diff mbox series

OpenMP (C only): omp allocate - handle stack vars, improve diagnostic

Message ID f8363377-75b2-ad5d-5388-54c585dfaf39@codesourcery.com
State New
Headers show
Series OpenMP (C only): omp allocate - handle stack vars, improve diagnostic | expand

Commit Message

Tobias Burnus Aug. 29, 2023, 4:12 p.m. UTC 
This adds support for
   #pragma omp allocate(var-list) [allocator(..) align(..)]

While the spec permits stack and static variables, this patch only
adds support for stack variables - keeping the 'sorry' for static
variables.
It is also only C as I wanted to get this out before updating C++
parsing. However, if disable the 'sorry' + add the attribute, it
will also work for C++.

For Fortran, there is no expression associated with the declaration
such that it will not work with the gimplify.cc patch (it cannot
find the place to add it); however, I have a mostly working version
for FE generated support for stack _and_ static variables.

The main RFC question for this patch is whether to generate the
GOMP_alloc/free calls also for !TREE_USED() or not. This patch avoids
this to aid removal of such stack variables, but one could also argue
otherwise.

Thoughts, comments, remarks?

Tobias
-----------------
Siemens Electronic Design Automation GmbH; Anschrift: Arnulfstraße 201, 80634 München; Gesellschaft mit beschränkter Haftung; Geschäftsführer: Thomas Heurung, Frank Thürauf; Sitz der Gesellschaft: München; Registergericht München, HRB 106955

Comments

Jakub Jelinek Aug. 29, 2023, 4:28 p.m. UTC | #1 
On Tue, Aug 29, 2023 at 06:12:58PM +0200, Tobias Burnus wrote:
> This adds support for
>   #pragma omp allocate(var-list) [allocator(..) align(..)]
> 
> While the spec permits stack and static variables, this patch only
> adds support for stack variables - keeping the 'sorry' for static
> variables.
> It is also only C as I wanted to get this out before updating C++
> parsing. However, if disable the 'sorry' + add the attribute, it
> will also work for C++.
> 
> For Fortran, there is no expression associated with the declaration
> such that it will not work with the gimplify.cc patch (it cannot
> find the place to add it); however, I have a mostly working version
> for FE generated support for stack _and_ static variables.
> 
> The main RFC question for this patch is whether to generate the
> GOMP_alloc/free calls also for !TREE_USED() or not. This patch avoids
> this to aid removal of such stack variables, but one could also argue
> otherwise.
> 
> Thoughts, comments, remarks?

Don't have time to look through it in detail right now, so just
2 quick comments.

One thing is that for C++ one needs to be careful about vars optimized
by NRV by the FE.  Dunno if we can simply disregard vars mentioned
in allocate directive from FE NRV, or whether we instead should ignore
the allocate directive, or sorry, etc.  Because NRV optimized vars
are turned into RESULT_DECL and live usually in the caller's stack frame
rather than current function, so they can't be heap allocated...

And, just from the gimplify_bind_expr function name, I'm guessing you are
adding the allocations at the start of surrounding BIND_EXPR, is that where
we generally want to allocate them?
Other option is on DECL_EXPR, e.g. for C++ right before they are
constructed, or for C initialized if they have initializer.  Though, that
can have a drawback, one can e.g. in C (and I think in C++ too as long as
the vars don't need any kind of construction) jump across that
initialization and one would then bypass this allocation.  Though, what
happens with say
void
foo (int i)
{
  switch (i)
    {
      int j;
      #pragma omp allocate (j) ...
    case 42:
      j = 5;
      use (&j);
      break;
    default:
      j = 24;
      use (&j);
      break;
    }
}
Where will j be allocated and where will j be deallocated with your patch?
If it is try/finally, then perhaps break; as jump out of the BIND_EXPR will
destruct it fine, but I have no idea how it would work when jumping into the
switch (or goto into middle of some compound statement etc.).

	Jakub
Tobias Burnus Aug. 29, 2023, 4:56 p.m. UTC | #2 
On 29.08.23 18:28, Jakub Jelinek wrote:
> One thing is that for C++ one needs to be careful about vars optimized
> by NRV by the FE.
Thanks for the warning.
> And, just from the gimplify_bind_expr function name,

(I forgot that one change in there which makes it look larger is that
I moved the stack handling below the variable cleaning as it makes sense
to GOMP_free before instead of after wiping the stack via __builtin_stack_restore.)

> I'm guessing you are adding the allocations at the start of surrounding BIND_EXPR, is that where
> we generally want to allocate them?

I actually started with this in a FE implementation, but that
does not work in general. I need to put the GOMP_alloc right
before the associated DECL_EXPR, otherwise, it will break for
code like:

   omp_allocator_handle_t my_allocator = omp_low_lat_mem_alloc;
   int n = 5;
   pragma omp allocate(n) allocator(my_allocator)
   int B[n]

where my_allocator strictly needs to be (initialized) before
GOMP_alloc and the 'n = 5' - but that needs in turn to happen
before the 'int B[n]' to make sure 'n' is allocated and has
the value 5.

> Other option is on DECL_EXPR, e.g. for C++ right before they are
> constructed, or for C initialized if they have initializer.  Though, that
> can have a drawback, one can e.g. in C (and I think in C++ too as long as
> the vars don't need any kind of construction) jump across that
> initialization and one would then bypass this allocation.

For your example, I get (the full gimple dump is below):

foo.c:7:11: warning: statement will never be executed [-Wswitch-unreachable]
     7 |       int j;

I wonder whether we should just error out as GCC does when using a VLA
at that place:

   error: switch jumps into scope of identifier with variably modified type

(with a similar but modified msg wording.) In a sense, the 'allocate' is very
similar to VLA such that we should be able to piggyback on the VLA diagnostic.

Tobias

PS: The gimple dump for your program:

void foo (int i)
{
   switch (i) <default: <D.2896>, case 42: <D.2895>>
   {
     int j [value-expr: *D.2899];

     try
       {
         D.2899 = __builtin_GOMP_alloc (0, 4, 0B);
         <D.2895>:
         *D.2899 = 5;
         use (D.2899);
         goto <D.2897>;
         <D.2896>:
         *D.2899 = 24;
         use (D.2899);
         goto <D.2897>;
       }
     finally
       {
         __builtin_GOMP_free (D.2899, 0B);
       }
   }
   <D.2897>:
}

