openmp: Add support for allocator and align modifiers on allocate clauses

As the allocate-2.c testcase shows, this change isn't 100% backwards compatible,
one could have allocate and/or align functions that return an OpenMP allocator
handle and previously it would call those functions and now would use those
names as keywords for the modifiers.  But it allows specify extra alignment
requirements for the allocations.

2021-09-22  Jakub Jelinek  <jakub@redhat.com>

gcc/
	* tree.h (OMP_CLAUSE_ALLOCATE_ALIGN): Define.
	* tree.c (omp_clause_num_ops): Change number of OMP_CLAUSE_ALLOCATE
	arguments from 2 to 3.
	* tree-pretty-print.c (dump_omp_clause): Print allocator() around
	allocate clause allocator and print align if present.
	* omp-low.c (scan_sharing_clauses): Force allocate_map entry even
	for omp_default_mem_alloc if align modifier is present.  If align
	modifier is present, use TREE_LIST to encode both allocator and
	align.
	(lower_private_allocate, lower_rec_input_clauses, create_task_copyfn):
	Handle align modifier on allocator clause if present.
gcc/c-family/
	* c-omp.c (c_omp_split_clauses): Copy over OMP_CLAUSE_ALLOCATE_ALIGN.
gcc/c/
	* c-parser.c (c_parser_omp_clause_allocate): Parse allocate clause
	modifiers.
gcc/cp/
	* parser.c (cp_parser_omp_clause_allocate): Parse allocate clause
	modifiers.
	* semantics.c (finish_omp_clauses) <OMP_CLAUSE_ALLOCATE>: Perform
	semantic analysis of OMP_CLAUSE_ALLOCATE_ALIGN.
	* pt.c (tsubst_omp_clauses) <case OMP_CLAUSE_ALLOCATE>: Handle
	also OMP_CLAUSE_ALLOCATE_ALIGN.
gcc/testsuite/
	* c-c++-common/gomp/allocate-6.c: New test.
	* c-c++-common/gomp/allocate-7.c: New test.
	* g++.dg/gomp/allocate-4.C: New test.
libgomp/
	* testsuite/libgomp.c-c++-common/allocate-2.c: New test.
	* testsuite/libgomp.c-c++-common/allocate-3.c: New test.

12 files changed, 544 insertions, 33 deletions

diff --git a/gcc/c-family/c-omp.c b/gcc/c-family/c-omp.c
index b606cf4..1f07a0a 100644
--- a/gcc/c-family/c-omp.c
+++ b/gcc/c-family/c-omp.c
@@ -2511,6 +2511,8 @@ c_omp_split_clauses (location_t loc, enum tree_code code,
 		    = OMP_CLAUSE_DECL (clauses);
 		  OMP_CLAUSE_ALLOCATE_ALLOCATOR (c)
 		    = OMP_CLAUSE_ALLOCATE_ALLOCATOR (clauses);
+		  OMP_CLAUSE_ALLOCATE_ALIGN (c)
+		    = OMP_CLAUSE_ALLOCATE_ALIGN (clauses);
 		  OMP_CLAUSE_CHAIN (c) = cclauses[s];
 		  cclauses[s] = c;
 		  has_dup_allocate = true;
diff --git a/gcc/c/c-parser.c b/gcc/c/c-parser.c
index fb1399e..fa29d2c 100644
--- a/gcc/c/c-parser.c
+++ b/gcc/c/c-parser.c
@@ -15305,7 +15305,15 @@ c_parser_omp_clause_aligned (c_parser *parser, tree list)
 
 /* OpenMP 5.0:
    allocate ( variable-list )
-   allocate ( expression : variable-list ) */
+   allocate ( expression : variable-list )
+
+   OpenMP 5.1:
+   allocate ( allocator-modifier : variable-list )
+   allocate ( allocator-modifier , allocator-modifier : variable-list )
+
+   allocator-modifier:
+   allocator ( expression )
+   align ( expression )  */
 
 static tree
 c_parser_omp_clause_allocate (c_parser *parser, tree list)
@@ -15313,6 +15321,7 @@ c_parser_omp_clause_allocate (c_parser *parser, tree list)
   location_t clause_loc = c_parser_peek_token (parser)->location;
   tree nl, c;
   tree allocator = NULL_TREE;
+  tree align = NULL_TREE;
 
