must be
+      // either 'private' or 'threadprivate' in enclosing context.
+      if (!symbol.test(Symbol::Flag::OmpThreadprivate) &&
+          !(HasSymbolInEnclosingScope(symbol, currScope()) &&
+              symbol.test(Symbol::Flag::OmpPrivate))) {
+        context_.Say(name.source,
+            "COPYPRIVATE variable '%s' is not PRIVATE or THREADPRIVATE in "
+            "outer context"_err_en_US,
+            symbol.name());
+      }
+    }
   }
 }
 
 void OmpAttributeVisitor::CheckPrivateDSAObject(
     const parser::Name &name, const Symbol &symbol, Symbol::Flag ompFlag) {
   const auto &ultimateSymbol{symbol.GetUltimate()};
   llvm::StringRef clauseName{"PRIVATE"};
   if (ompFlag == Symbol::Flag::OmpFirstPrivate)
if (stmtFunctionExprSymbols_.find(ultimateSymbol) !=
       stmtFunctionExprSymbols_.end()) {
     context_.Say(name.source,
         "Variable '%s' in STATEMENT FUNCTION expression cannot be in a "
         "%s clause"_err_en_US,
         name.ToString(), clauseName.str());
   }
 }
 
+bool OmpAttributeVisitor::HasSymbolInEnclosingScope(
+    const Symbol &symbol, Scope &scope) {
+  const auto symbols{scope.parent().GetSymbols()};
+  auto it{std::find(symbols.begin(), symbols.end(), symbol)};
+  return it != symbols.end();
+}
+
 } // namespace Fortran::semantics

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

 !                        nowait-clause
 
   !$omp parallel
   b = 1
   !ERROR: LASTPRIVATE clause is not allowed on the SINGLE directive
   !$omp single private(a) lastprivate(c)
   a = 3.14
   !ERROR: Clause NOWAIT is not allowed if clause COPYPRIVATE appears on the END SINGLE directive
+  !ERROR: COPYPRIVATE variable 'a' may not appear on a PRIVATE or FIRSTPRIVATE clause on a SINGLE construct
   !ERROR: At most one NOWAIT clause can appear on the END SINGLE directive
   !$omp end single copyprivate(a) nowait nowait
   c = 2
   !$omp end parallel
 
 ! 2.7.4 workshare
 
   !$omp parallel
! 2.13.1 master
   !$omp end parallel
 
 ! Standalone Directives (basic)
 
   !$omp taskyield
   !$omp barrier
   !$omp taskwait
   !$omp taskwait depend(source)
-  !ERROR: Internal: no symbol found for 'i'
-  !$omp taskwait depend(sink:i-1)
+  ! !$omp taskwait depend(sink:i-1)
   ! !$omp target enter data map(to:arrayA) map(alloc:arrayB)
   ! !$omp target update from(arrayA) to(arrayB)
   ! !$omp target exit data map(from:arrayA) map(delete:arrayB)
   !$omp ordered depend(source)
-  !ERROR: Internal: no symbol found for 'i'
-  !$omp ordered depend(sink:i-1)
+  ! !$omp ordered depend(sink:i-1)
   !$omp flush (c)
   !$omp flush acq_rel
   !$omp flush release
   !$omp flush acquire
   !ERROR: If memory-order-clause is RELEASE, ACQUIRE, or ACQ_REL, list items must not be specified on the FLUSH directive
   !$omp flush release (c)
   !ERROR: SEQ_CST clause is not allowed on the FLUSH directive
   !$omp flush seq_cst
   !ERROR: RELAXED clause is not allowed on the FLUSH directive
   !$omp flush relaxed
 
   !$omp cancel DO
   !$omp cancellation point parallel
 
 ! 2.13.2 critical Construct
 
-  !ERROR: Internal: no symbol found for 'first'
-  !$omp critical (first)
+  ! !$omp critical (first)
   a = 3.14
-  !ERROR: Internal: no symbol found for 'first'
-  !$omp end critical (first)
+  ! !$omp end critical (first)
 