-----------------
Siemens Electronic Design Automation GmbH; Anschrift: Arnulfstraße 201, 80634 München; Gesellschaft mit beschränkter Haftung; Geschäftsführer: Thomas Heurung, Frank Thürauf; Sitz der Gesellschaft: München; Registergericht München, HRB 106955
Jakub Jelinek Aug. 29, 2023, 5:14 p.m. UTC | #3 
On Tue, Aug 29, 2023 at 06:56:40PM +0200, Tobias Burnus wrote:
> On 29.08.23 18:28, Jakub Jelinek wrote:
> > One thing is that for C++ one needs to be careful about vars optimized
> > by NRV by the FE.
> Thanks for the warning.
> > And, just from the gimplify_bind_expr function name,
> 
> (I forgot that one change in there which makes it look larger is that
> I moved the stack handling below the variable cleaning as it makes sense
> to GOMP_free before instead of after wiping the stack via __builtin_stack_restore.)
> 
> > I'm guessing you are adding the allocations at the start of surrounding BIND_EXPR, is that where
> > we generally want to allocate them?
> 
> I actually started with this in a FE implementation, but that
> does not work in general. I need to put the GOMP_alloc right
> before the associated DECL_EXPR, otherwise, it will break for
> code like:
> 
>   omp_allocator_handle_t my_allocator = omp_low_lat_mem_alloc;
>   int n = 5;
>   pragma omp allocate(n) allocator(my_allocator)

What about
  int n = 5;
  omp_allocator_handle_t my_allocator = omp_low_lat_mem_alloc;
  #pragma omp allocate(n) allocator(my_allocator)
?  What we do in that case?  Is that invalid because my_allocator
isn't in scope when n is defined?

> For your example, I get (the full gimple dump is below):
> 
> foo.c:7:11: warning: statement will never be executed [-Wswitch-unreachable]
>     7 |       int j;
> 
> I wonder whether we should just error out as GCC does when using a VLA
> at that place:
> 
>   error: switch jumps into scope of identifier with variably modified type
> 
> (with a similar but modified msg wording.) In a sense, the 'allocate' is very
> similar to VLA such that we should be able to piggyback on the VLA diagnostic.

Well, in this case the warning is there just because the patch chose to put
it at the start of the BIND_EXPR rather than right before DECL_EXPR.

For switches, there is the case of the switch jumping across declaration
of an automatic var which is not initialized/constructed (I think in that
case there is normally no warning/error and happens a lot in the wild
including GCC sources) but perhaps one could treat those cases with
#pragma omp allocate as if they are actually constructed (though, I'd still
raise it at OpenMP F2F), and another case with switch which doesn't even do
that.
Consider
  switch (i)
    {
    case 42:
      bar ();
      break;
    case 51:
      int j = 5;
      use (&j);
      break;
    }
This is valid for both C and C++, one doesn't jump across any initialization
in there.  Yet if the j allocation is done at the start of the BIND_EXPR, it
will jump across that initialization.
So, to me this feels like we should treat for the crossing initialization
stuff vars mentioned in allocate directive as having a non-trivial
constructor.
If you add default: break; before } above, C++ will reject it with an error,
but C will accept it, it is up to the user to initialize the var again in C
or not use it.

And, as I said earlier, this isn't just about switch, but user gotos as
well.

	Jakub
Tobias Burnus Aug. 30, 2023, 10:47 a.m. UTC | #4 
Revised patch included - addresses part of the issues:
* gimplify.cc: Fix placement of GOMP_alloc by really checking for
   DECL_EXPR (experimented with it before but settled
   for a different pattern)
* c/ Add it to has_jump_unsafe_decl similar to VLA
   + added msg to the switch/goto error handling.
   + new c-c++-common/gomp/allocate-11.c testcase for it

But not all discussion points are solved, yet. Namely,
how to diagnose:

   omp_allocator_handle_t uninit;
   int var, var2;
   uninit = omp_low_lat_mem_alloc;
   omp_allocator_handle_t late_declared = omp_low_lat_mem_alloc;
#pragma omp allocate(var) allocator(uninit)
#pragma omp allocate(var) allocator(late_declared)

(Currently, it is only diagnosed with -Wuninitialized (or -Wextra).)

Otherwise, I think all is covered, but I might have missed something.

Regarding:

On 29.08.23 19:14, Jakub Jelinek wrote:

> What about
>    int n = 5;
>    omp_allocator_handle_t my_allocator = omp_low_lat_mem_alloc;
>    #pragma omp allocate(n) allocator(my_allocator)
> ?  What we do in that case?  Is that invalid because my_allocator
> isn't in scope when n is defined?

IMHO yes - as we can always construct cases with loops in the
execution-order dependence.

The current implementation handles it as VLA,
i.e. warning (VLA vs. allocator):

foo.c:5:3: warning: ‘m’ is used uninitialized [-Wuninitialized]
     5 |   int A[m];
       |   ^~~

foo.c:7:7: warning: ‘my_allocator’ is used uninitialized [-Wuninitialized]

     7 |   int n = 5;

As this is more surprising than VLA, it is probably worth an error
and a more explicit error. The question is how to best detect that
the allocator is either declared after the to-be-omp-allocated variable
declaration or that it is declared before but the assignment happens
only after the declaration (either before 'omp allocate' or not).


I wonder how to best check for this. During parsing, we can check for
TREE_USE and whether an initializer exists, for DECL_EXPR and for
c_check_in_current_scope, but this is not sufficient. We could use
DECL_SOURCE_LOCATION () with '<' comparisons, but I am not sure how it
plays with #line and #include - nor will it catch:
   int var;
   allocator = ...
   #pragma omp allocate(var) allocator(allocator)
where 'allocator' is TREE_USED and declared before 'var' but still wrong.
The 'is used initialized' is done deep in the middle end, but I don't think
we want to add some flag_openmp always-run special case there.

[Admittedly, we do not need to catch all issues and -Wuninitialized, enabled
by -Wextra, will also find this usage. Still, we should try to diagnose the
most common issues.]


> Well, in this case the warning is there just because the patch chose to put
> it at the start of the BIND_EXPR rather than right before DECL_EXPR.

Granted; the way the code before the DECL_EXPR was found wasn't ideal; I
now changed it back to what I previously had - where for C++, I need to
handle also a cleanup_point_expr around DECL_EXPR.

> For switches, there is the case of the switch jumping across declaration
> of an automatic var which is not initialized/constructed (I think in that
> case there is normally no warning/error and happens a lot in the wild
> including GCC sources) but perhaps one could treat those cases with
> #pragma omp allocate as if they are actually constructed (though, I'd still
> raise it at OpenMP F2F),
Can you open an OpenMP spec issue?
> and another case with switch which doesn't even do
> that.
> Consider
>    switch (i)
>      {
>      case 42:
>        bar ();
>        break;
>      case 51:
>        int j = 5;
>        use (&j);
>        break;
>      }
> This is valid for both C and C++, one doesn't jump across any initialization
> in there.  Yet if the j allocation is done at the start of the BIND_EXPR, it
> will jump across that initialization.

With the previously posted patch (and also the current one), that yields:

         <D.2896>:
         D.2900 = __builtin_GOMP_alloc (0, 4, 0B);
         *D.2900 = 5;

which looks fine to me.