   matching_parens parens;
   if (!parens.require_open (parser))
@@ -15323,17 +15332,128 @@ c_parser_omp_clause_allocate (c_parser *parser, tree list)
       || (c_parser_peek_2nd_token (parser)->type != CPP_COMMA
 	  && c_parser_peek_2nd_token (parser)->type != CPP_CLOSE_PAREN))
     {
-      location_t expr_loc = c_parser_peek_token (parser)->location;
-      c_expr expr = c_parser_expr_no_commas (parser, NULL);
-      expr = convert_lvalue_to_rvalue (expr_loc, expr, false, true);
-      allocator = expr.value;
-      allocator = c_fully_fold (allocator, false, NULL);
-      tree orig_type
-	= expr.original_type ? expr.original_type : TREE_TYPE (allocator);
-      orig_type = TYPE_MAIN_VARIANT (orig_type);
-      if (!INTEGRAL_TYPE_P (TREE_TYPE (allocator))
-	  || TREE_CODE (orig_type) != ENUMERAL_TYPE
-	  || TYPE_NAME (orig_type) != get_identifier ("omp_allocator_handle_t"))
+      bool has_modifiers = false;
+      tree orig_type = NULL_TREE;
+      if (c_parser_next_token_is (parser, CPP_NAME)
+	  && c_parser_peek_2nd_token (parser)->type == CPP_OPEN_PAREN)
+	{
+	  unsigned int n = 3;
+	  const char *p
+	    = IDENTIFIER_POINTER (c_parser_peek_token (parser)->value);
+	  if ((strcmp (p, "allocator") == 0 || strcmp (p, "align") == 0)
+	      && c_parser_check_balanced_raw_token_sequence (parser, &n)
+	      && (c_parser_peek_nth_token_raw (parser, n)->type
+		  == CPP_CLOSE_PAREN))
+	    {
+	      if (c_parser_peek_nth_token_raw (parser, n + 1)->type
+		  == CPP_COLON)
+		has_modifiers = true;
+	      else if (c_parser_peek_nth_token_raw (parser, n + 1)->type
+		       == CPP_COMMA
+		       && (c_parser_peek_nth_token_raw (parser, n + 2)->type
+			   == CPP_NAME)
+		       && (c_parser_peek_nth_token_raw (parser, n + 3)->type
+			   == CPP_OPEN_PAREN))
+		{
+		  c_token *tok = c_parser_peek_nth_token_raw (parser, n + 2);
+		  const char *q = IDENTIFIER_POINTER (tok->value);
+		  n += 4;
+		  if ((strcmp (q, "allocator") == 0
+		       || strcmp (q, "align") == 0)
+		      && c_parser_check_balanced_raw_token_sequence (parser,
+								     &n)
+		      && (c_parser_peek_nth_token_raw (parser, n)->type
+			  == CPP_CLOSE_PAREN)
+		      && (c_parser_peek_nth_token_raw (parser, n + 1)->type
+			  == CPP_COLON))
+		    has_modifiers = true;
+		}
+	    }
+	  if (has_modifiers)
+	    {
+	      c_parser_consume_token (parser);
+	      matching_parens parens2;;
+	      parens2.require_open (parser);
+	      location_t expr_loc = c_parser_peek_token (parser)->location;
+	      c_expr expr = c_parser_expr_no_commas (parser, NULL);
+	      expr = convert_lvalue_to_rvalue (expr_loc, expr, false, true);
+	      if (strcmp (p, "allocator") == 0)
+		{
+		  allocator = expr.value;
+		  allocator = c_fully_fold (allocator, false, NULL);
+		  orig_type = expr.original_type
+			      ? expr.original_type : TREE_TYPE (allocator);
+		  orig_type = TYPE_MAIN_VARIANT (orig_type);
+		}
+	      else
+		{
+		  align = expr.value;
+		  align = c_fully_fold (align, false, NULL);
+		}
+	      parens2.skip_until_found_close (parser);
+	      if (c_parser_next_token_is (parser, CPP_COMMA))
+		{
+		  c_parser_consume_token (parser);
+		  c_token *tok = c_parser_peek_token (parser);
+		  const char *q = "";
+		  if (c_parser_next_token_is (parser, CPP_NAME))
+		    q = IDENTIFIER_POINTER (tok->value);
+		  if (strcmp (q, "allocator") != 0 && strcmp (q, "align") != 0)
+		    {
+		      c_parser_error (parser, "expected %<allocator%> or "
+					      "%<align%>");
+		      parens.skip_until_found_close (parser);
+		      return list;
+		    }
+		  else if (strcmp (p, q) == 0)
+		    {
+		      error_at (tok->location, "duplicate %qs modifier", p);
+		      parens.skip_until_found_close (parser);
+		      return list;
+		    }
+		  c_parser_consume_token (parser);
+		  if (!parens2.require_open (parser))
+		    {
+		      parens.skip_until_found_close (parser);
+		      return list;
+		    }
+		  expr_loc = c_parser_peek_token (parser)->location;
+		  expr = c_parser_expr_no_commas (parser, NULL);
+		  expr = convert_lvalue_to_rvalue (expr_loc, expr, false,
+						   true);
+		  if (strcmp (q, "allocator") == 0)
+		    {
+		      allocator = expr.value;
+		      allocator = c_fully_fold (allocator, false, NULL);
+		      orig_type = expr.original_type
+				  ? expr.original_type : TREE_TYPE (allocator);
+		      orig_type = TYPE_MAIN_VARIANT (orig_type);
+		    }
+		  else
+		    {
+		      align = expr.value;
+		      align = c_fully_fold (align, false, NULL);
+		    }
+		  parens2.skip_until_found_close (parser);
+		}
+	    }
+	}
+      if (!has_modifiers)
+	{
+	  location_t expr_loc = c_parser_peek_token (parser)->location;
+	  c_expr expr = c_parser_expr_no_commas (parser, NULL);
+	  expr = convert_lvalue_to_rvalue (expr_loc, expr, false, true);
+	  allocator = expr.value;
+	  allocator = c_fully_fold (allocator, false, NULL);
+	  orig_type = expr.original_type
+		      ? expr.original_type : TREE_TYPE (allocator);
+	  orig_type = TYPE_MAIN_VARIANT (orig_type);
+	}
+      if (allocator
+	  && (!INTEGRAL_TYPE_P (TREE_TYPE (allocator))
+	      || TREE_CODE (orig_type) != ENUMERAL_TYPE
+	      || (TYPE_NAME (orig_type)
+		  != get_identifier ("omp_allocator_handle_t"))))
         {
           error_at (clause_loc, "%<allocate%> clause allocator expression "
 				"has type %qT rather than "
@@ -15341,6 +15461,16 @@ c_parser_omp_clause_allocate (c_parser *parser, tree list)
 				TREE_TYPE (allocator));
           allocator = NULL_TREE;
         }
+      if (align
+	  && (!INTEGRAL_TYPE_P (TREE_TYPE (align))
+	      || !tree_fits_uhwi_p (align)
+	      || !integer_pow2p (align)))
+	{
+	  error_at (clause_loc, "%<allocate%> clause %<align%> modifier "
+				"argument needs to be positive constant "
+				"power of two integer expression");
+	  align = NULL_TREE;
+	}
       if (!c_parser_require (parser, CPP_COLON, "expected %<:%>"))
 	{
 	  parens.skip_until_found_close (parser);
@@ -15351,9 +15481,12 @@ c_parser_omp_clause_allocate (c_parser *parser, tree list)
   nl = c_parser_omp_variable_list (parser, clause_loc,
 				   OMP_CLAUSE_ALLOCATE, list);
 