 ! 2.9.1 task-clause -> if-clause |
 !                      final-clause |
 !                      untied-clause |
 !                      default-clause |
 !                      mergeable-clause |
 !                      private-clause |
 !                      firstprivate-clause |

flang/test/Semantics/omp-copyprivate01.f90

+! RUN: %S/test_errors.sh %s %t %f18 -fopenmp
+! OpenMP Version 4.5
+! 2.15.4.2 copyprivate Clause
+! A list item that appears in a copyprivate clause may not appear in a
+! private or firstprivate clause on the single construct.
+
+program omp_copyprivate
+  integer :: a(10), b(10), k
+
+  k = 10
+  a = 10
+  b = a * 10
+
+  !$omp parallel
+  !$omp single private(k)
+  a = a + k
+  !ERROR: COPYPRIVATE variable 'k' may not appear on a PRIVATE or FIRSTPRIVATE clause on a SINGLE construct
+  !$omp end single copyprivate(k)
+  !$omp single firstprivate(k)
+  b = a - k
+  !ERROR: COPYPRIVATE variable 'k' may not appear on a PRIVATE or FIRSTPRIVATE clause on a SINGLE construct
+  !$omp end single copyprivate(k)
+  !$omp end parallel
+
+  print *, a, b
+
+end program omp_copyprivate

flang/test/Semantics/omp-copyprivate02.f90

+! RUN: %S/test_errors.sh %s %t %f18 -fopenmp
+! OpenMP Version 4.5
+! 2.15.4.2 copyprivate Clause
+! Pointers with the INTENT(IN) attribute may not appear in a copyprivate clause.
+
+subroutine omp_copyprivate(p)
+  integer :: a(10), b(10), c(10)
+  integer, pointer, intent(in) :: p
+
+  a = 10
+  b = 20
+
+  !$omp parallel
+  !$omp single
+  c = a + b + p
+  !ERROR: COPYPRIVATE variable 'p' is not PRIVATE or THREADPRIVATE in outer context
+  !ERROR: Pointer 'p' with the INTENT(IN) attribute may not appear in a COPYPRIVATE clause
+  !$omp end single copyprivate(p)
+  !$omp end parallel
+
+  print *, c
+
+end subroutine omp_copyprivate

flang/test/Semantics/omp-copyprivate03.f90

+! RUN: %S/test_errors.sh %s %t %f18 -fopenmp
+! OpenMP Version 4.5
+! 2.15.4.2 copyprivate Clause
+! All list items that appear in the copyprivate clause must be either
+! threadprivate or private in the enclosing context.
+
+program omp_copyprivate
+  integer :: a(10), b(10)
+  integer, save :: k
+
+  !$omp threadprivate(k)
+  
+  k = 10
+  a = 10
+  b = a + 10
+
+  !$omp parallel 
+  !$omp single
+  a = a + k
+  !$omp end single copyprivate(k)
+  !$omp single
+  b = b - a
+  !ERROR: COPYPRIVATE variable 'b' is not PRIVATE or THREADPRIVATE in outer context
+  !$omp end single copyprivate(b)
+  !$omp end parallel
+
+  !$omp parallel sections private(k)
+  !$omp section
+  !$omp parallel
+  !$omp single
+  a = a * b + k
+  !ERROR: COPYPRIVATE variable 'a' is not PRIVATE or THREADPRIVATE in outer context
+  !$omp end single copyprivate(a)
+  !$omp end parallel
+  !$omp end parallel sections
+
+  print *, a, b
+
+end program omp_copyprivate