Tobias
-----------------
Siemens Electronic Design Automation GmbH; Anschrift: Arnulfstraße 201, 80634 München; Gesellschaft mit beschränkter Haftung; Geschäftsführer: Thomas Heurung, Frank Thürauf; Sitz der Gesellschaft: München; Registergericht München, HRB 106955
Jakub Jelinek Aug. 30, 2023, 11:13 a.m. UTC | #5 
On Wed, Aug 30, 2023 at 12:47:42PM +0200, Tobias Burnus wrote:
> > For switches, there is the case of the switch jumping across declaration
> > of an automatic var which is not initialized/constructed (I think in that
> > case there is normally no warning/error and happens a lot in the wild
> > including GCC sources) but perhaps one could treat those cases with
> > #pragma omp allocate as if they are actually constructed (though, I'd still
> > raise it at OpenMP F2F),
> Can you open an OpenMP spec issue?

https://github.com/OpenMP/spec/issues/3676

Feel freeo to amend it.

	Jakub
Tobias Burnus Aug. 30, 2023, 12:43 p.m. UTC | #6 
Attached is an incremental patch to add diagnostic for the in-between
allocator issues, i.e.

On 30.08.23 12:47, Tobias Burnus wrote:
> omp_allocator_handle_t uninit;
>   int var, var2;
>   uninit = omp_low_lat_mem_alloc;
>   omp_allocator_handle_t late_declared = omp_low_lat_mem_alloc;
> #pragma omp allocate(var) allocator(uninit)
> #pragma omp allocate(var) allocator(late_declared)

Further comments, remarks and suggestions to this patch - or the base
patch (v2 in previous email) are highly welcome.

Tobias
-----------------
Siemens Electronic Design Automation GmbH; Anschrift: Arnulfstraße 201, 80634 München; Gesellschaft mit beschränkter Haftung; Geschäftsführer: Thomas Heurung, Frank Thürauf; Sitz der Gesellschaft: München; Registergericht München, HRB 106955
diff mbox series

Patch

OpenMP (C only): omp allocate - handle stack vars, improve diagnostic

The 'allocate' directive can be used for both stack and static variables.
While the parser in C and C++ was pre-existing, it missed several
diagnostics, which this commit adds - for now only for C.
Additionally, it stopped with a sorry after parsing.

For C only, the sorry is now restricted to static variables, the stack
variable declarations are now tagged by the 'omp allocate' attribute and
in gimplify_bind_expr the GOMP_alloc/GOMP_free allocation will now be
added.

Follow up: Add the same parser additions for C++ and update the testcases.
And add Fortran support, where also parsing support exists, where also
diagnostic updates are required.

gcc/c/ChangeLog:

	* c-parser.cc (c_parser_omp_construct): Move call to
	c_parser_omp_allocate to ...
	(c_parser_pragma): ... here.
	(c_parser_omp_allocate): Avoid ICE is allocator could not
	be parsed; set 'omp allocate' attribute for stack variables
	and only reject stack variables; add several additional
	restriction checks.

gcc/ChangeLog:

	* gimplify.cc (gimplify_bind_expr): Convert 'omp allocate'
	var-decl attribute to GOMP_alloc/GOMP_free calls.

libgomp/ChangeLog:

	* libgomp.texi (Impl.Status): Mark allocate directive as 'P'.
	(OMP_ALLOCATOR): Fix name of ICV variable.
	* testsuite/libgomp.c-c++-common/allocate-4.c: New test.

gcc/testsuite/ChangeLog:

	* c-c++-common/gomp/allocate-5.c: Fix testcase; make some
	dg-messages for 'sorry' as c++, only.
	* c-c++-common/gomp/directive-1.c: Make a 'sorry' c++ only.
	* c-c++-common/gomp/allocate-10.c: New test.
	* c-c++-common/gomp/allocate-9.c: New test.

 gcc/c/c-parser.cc                                  | 73 ++++++++++++++---
 gcc/gimplify.cc                                    | 82 +++++++++++++++----
 gcc/testsuite/c-c++-common/gomp/allocate-10.c      | 46 +++++++++++
 gcc/testsuite/c-c++-common/gomp/allocate-5.c       | 58 ++++++-------
 gcc/testsuite/c-c++-common/gomp/allocate-9.c       | 94 ++++++++++++++++++++++
 gcc/testsuite/c-c++-common/gomp/directive-1.c      |  2 +-
 libgomp/libgomp.texi                               |  7 +-
 .../testsuite/libgomp.c-c++-common/allocate-4.c    | 75 +++++++++++++++++
 8 files changed, 377 insertions(+), 60 deletions(-)

diff --git a/gcc/c/c-parser.cc b/gcc/c/c-parser.cc
index cae10ba9c80..a04fd474a28 100644
--- a/gcc/c/c-parser.cc
+++ b/gcc/c/c-parser.cc
@@ -1679,6 +1679,7 @@  static bool c_parser_omp_declare (c_parser *, enum pragma_context);
 static void c_parser_omp_requires (c_parser *);
 static bool c_parser_omp_error (c_parser *, enum pragma_context);
 static void c_parser_omp_assumption_clauses (c_parser *, bool);
+static void c_parser_omp_allocate (c_parser *);
 static void c_parser_omp_assumes (c_parser *);
 static bool c_parser_omp_ordered (c_parser *, enum pragma_context, bool *);
 static void c_parser_oacc_routine (c_parser *, enum pragma_context);
@@ -13620,6 +13621,10 @@  c_parser_pragma (c_parser *parser, enum pragma_context context, bool *if_p)
       c_parser_omp_requires (parser);
       return false;
 
+    case PRAGMA_OMP_ALLOCATE:
+      c_parser_omp_allocate (parser);
+      return false;
+
     case PRAGMA_OMP_ASSUMES:
       if (context != pragma_external)
 	{
@@ -19316,10 +19321,13 @@  c_parser_oacc_wait (location_t loc, c_parser *parser, char *p_name)
    align (constant-expression)]  */
 
 static void
-c_parser_omp_allocate (location_t loc, c_parser *parser)
+c_parser_omp_allocate (c_parser *parser)
 {
   tree alignment = NULL_TREE;
   tree allocator = NULL_TREE;
+  c_parser_consume_pragma (parser);
+  location_t loc = c_parser_peek_token (parser)->location;
+  location_t allocator_loc = UNKNOWN_LOCATION;
   tree nl = c_parser_omp_var_list_parens (parser, OMP_CLAUSE_ALLOCATE, NULL_TREE);
   do
     {
@@ -19344,7 +19352,9 @@  c_parser_omp_allocate (location_t loc, c_parser *parser)
       c_expr expr = c_parser_expr_no_commas (parser, NULL);
       expr = convert_lvalue_to_rvalue (expr_loc, expr, false, true);
       expr_loc = c_parser_peek_token (parser)->location;
-      if (p[2] == 'i' && alignment)
+      if (expr.value == error_mark_node)
+	;
+      else if (p[2] == 'i' && alignment)
 	{
 	  error_at (expr_loc, "too many %qs clauses", "align");
 	  break;
@@ -19371,6 +19381,7 @@  c_parser_omp_allocate (location_t loc, c_parser *parser)
       else
 	{
 	  allocator = c_fully_fold (expr.value, false, NULL);
+	  allocator_loc = expr_loc;
 	  tree orig_type
 	    = expr.original_type ? expr.original_type : TREE_TYPE (allocator);
 	  orig_type = TYPE_MAIN_VARIANT (orig_type);
@@ -19390,14 +19401,53 @@  c_parser_omp_allocate (location_t loc, c_parser *parser)
     } while (true);
   c_parser_skip_to_pragma_eol (parser);
 