-  if (allocator)
+  if (allocator || align)
     for (c = nl; c != list; c = OMP_CLAUSE_CHAIN (c))
-      OMP_CLAUSE_ALLOCATE_ALLOCATOR (c) = allocator;
+      {
+	OMP_CLAUSE_ALLOCATE_ALLOCATOR (c) = allocator;
+	OMP_CLAUSE_ALLOCATE_ALIGN (c) = align;
+      }
 
   parens.skip_until_found_close (parser);
   return nl;
diff --git a/gcc/cp/parser.c b/gcc/cp/parser.c
index 62908da..26d925c 100644
--- a/gcc/cp/parser.c
+++ b/gcc/cp/parser.c
@@ -38337,13 +38337,21 @@ cp_parser_omp_clause_aligned (cp_parser *parser, tree list)
 
 /* OpenMP 5.0:
    allocate ( variable-list )
-   allocate ( expression : variable-list )  */
+   allocate ( expression : variable-list )
+
+   OpenMP 5.1:
+   allocate ( allocator-modifier : variable-list )
+   allocate ( allocator-modifier , allocator-modifier : variable-list )
+
+   allocator-modifier:
+   allocator ( expression )
+   align ( expression )  */
 
 static tree
 cp_parser_omp_clause_allocate (cp_parser *parser, tree list)
 {
-  tree nlist, c, allocator = NULL_TREE;
-  bool colon;
+  tree nlist, c, allocator = NULL_TREE, align = NULL_TREE;
+  bool colon, has_modifiers = false;
 