flang/test/Semantics/omp-firstprivate01.f90

+! RUN: %S/test_errors.sh %s %t %f18 -fopenmp
+! OpenMP Version 4.5
+! 2.15.3.4 firstprivate Clause
+! Variables that appear in a firstprivate clause on a distribute or
+! worksharing constructs must not appear in the private or 
+! reduction clause in a teams or parallel constructs in the outer context
+
+program omp_firstprivate
+
+  integer :: i, a(10), b(10), c(10)
+  
+  a = 10
+  b = 20
+
+  !$omp target
+  !$omp teams private(a, b)
+  !ERROR: FIRSTPRIVATE variable 'a' is PRIVATE in outer context
+  !$omp distribute firstprivate(a)
+  do i = 1, 10
+    a(i) = a(i) + b(i) - i
+  end do
+  !$omp end distribute
+  !$omp end teams
+  !$omp teams reduction(+:a)
+  !ERROR: FIRSTPRIVATE variable 'a' is PRIVATE in outer context
+  !$omp distribute firstprivate(a)
+  do i = 1, 10
+   b(i) = b(i) + a(i) + i
+  end do
+  !$omp end distribute
+  !$omp end teams
+  !$omp end target
+
+  print *, a, b
+
+  !$omp parallel private(a,b)
+  !ERROR: FIRSTPRIVATE variable 'b' is PRIVATE in outer context
+  !$omp do firstprivate(b)
+  do i = 1, 10
+    c(i) = a(i) + b(i) + i
+  end do
+  !$omp end do
+  !$omp end parallel
+
+  !$omp parallel reduction(-:a)
+  !ERROR: FIRSTPRIVATE variable 'a' is PRIVATE in outer context
+  !$omp do firstprivate(a,b)
+  do i = 1, 10
+    c(i) =  c(i) - a(i) * b(i) * i
+  end do
+  !$omp end do
+  !$omp end parallel
+
+  !$omp parallel reduction(+:a)
+  !ERROR: FIRSTPRIVATE variable 'a' is PRIVATE in outer context
+  !$omp sections firstprivate(a, b)
+  !$omp section
+  c = c * a + b
+  !$omp end sections
+  !$omp end parallel
+
+  !$omp parallel reduction(-:a)
+  !ERROR: FIRSTPRIVATE variable 'a' is PRIVATE in outer context
+  !$omp task firstprivate(a,b)
+  c =  c - a * b
+  !$omp end task
+  !$omp end parallel
+  
+  !$omp parallel reduction(+:b)
+  !ERROR: FIRSTPRIVATE variable 'b' is PRIVATE in outer context
+  !$omp taskloop firstprivate(b)
+  do i = 1, 10
+    c(i) = a(i) + b(i) + i
+    a = a+i
+    b = b-i
+  end do
+  !$omp end taskloop
+  !$omp end parallel
+
+  print *, c
+
+end program omp_firstprivate

flang/test/Semantics/omp-lastprivate01.f90

+! RUN: %S/test_errors.sh %s %t %f18 -fopenmp
+! OpenMP Version 4.5
+! 2.15.3.5 lastprivate Clause
+! A variable that appears in a lastprivate clause must be definable.
+
+module protected_var
+  integer, protected :: p
+end module protected_var
+
+program omp_lastprivate
+  use protected_var
+  integer :: i, a(10), b(10), c(10)
+  integer, parameter :: k = 10
+
+  a = 10
+  b = 20
+
+  !ERROR: Variable 'k' on the LASTPRIVATE clause is not definable
+  !$omp parallel do lastprivate(k)
+  do i = 1, 10
+    c(i) = a(i) + b(i) + k
+  end do
+  !$omp end parallel do
+
+  !ERROR: Variable 'p' on the LASTPRIVATE clause is not definable
+  !$omp parallel do lastprivate(p)
+  do i = 1, 10
+    c(i) = a(i) + b(i) + k
+  end do
+  !$omp end parallel do
+ 
+  call omp_lastprivate_sb(i)
+
+  print *, c
+
+end program omp_lastprivate
+
+subroutine omp_lastprivate_sb(m)
+  integer :: i, a(10), b(10), c(10)
+  integer, intent(in) :: m
+
+  a = 10
+  b = 20
+
+  !ERROR: Variable 'm' on the LASTPRIVATE clause is not definable
+  !$omp parallel do lastprivate(m)
+  do i = 1, 10
+    c(i) = a(i) + b(i) + m
+  end do
+  !$omp end parallel do
+
+  print *, c
+
+end subroutine omp_lastprivate_sb

flang/test/Semantics/omp-lastprivate02.f90

+! RUN: %S/test_errors.sh %s %t %f18 -fopenmp
+! OpenMP Version 4.5
+! 2.15.3.5 lastprivate Clause
+! A list item that is private within a parallel region, or that appears in
+! reduction clause of a parallel construct, must not appear in a
+! lastprivate clause on a worksharing construct if any of the corresponding
+! worksharing regions ever binds to any of the corresponding parallel regions.
+
+program omp_lastprivate
+
+  integer :: a(10), b(10), c(10)
+
+  a = 10
+  b = 20
+
+  !$omp parallel reduction(+:a)
+  !ERROR: LASTPRIVATE variable 'a' is PRIVATE in outer context
+  !$omp sections lastprivate(a, b)
+  !$omp section
+  c = a + b
+  !$omp end sections
+  !$omp end parallel
+
+  !$omp parallel private(a,b)
+  !ERROR: LASTPRIVATE variable 'a' is PRIVATE in outer context
+  !ERROR: LASTPRIVATE variable 'b' is PRIVATE in outer context
+  !$omp do lastprivate(a,b)
+  do i = 1, 10
+    c(i) = a(i) + b(i) + i
+  end do
+  !$omp end do
+  !$omp end parallel
+
+  print *, c
+
+end program omp_lastprivate