-  if (allocator || alignment)
-    for (tree c = nl; c != NULL_TREE; c = OMP_CLAUSE_CHAIN (c))
-      {
-	OMP_CLAUSE_ALLOCATE_ALLOCATOR (c) = allocator;
-	OMP_CLAUSE_ALLOCATE_ALIGN (c) = alignment;
-      }
-
-  sorry_at (loc, "%<#pragma omp allocate%> not yet supported");
+  for (tree c = nl; c != NULL_TREE; c = OMP_CLAUSE_CHAIN (c))
+    {
+      tree var = OMP_CLAUSE_DECL (c);
+      if (TREE_CODE (var) == PARM_DECL)
+	{
+	  error_at (OMP_CLAUSE_LOCATION (nl),
+		    "function parameter %qD may not appear as list item in an "
+		    "%<allocate%> directive", var);
+	  continue;
+	}
+      if (!c_check_in_current_scope (var))
+	{
+	  error_at (OMP_CLAUSE_LOCATION (nl),
+		    "%<allocate%> directive must be in the same scope as %qD",
+		    var);
+	  inform (DECL_SOURCE_LOCATION (var), "declared here");
+	  continue;
+	}
+      if (lookup_attribute ("omp allocate", DECL_ATTRIBUTES (var)))
+	{
+	  error_at (OMP_CLAUSE_LOCATION (nl),
+		    "%qD already appeared as list item in an "
+		    "%<allocate%> directive", var);
+	  continue;
+	}
+      if (TREE_STATIC (var))
+	{
+	  if (allocator == NULL_TREE && allocator_loc == UNKNOWN_LOCATION)
+	    error_at (loc, "%<allocator%> clause required for "
+			   "static variable %qD", var);
+	  else if (allocator
+		   && (tree_int_cst_sgn (allocator) != 1
+		       || tree_to_shwi (allocator) > 8))
+	    /* 8 = largest predefined memory allocator. */
+	    error_at (allocator_loc,
+		      "%<allocator%> clause requires a predefined allocator as "
+		      "%qD is static", var);
+	  else
+	    sorry_at (OMP_CLAUSE_LOCATION (nl),
+		      "%<#pragma omp allocate%> for static variables like "
+		      "%qD not yet supported", var);
+	  continue;
+	}
+      DECL_ATTRIBUTES (var) = tree_cons (get_identifier ("omp allocate"),
+					 build_tree_list (allocator, alignment),
+					 DECL_ATTRIBUTES (var));
+    }
 }
 
 /* OpenMP 2.5:
@@ -24894,9 +24944,6 @@  c_parser_omp_construct (c_parser *parser, bool *if_p)
       strcpy (p_name, "#pragma wait");
       stmt = c_parser_oacc_wait (loc, parser, p_name);
       break;
-    case PRAGMA_OMP_ALLOCATE:
-      c_parser_omp_allocate (loc, parser);
-      return;
     case PRAGMA_OMP_ATOMIC:
       c_parser_omp_atomic (loc, parser, false);
       return;
diff --git a/gcc/gimplify.cc b/gcc/gimplify.cc
index a49b50bc857..e48eb404bcb 100644
--- a/gcc/gimplify.cc
+++ b/gcc/gimplify.cc
@@ -1364,6 +1364,44 @@  gimplify_bind_expr (tree *expr_p, gimple_seq *pre_p)
 	{
 	  struct gimplify_omp_ctx *ctx = gimplify_omp_ctxp;
 
+	  tree attr;
+	  if (flag_openmp
+	      && !is_global_var (t)
+	      && DECL_CONTEXT (t) == current_function_decl
+	      && TREE_USED (t)
+	      && (attr = lookup_attribute ("omp allocate", DECL_ATTRIBUTES (t)))
+		 != NULL_TREE)
+	    {
+	      tree v = create_tmp_var_raw (build_pointer_type (TREE_TYPE (t)));
+	      DECL_IGNORED_P (v) = 0;
+	      tree tmp = build_fold_indirect_ref (v);
+	      TREE_THIS_NOTRAP (tmp) = 1;
+	      SET_DECL_VALUE_EXPR (t, tmp);
+	      DECL_HAS_VALUE_EXPR_P (t) = 1;
+	      attr = TREE_VALUE (attr);
+	      tree sz = TYPE_SIZE_UNIT (TREE_TYPE (t));
+	      tree alloc = (TREE_PURPOSE (attr)
+			    ? TREE_PURPOSE (attr)
+			    : build_zero_cst (ptr_type_node));
+	      tree align = (TREE_VALUE (attr)
+			    ? TREE_VALUE (attr) : build_zero_cst (size_type_node));
+	      tmp = builtin_decl_explicit (BUILT_IN_GOMP_ALLOC);
+	      tmp = build_call_expr_loc (DECL_SOURCE_LOCATION (t), tmp, 3, align, sz, alloc);
+	      tmp = fold_build2_loc (DECL_SOURCE_LOCATION (t), MODIFY_EXPR, TREE_TYPE (v), v,
+				     fold_convert (TREE_TYPE (v), tmp));
+	      gcc_assert (BIND_EXPR_BODY (bind_expr) != NULL_TREE
+			  && TREE_CODE (BIND_EXPR_BODY (bind_expr)) == STATEMENT_LIST);
+	      tree_stmt_iterator e = tsi_start (BIND_EXPR_BODY (bind_expr));
+	      while (!tsi_end_p (e))
+		{
+		  if (EXPR_LOCATION (*e) == DECL_SOURCE_LOCATION (t))
+		    break;
+		  ++e;
+		}
+	      gcc_assert (!tsi_end_p (e));
+	      tsi_link_before (&e, tmp, TSI_SAME_STMT);
+	    }
+
 	  /* Mark variable as local.  */
 	  if (ctx && ctx->region_type != ORT_NONE && !DECL_EXTERNAL (t))
 	    {
@@ -1446,22 +1484,6 @@  gimplify_bind_expr (tree *expr_p, gimple_seq *pre_p)
   cleanup = NULL;
   stack_save = NULL;
 