   matching_parens parens;
   if (!parens.require_open (parser))
@@ -38352,7 +38360,51 @@ cp_parser_omp_clause_allocate (cp_parser *parser, tree list)
   cp_parser_parse_tentatively (parser);
   bool saved_colon_corrects_to_scope_p = parser->colon_corrects_to_scope_p;
   parser->colon_corrects_to_scope_p = false;
-  allocator = cp_parser_assignment_expression (parser);
+  for (int mod = 0; mod < 2; mod++)
+    if (cp_lexer_next_token_is (parser->lexer, CPP_NAME)
+	&& cp_lexer_nth_token_is (parser->lexer, 2, CPP_OPEN_PAREN))
+      {
+	tree id = cp_lexer_peek_token (parser->lexer)->u.value;
+	const char *p = IDENTIFIER_POINTER (id);
+	if (strcmp (p, "allocator") != 0 && strcmp (p, "align") != 0)
+	  break;
+	cp_lexer_consume_token (parser->lexer);
+	matching_parens parens2;
+	if (!parens2.require_open (parser))
+	  break;
+	if (strcmp (p, "allocator") == 0)
+	  {
+	    if (allocator != NULL_TREE)
+	      break;
+	    allocator = cp_parser_assignment_expression (parser);
+	  }
+	else
+	  {
+	    if (align != NULL_TREE)
+	      break;
+	    align = cp_parser_assignment_expression (parser);
+	  }
+	if (!parens2.require_close (parser))
+	  break;
+	if (cp_lexer_next_token_is (parser->lexer, CPP_COLON))
+	  {
+	    has_modifiers = true;
+	    break;
+	  }
+	if (mod != 0 || cp_lexer_next_token_is_not (parser->lexer, CPP_COMMA))
+	  break;
+	cp_lexer_consume_token (parser->lexer);
+      }
+    else
+      break;
+  if (!has_modifiers)
+    {
+      cp_parser_abort_tentative_parse (parser);
+      align = NULL_TREE;
+      allocator = NULL_TREE;
+      cp_parser_parse_tentatively (parser);
+      allocator = cp_parser_assignment_expression (parser);
+    }
   parser->colon_corrects_to_scope_p = saved_colon_corrects_to_scope_p;
   if (cp_lexer_next_token_is (parser->lexer, CPP_COLON))
     {
@@ -38360,18 +38412,25 @@ cp_parser_omp_clause_allocate (cp_parser *parser, tree list)
       cp_lexer_consume_token (parser->lexer);
       if (allocator == error_mark_node)
 	allocator = NULL_TREE;
+      if (align == error_mark_node)
+	align = NULL_TREE;
     }
   else
     {
       cp_parser_abort_tentative_parse (parser);
       allocator = NULL_TREE;
+      align = NULL_TREE;
     }
 
   nlist = cp_parser_omp_var_list_no_open (parser, OMP_CLAUSE_ALLOCATE, list,
 					  &colon);
 
-  for (c = nlist; c != list; c = OMP_CLAUSE_CHAIN (c))
-    OMP_CLAUSE_ALLOCATE_ALLOCATOR (c) = allocator;
+  if (allocator || align)
+    for (c = nlist; c != list; c = OMP_CLAUSE_CHAIN (c))
+      {
+	OMP_CLAUSE_ALLOCATE_ALLOCATOR (c) = allocator;
+	OMP_CLAUSE_ALLOCATE_ALIGN (c) = align;
+      }
 