-  /* If the code both contains VLAs and calls alloca, then we cannot reclaim
-     the stack space allocated to the VLAs.  */
-  if (gimplify_ctxp->save_stack && !gimplify_ctxp->keep_stack)
-    {
-      gcall *stack_restore;
-
-      /* Save stack on entry and restore it on exit.  Add a try_finally
-	 block to achieve this.  */
-      build_stack_save_restore (&stack_save, &stack_restore);
-
-      gimple_set_location (stack_save, start_locus);
-      gimple_set_location (stack_restore, end_locus);
-
-      gimplify_seq_add_stmt (&cleanup, stack_restore);
-    }
-
   /* Add clobbers for all variables that go out of scope.  */
   for (t = BIND_EXPR_VARS (bind_expr); t ; t = DECL_CHAIN (t))
     {
@@ -1469,6 +1491,17 @@  gimplify_bind_expr (tree *expr_p, gimple_seq *pre_p)
 	  && !is_global_var (t)
 	  && DECL_CONTEXT (t) == current_function_decl)
 	{
+	  if (flag_openmp
+	      && DECL_HAS_VALUE_EXPR_P (t)
+	      && TREE_USED (t)
+	      && lookup_attribute ("omp allocate", DECL_ATTRIBUTES (t)))
+	    {
+	      tree tmp = builtin_decl_explicit (BUILT_IN_GOMP_FREE);
+	      tmp = build_call_expr_loc (end_locus, tmp, 2,
+					 TREE_OPERAND (DECL_VALUE_EXPR (t), 0),
+					 build_zero_cst (ptr_type_node));
+	      gimplify_and_add (tmp, &cleanup);
+	    }
 	  if (!DECL_HARD_REGISTER (t)
 	      && !TREE_THIS_VOLATILE (t)
 	      && !DECL_HAS_VALUE_EXPR_P (t)
@@ -1525,6 +1558,23 @@  gimplify_bind_expr (tree *expr_p, gimple_seq *pre_p)
 	gimplify_ctxp->live_switch_vars->remove (t);
     }
 
+  /* If the code both contains VLAs and calls alloca, then we cannot reclaim
+     the stack space allocated to the VLAs.  */
+  if (gimplify_ctxp->save_stack && !gimplify_ctxp->keep_stack)
+    {
+      gcall *stack_restore;
+
+      /* Save stack on entry and restore it on exit.  Add a try_finally
+	 block to achieve this.  */
+      build_stack_save_restore (&stack_save, &stack_restore);
+
+      gimple_set_location (stack_save, start_locus);
+      gimple_set_location (stack_restore, end_locus);
+
+      gimplify_seq_add_stmt (&cleanup, stack_restore);
+    }
+
+
   if (ret_clauses)
     {
       gomp_target *stmt;
diff --git a/gcc/testsuite/c-c++-common/gomp/allocate-10.c b/gcc/testsuite/c-c++-common/gomp/allocate-10.c
new file mode 100644
index 00000000000..6fe16b1eb03
--- /dev/null
+++ b/gcc/testsuite/c-c++-common/gomp/allocate-10.c
@@ -0,0 +1,46 @@ 
+/* TODO: enable for C++ once implemented. */
+/* { dg-do compile { target c } } */
+/* { dg-additional-options "-Wall -fdump-tree-gimple" } */
+
+typedef enum omp_allocator_handle_t
+{
+  omp_default_mem_alloc = 1,
+  __omp_allocator_handle_t_max__ = __UINTPTR_MAX__
+} omp_allocator_handle_t;
+
+void
+f()
+{
+  int n;
+  int A[n]; /* { dg-warning "'n' is used uninitialized" } */
+  /* { dg-warning "unused variable 'A'" "" { target *-*-* } .-1 } */
+}
+
+void
+h1()
+{
+  omp_allocator_handle_t my_handle;
+  int B1[3]; /* { dg-warning "'my_handle' is used uninitialized" } */
+  /* { dg-warning "variable 'B1' set but not used" "" { target *-*-* } .-1 } */
+  #pragma omp allocate(B1) allocator(my_handle)
+  B1[0] = 5;
+  /* { dg-final { scan-tree-dump-times "__builtin_GOMP_alloc" 1 "gimple" } } */
+  /* { dg-final { scan-tree-dump-times "D.\[0-9\]+ = __builtin_GOMP_alloc \\(0, 12, my_handle\\);" 1 "gimple" } } */
+  /* { dg-final { scan-tree-dump-times "__builtin_GOMP_free \\(D.\[0-9\]+, 0B\\);" 1 "gimple" } } */
+}
+
+void
+h2()
+{
+  omp_allocator_handle_t my_handle;
+  int B2[3];  /* { dg-warning "unused variable 'B2'" } */
+  #pragma omp allocate(B2) allocator(my_handle) /* No warning as 'B2' is unused */
+}
+
+void
+h3()
+{
+  omp_allocator_handle_t my_handle;
+  int B3[3] = {1,2,3};  /* { dg-warning "unused variable 'B3'" } */
+  #pragma omp allocate(B3) allocator(my_handle) /* No warning as 'B3' is unused */
+}
diff --git a/gcc/testsuite/c-c++-common/gomp/allocate-5.c b/gcc/testsuite/c-c++-common/gomp/allocate-5.c
index 8a9181205f7..de1efc6832d 100644
--- a/gcc/testsuite/c-c++-common/gomp/allocate-5.c
+++ b/gcc/testsuite/c-c++-common/gomp/allocate-5.c
@@ -19,59 +19,63 @@  void
 foo ()
 {
   int a, b;
+  static int c;
   omp_allocator_handle_t my_allocator;
-#pragma omp allocate (a)  /* { dg-message "sorry, unimplemented: '#pragma omp allocate' not yet supported" } */
-#pragma omp allocate (b) allocator(my_allocator)  /* { dg-message "sorry, unimplemented: '#pragma omp allocate' not yet supported" } */
+#pragma omp allocate (a)  /* { dg-message "sorry, unimplemented: '#pragma omp allocate' not yet supported" "" { target c++ } } */
+#pragma omp allocate (b) allocator(my_allocator)  /* { dg-message "sorry, unimplemented: '#pragma omp allocate' not yet supported" "" { target c++ } } */
+#pragma omp allocate(c) align(32)
+  /* { dg-message "'allocator' clause required for static variable 'c'" "" { target c } .-1 } */
+  /* { dg-message "sorry, unimplemented: '#pragma omp allocate' not yet supported" "" { target c++ } .-2 } */
 }
 
 void
 bar ()
 {
-  int a, b;
+  int a, a2, b;
   omp_allocator_handle_t my_allocator;
 #pragma omp allocate  /* { dg-error "expected '\\(' before end of line" } */
-  /* { dg-message "sorry, unimplemented: '#pragma omp allocate' not yet supported" "" { target *-*-* } .-1 } */
+  /* { dg-message "sorry, unimplemented: '#pragma omp allocate' not yet supported" "" { target c++ } .-1 } */
 #pragma omp allocate allocator(my_allocator)  /* { dg-error "expected '\\(' before 'allocator'" } */
-  /* { dg-message "sorry, unimplemented: '#pragma omp allocate' not yet supported" "" { target *-*-* } .-1 } */
+  /* { dg-message "sorry, unimplemented: '#pragma omp allocate' not yet supported" "" { target c++ } .-1 } */
 #pragma omp allocate(a) foo(my_allocator) /* { dg-error "expected 'allocator'" } */
   /* { dg-error "expected end of line before '\\(' token" "" { target *-*-* } .-1 } */
-  /* { dg-message "sorry, unimplemented: '#pragma omp allocate' not yet supported" "" { target *-*-* } .-2 } */
-#pragma omp allocate(a) allocator(b)  /* { dg-error "'allocator' clause allocator expression has type 'int' rather than 'omp_allocator_handle_t'" "todo: cp/semantics.c" { xfail c++ } } */
-  /* { dg-message "sorry, unimplemented: '#pragma omp allocate' not yet supported" "" { target *-*-* } .-1 } */
+  /* { dg-message "sorry, unimplemented: '#pragma omp allocate' not yet supported" "" { target c++ } .-2 } */
+#pragma omp allocate(a2) allocator(b)  /* { dg-error "'allocator' clause allocator expression has type 'int' rather than 'omp_allocator_handle_t'" "todo: cp/semantics.c" { xfail c++ } } */
+  /* { dg-message "sorry, unimplemented: '#pragma omp allocate' not yet supported" "" { target c++ } .-1 } */
 }
 
 
 void
 align_test ()
 {
-  int i;
-  #pragma omp allocate(i) allocator(omp_default_mem_alloc), align(32)
-  /* { dg-message "sorry, unimplemented: '#pragma omp allocate' not yet supported" "" { target *-*-* } .-1 } */
-  #pragma omp allocate(i) align ( 32 ),allocator(omp_default_mem_alloc)
-  /* { dg-message "sorry, unimplemented: '#pragma omp allocate' not yet supported" "" { target *-*-* } .-1 } */
-  #pragma omp allocate(i),allocator(omp_default_mem_alloc) align(32)
-  /* { dg-message "sorry, unimplemented: '#pragma omp allocate' not yet supported" "" { target *-*-* } .-1 } */
-  #pragma omp allocate(i) align ( 32 ) allocator(omp_default_mem_alloc)
-  /* { dg-message "sorry, unimplemented: '#pragma omp allocate' not yet supported" "" { target *-*-* } .-1 } */
+  int i1,i2,i3,i4,i5,i6;
+  #pragma omp allocate(i1) allocator(omp_default_mem_alloc), align(32)
+  /* { dg-message "sorry, unimplemented: '#pragma omp allocate' not yet supported" "" { target c++ } .-1 } */
+  #pragma omp allocate(i2) align ( 32 ),allocator(omp_default_mem_alloc)
+  /* { dg-message "sorry, unimplemented: '#pragma omp allocate' not yet supported" "" { target c++ } .-1 } */
+  #pragma omp allocate(i3),allocator(omp_default_mem_alloc) align(32)
+  /* { dg-message "sorry, unimplemented: '#pragma omp allocate' not yet supported" "" { target c++ } .-1 } */
+  #pragma omp allocate(i4) align ( 32 ) allocator(omp_default_mem_alloc)
+  /* { dg-message "sorry, unimplemented: '#pragma omp allocate' not yet supported" "" { target c++ } .-1 } */
 
-  #pragma omp allocate(i) allocator ( omp_high_bw_mem_alloc ), align ( 32 ) allocator(omp_default_mem_alloc)
+  #pragma omp allocate(i5) allocator ( omp_high_bw_mem_alloc ), align ( 32 ) allocator(omp_default_mem_alloc)
   /* { dg-error "too many 'allocator' clauses" "" { target *-*-* } .-1 } */
   /* { dg-error "expected end of line before '\\)' token" "" { target *-*-* } .-2 } */
-  /* { dg-message "sorry, unimplemented: '#pragma omp allocate' not yet supported" "" { target *-*-* } .-3 } */
-  #pragma omp allocate(i) align ( 32 ), align(32) allocator(omp_default_mem_alloc)
+  /* { dg-message "sorry, unimplemented: '#pragma omp allocate' not yet supported" "" { target c++ } .-3 } */
+  #pragma omp allocate(i6) align ( 32 ), align(32) allocator(omp_default_mem_alloc)
   /* { dg-error "too many 'align' clauses" "" { target *-*-* } .-1 } */
   /* { dg-error "expected end of line before '\\)' token" "" { target *-*-* } .-2 } */
-  /* { dg-message "sorry, unimplemented: '#pragma omp allocate' not yet supported" "" { target *-*-* } .-3 } */
+  /* { dg-message "sorry, unimplemented: '#pragma omp allocate' not yet supported" "" { target c++ } .-3 } */
 }
 
 void
 align_test2 ()
 {
-  int i;
+  int i, i2,i3;
   #pragma omp allocate(i) align (32.0)  /* { dg-error "'align' clause argument needs to be positive constant power of two integer expression" } */
-  /* { dg-message "sorry, unimplemented: '#pragma omp allocate' not yet supported" "" { target *-*-* } .-1 } */
-  #pragma omp allocate(i) align ( 31 )  /* { dg-error "'align' clause argument needs to be positive constant power of two integer expression" } */
-  /* { dg-message "sorry, unimplemented: '#pragma omp allocate' not yet supported" "" { target *-*-* } .-1 } */
-  #pragma omp allocate(i) align ( -32 )  /* { dg-error "'align' clause argument needs to be positive constant power of two integer expression" } */
-  /* { dg-message "sorry, unimplemented: '#pragma omp allocate' not yet supported" "" { target *-*-* } .-1 } */
+  /* { dg-message "sorry, unimplemented: '#pragma omp allocate' not yet supported" "" { target c++ } .-1 } */
+  #pragma omp allocate(i2) align ( 31 )  /* { dg-error "'align' clause argument needs to be positive constant power of two integer expression" } */
+  /* { dg-message "sorry, unimplemented: '#pragma omp allocate' not yet supported" "" { target c++ } .-1 } */
+  #pragma omp allocate(i3) align ( -32 )  /* { dg-error "'align' clause argument needs to be positive constant power of two integer expression" } */
+  /* { dg-message "sorry, unimplemented: '#pragma omp allocate' not yet supported" "" { target c++ } .-1 } */
 }
diff --git a/gcc/testsuite/c-c++-common/gomp/allocate-9.c b/gcc/testsuite/c-c++-common/gomp/allocate-9.c
new file mode 100644
index 00000000000..e782b48bdf4
--- /dev/null
+++ b/gcc/testsuite/c-c++-common/gomp/allocate-9.c
@@ -0,0 +1,94 @@ 
+typedef enum omp_allocator_handle_t
+{
+  omp_null_allocator = 0,
+  omp_default_mem_alloc = 1,
+  omp_large_cap_mem_alloc = 2,
+  omp_const_mem_alloc = 3,
+  omp_high_bw_mem_alloc = 4,
+  omp_low_lat_mem_alloc = 5,
+  omp_cgroup_mem_alloc = 6,
+  omp_pteam_mem_alloc = 7,
+  omp_thread_mem_alloc = 8,
+  __ompx_last_mem_alloc = omp_thread_mem_alloc,
+  __omp_allocator_handle_t_max__ = __UINTPTR_MAX__
+} omp_allocator_handle_t;
+
+
+static int A[5] = {1,2,3,4,5};
+int B, C, D;
+
+/* If the following fails bacause of added predefined allocators, please update
+   - c/c-parser.c's c_parser_omp_allocate
+   - fortran/openmp.cc's is_predefined_allocator
+   - libgomp/env.c's parse_allocator
+   - libgomp/libgomp.texi (document the new values - multiple locations)
+   + ensure that the memory-spaces are also up to date. */
+
+#pragma omp allocate(A) align(32) allocator((omp_allocator_handle_t) 9) /* { dg-error "'allocator' clause requires a predefined allocator as 'A' is static" "" { xfail c++ } } */
+/* { dg-message "sorry, unimplemented: '#pragma omp allocate' not yet supported" "" { target c++ } .-1 } */
+
+
+// typo in allocator name:
+#pragma omp allocate(A) allocator(omp_low_latency_mem_alloc)
+/* { dg-error "'omp_low_latency_mem_alloc' undeclared here \\(not in a function\\); did you mean 'omp_low_lat_mem_alloc'\\?" "" { target c } .-1 } */
+/* { dg-error "'omp_low_latency_mem_alloc' was not declared in this scope; did you mean 'omp_low_lat_mem_alloc'\\?" "" { target c++ } .-2 } */
+/* { dg-error "'allocator' clause required for static variable 'A'" "" { target c } .-3 } */
+/* { dg-message "sorry, unimplemented: '#pragma omp allocate' not yet supported" "" { target c++ } .-4 } */
+
+/* align be const multiple of 2 */
+#pragma omp allocate(A) align(31) allocator(omp_default_mem_alloc) /* { dg-error "'align' clause argument needs to be positive constant power of two integer expression" } */
+/* { dg-message "sorry, unimplemented: '#pragma omp allocate' for static variables like 'A' not yet supported" "" { target c } .-1 } */
+/* { dg-message "sorry, unimplemented: '#pragma omp allocate' not yet supported" "" { target c++ } .-2 } */
+
+/* allocator missing (required as A is static) */
+#pragma omp allocate(A) align(32) /* { dg-error "'allocator' clause required for static variable 'A'" "" { xfail c++ } } */
+/* { dg-message "sorry, unimplemented: '#pragma omp allocate' not yet supported" "" { target c++ } .-1 } */
+
+/* "expression in the clause must be a constant expression that evaluates to one of the
+   predefined memory allocator values -> omp_low_lat_mem_alloc"  */
+#pragma omp allocate(B) allocator((omp_allocator_handle_t) (omp_high_bw_mem_alloc+1)) align(32) /* OK: omp_low_lat_mem_alloc */
+/* { dg-message "sorry, unimplemented: '#pragma omp allocate' for static variables like 'B' not yet supported" "" { target c } .-1 } */
+/* { dg-message "sorry, unimplemented: '#pragma omp allocate' not yet supported" "" { target c++ } .-2 } */
+
+#pragma omp allocate(C) allocator((omp_allocator_handle_t) 2) /* OK: omp_large_cap_mem_alloc */
+/* { dg-message "sorry, unimplemented: '#pragma omp allocate' for static variables like 'C' not yet supported" "" { target c } .-1 } */
+/* { dg-message "sorry, unimplemented: '#pragma omp allocate' not yet supported" "" { target c++ } .-2 } */
+
+#pragma omp allocate(A) align(32) allocator(omp_null_allocator) /* { dg-error "'allocator' clause requires a predefined allocator as 'A' is static" "" { xfail c++ } } */
+/* { dg-message "sorry, unimplemented: '#pragma omp allocate' not yet supported" "" { target c++ } .-1 } */
+
+#pragma omp allocate(C) align(32) allocator(omp_large_cap_mem_alloc)  /* { dg-error "'C' already appeared as list item in an 'allocate' directive" "" { xfail *-*-* } } */
+/* { dg-message "sorry, unimplemented: '#pragma omp allocate' for static variables like 'C' not yet supported" "" { target c } .-1 } */
+/* { dg-message "sorry, unimplemented: '#pragma omp allocate' not yet supported" "" { target c++ } .-2 } */
+
+// allocate directive in same TU
+int f()
+{
+  #pragma omp allocate(D) align(32) allocator(omp_large_cap_mem_alloc) /* { dg-error "'allocate' directive must be in the same scope as 'D'" "" { xfail c++ } } */
+/* { dg-note "declared here" "" { target c } 18 } */
+/* { dg-message "sorry, unimplemented: '#pragma omp allocate' not yet supported" "" { target c++ } .-2 } */
+  return A[0];
+}
+
+int g()
+{
+  int a2=1, b2=2;
+  #pragma omp allocate(a2)
+/* { dg-message "sorry, unimplemented: '#pragma omp allocate' not yet supported" "" { target c++ } .-1 } */
+  #pragma omp allocate(a2)  /* { dg-error "'a2' already appeared as list item in an 'allocate' directive" "" { xfail c++ } } */
+/* { dg-message "sorry, unimplemented: '#pragma omp allocate' not yet supported" "" { target c++ } .-1 } */
+  {
+    int c2=3;
+    #pragma omp allocate(c2, b2) /* { dg-error "'allocate' directive must be in the same scope as 'b2'" "" { xfail c++ } } */
+/* { dg-note "declared here" "" { target c } .-8 } */
+/* { dg-message "sorry, unimplemented: '#pragma omp allocate' not yet supported" "" { target c++ } .-2 } */
+    return c2+a2+b2;
+  }
+}
+
+int h(int q)
+{
+  #pragma omp allocate(q)  /* { dg-error "function parameter 'q' may not appear as list item in an 'allocate' directive" "" { xfail c++ } } */
+/* { dg-message "sorry, unimplemented: '#pragma omp allocate' not yet supported" "" { target c++ } .-1 } */
+  return q;
+}
diff --git a/gcc/testsuite/c-c++-common/gomp/directive-1.c b/gcc/testsuite/c-c++-common/gomp/directive-1.c
index fc441538778..21ca319b9f9 100644
--- a/gcc/testsuite/c-c++-common/gomp/directive-1.c
+++ b/gcc/testsuite/c-c++-common/gomp/directive-1.c
@@ -19,7 +19,7 @@  foo (void)
   int i, k = 0, l = 0;
   #pragma omp allocate, (i)			/* { dg-error "expected '\\\(' before ',' token" } */
 						/* { dg-error "expected end of line before ',' token" "" { target c++ } .-1 } */
-						/* { dg-message "not yet supported" "" { target *-*-* } .-2 } */
+						/* { dg-message "not yet supported" "" { target c++ } .-2 } */
   #pragma omp critical, (bar)			/* { dg-error "expected an OpenMP clause before '\\\(' token" } */
   ;
   #pragma omp flush, (k, l)			/* { dg-error "expected '\\\(' or end of line before ',' token" "" { target c } } */
diff --git a/libgomp/libgomp.texi b/libgomp/libgomp.texi
index 4aad8cc52f4..6dab873337e 100644
--- a/libgomp/libgomp.texi
+++ b/libgomp/libgomp.texi
@@ -225,7 +225,7 @@  The OpenMP 4.5 specification is fully supported.
 @item Predefined memory spaces, memory allocators, allocator traits
       @tab Y @tab See also @ref{Memory allocation}
 @item Memory management routines @tab Y @tab
-@item @code{allocate} directive @tab N @tab
+@item @code{allocate} directive @tab P @tab Only C, only stack variables
 @item @code{allocate} clause @tab P @tab Initial support
 @item @code{use_device_addr} clause on @code{target data} @tab Y @tab
 @item @code{ancestor} modifier on @code{device} clause @tab Y @tab
@@ -296,7 +296,8 @@  The OpenMP 4.5 specification is fully supported.
 @item Loop transformation constructs @tab N @tab
 @item @code{strict} modifier in the @code{grainsize} and @code{num_tasks}
       clauses of the @code{taskloop} construct @tab Y @tab
-@item @code{align} clause in @code{allocate} directive @tab N @tab
+@item @code{align} clause in @code{allocate} directive @tab P
+      @tab Only C (and only stack variables)
 @item @code{align} modifier in @code{allocate} clause @tab Y @tab
 @item @code{thread_limit} clause to @code{target} construct @tab Y @tab
 @item @code{has_device_addr} clause to @code{target} construct @tab Y @tab
@@ -2204,7 +2205,7 @@  variable is not set.
 @section @env{OMP_ALLOCATOR} -- Set the default allocator
 @cindex Environment Variable
 @table @asis
-@item @emph{ICV:} @var{available-devices-var}
+@item @emph{ICV:} @var{def-allocator-var}
 @item @emph{Scope:} data environment
 @item @emph{Description}:
 Sets the default allocator that is used when no allocator has been specified
diff --git a/libgomp/testsuite/libgomp.c-c++-common/allocate-4.c b/libgomp/testsuite/libgomp.c-c++-common/allocate-4.c
new file mode 100644
index 00000000000..ebfdc78a56b
--- /dev/null
+++ b/libgomp/testsuite/libgomp.c-c++-common/allocate-4.c
@@ -0,0 +1,75 @@ 
+/* TODO: enable for C++ once implemented. */
+/* { dg-do run { target c } } */
+/* { dg-additional-options "-fdump-tree-gimple" } */
+
+#include <omp.h>
+#include <stdint.h>
+
+/* { dg-final { scan-tree-dump-times "__builtin_GOMP_free \\(D.\[0-9\]+, 0B\\);" 5 "gimple" } } */
+
+int one ()
+{
+  int sum = 0;
+  #pragma omp allocate(sum)
+  /* { dg-final { scan-tree-dump-times "D.\[0-9\]+ = __builtin_GOMP_alloc \\(0, 4, 0B\\);" 1 "gimple" } } */
+
+  /* NOTE: Initializer cannot be omp_init_allocator - as 'A' is
+     in the same scope and the auto-omp_free comes later than
+     any omp_destroy_allocator.  */
+  omp_allocator_handle_t my_allocator = omp_low_lat_mem_alloc;
+  int n = 25;
+  int A[n];
+  #pragma omp allocate(A) align(128) allocator(my_allocator)
+  /* { dg-final { scan-tree-dump-times "D.\[0-9\]+ = __builtin_GOMP_alloc \\(128, _\[0-9\]+, my_allocator\\);" 1 "gimple" } } */
+
+  if (((intptr_t)A) % 128 != 0)
+    __builtin_abort ();
+  for (int i = 0; i < n; ++i)
+    A[i] = i;
+
+  omp_alloctrait_t traits[1] = { { omp_atk_alignment, 64 } };
+  my_allocator = omp_init_allocator(omp_low_lat_mem_space,1,traits);
+  {
+    int B[n] = { };
+    int C[5] = {1,2,3,4,5};
+    #pragma omp allocate(B,C) allocator(my_allocator)
+    /* { dg-final { scan-tree-dump-times "D.\[0-9\]+ = __builtin_GOMP_alloc \\(0, _\[0-9\]+, my_allocator\\);" 1 "gimple" } } */
+    /* { dg-final { scan-tree-dump-times "D.\[0-9\]+ = __builtin_GOMP_alloc \\(0, 20, my_allocator\\);" 1 "gimple" } } */
+
+    int D[5] = {11,22,33,44,55};
+    #pragma omp allocate(D) align(256)
+    /* { dg-final { scan-tree-dump-times "D.\[0-9\]+ = __builtin_GOMP_alloc \\(256, 20, 0B\\);" 1 "gimple" } } */
+
+    if (((intptr_t) B) % 64 != 0)
+      __builtin_abort ();
+    if (((intptr_t) C) % 64 != 0)
+      __builtin_abort ();
+    if (((intptr_t) D) % 64 != 0)
+      __builtin_abort ();
+
+    for (int i = 0; i < 5; ++i)
+      {
+	if (C[i] != i+1)
+	  __builtin_abort ();
+	if (D[i] != i+1 + 10*(i+1))
+	  __builtin_abort ();
+      }
+
+    for (int i = 0; i < n; ++i)
+      {
+	if (B[i] != 0)
+	  __builtin_abort ();
+	sum += A[i]+B[i]+C[i%5]+D[i%5];
+      }
+  }
+  omp_destroy_allocator (my_allocator);
+  return sum;
+}
+
+int
+main ()
+{
+  if (one () != 1200)
+    __builtin_abort ();
+  return 0;
+}