   return nlist;
 }
diff --git a/gcc/cp/pt.c b/gcc/cp/pt.c
index 048bbc4..28ba32f4 100644
--- a/gcc/cp/pt.c
+++ b/gcc/cp/pt.c
@@ -17489,6 +17489,13 @@ tsubst_omp_clauses (tree clauses, enum c_omp_region_type ort,
 	  break;
 	case OMP_CLAUSE_GANG:
 	case OMP_CLAUSE_ALIGNED:
+	  OMP_CLAUSE_DECL (nc)
+	    = tsubst_omp_clause_decl (OMP_CLAUSE_DECL (oc), args, complain,
+				      in_decl, NULL);
+	  OMP_CLAUSE_OPERAND (nc, 1)
+	    = tsubst_expr (OMP_CLAUSE_OPERAND (oc, 1), args, complain,
+			   in_decl, /*integral_constant_expression_p=*/false);
+	  break;
 	case OMP_CLAUSE_ALLOCATE:
 	  OMP_CLAUSE_DECL (nc)
 	    = tsubst_omp_clause_decl (OMP_CLAUSE_DECL (oc), args, complain,
@@ -17496,6 +17503,9 @@ tsubst_omp_clauses (tree clauses, enum c_omp_region_type ort,
 	  OMP_CLAUSE_OPERAND (nc, 1)
 	    = tsubst_expr (OMP_CLAUSE_OPERAND (oc, 1), args, complain,
 			   in_decl, /*integral_constant_expression_p=*/false);
+	  OMP_CLAUSE_OPERAND (nc, 2)
+	    = tsubst_expr (OMP_CLAUSE_OPERAND (oc, 2), args, complain,
+			   in_decl, /*integral_constant_expression_p=*/false);
 	  break;
 	case OMP_CLAUSE_LINEAR:
 	  OMP_CLAUSE_DECL (nc)
diff --git a/gcc/cp/semantics.c b/gcc/cp/semantics.c
index 35a7b9f..0d8e5fd 100644
--- a/gcc/cp/semantics.c
+++ b/gcc/cp/semantics.c
@@ -7527,7 +7527,44 @@ finish_omp_clauses (tree clauses, enum c_omp_region_type ort)
 	      bitmap_set_bit (&aligned_head, DECL_UID (t));
 	      allocate_seen = true;
 	    }
-	  tree allocator;
+	  tree allocator, align;
+	  align = OMP_CLAUSE_ALLOCATE_ALIGN (c);
+	  if (error_operand_p (align))
+	    {
+	      remove = true;
+	      break;
+	    }
+	  if (align)
+	    {
+	      if (!type_dependent_expression_p (align)
+		  && !INTEGRAL_TYPE_P (TREE_TYPE (align)))
+		{
+		  error_at (OMP_CLAUSE_LOCATION (c),
+			    "%<allocate%> clause %<align%> modifier "
+			    "argument needs to be positive constant "
+			    "power of two integer expression");
+		  remove = true;
+		}
+	      else
+		{
+		  align = mark_rvalue_use (align);
+		  if (!processing_template_decl)
+		    {
+		      align = maybe_constant_value (align);
+		      if (TREE_CODE (align) != INTEGER_CST
+			  || !tree_fits_uhwi_p (align)
+			  || !integer_pow2p (align))
+			{
+			  error_at (OMP_CLAUSE_LOCATION (c),
+				    "%<allocate%> clause %<align%> modifier "
+				    "argument needs to be positive constant "
+				    "power of two integer expression");
+			  remove = true;
+			}
+		    }
+		}
+	      OMP_CLAUSE_ALLOCATE_ALIGN (c) = align;
+	    }
 	  allocator = OMP_CLAUSE_ALLOCATE_ALLOCATOR (c);
 	  if (error_operand_p (allocator))
 	    {
@@ -7552,6 +7589,7 @@ finish_omp_clauses (tree clauses, enum c_omp_region_type ort)
 			"type %qT rather than %<omp_allocator_handle_t%>",
 			TREE_TYPE (allocator));
 	      remove = true;
+	      break;
 	    }
 	  else
 	    {
diff --git a/gcc/omp-low.c b/gcc/omp-low.c
index 27a513e..26c5c02 100644
--- a/gcc/omp-low.c
+++ b/gcc/omp-low.c
@@ -1161,14 +1161,17 @@ scan_sharing_clauses (tree clauses, omp_context *ctx)
     if (OMP_CLAUSE_CODE (c) == OMP_CLAUSE_ALLOCATE
 	&& (OMP_CLAUSE_ALLOCATE_ALLOCATOR (c) == NULL_TREE
 	    /* omp_default_mem_alloc is 1 */
-	    || !integer_onep (OMP_CLAUSE_ALLOCATE_ALLOCATOR (c))))
+	    || !integer_onep (OMP_CLAUSE_ALLOCATE_ALLOCATOR (c))
+	    || OMP_CLAUSE_ALLOCATE_ALIGN (c) != NULL_TREE))
       {
 	if (ctx->allocate_map == NULL)
 	  ctx->allocate_map = new hash_map<tree, tree>;
-	ctx->allocate_map->put (OMP_CLAUSE_DECL (c),
-				OMP_CLAUSE_ALLOCATE_ALLOCATOR (c)
-				? OMP_CLAUSE_ALLOCATE_ALLOCATOR (c)
-				: integer_zero_node);
+	tree val = integer_zero_node;
+	if (OMP_CLAUSE_ALLOCATE_ALLOCATOR (c))
+	  val = OMP_CLAUSE_ALLOCATE_ALLOCATOR (c);
+	if (OMP_CLAUSE_ALLOCATE_ALIGN (c))
+	  val = build_tree_list (val, OMP_CLAUSE_ALLOCATE_ALIGN (c));
+	ctx->allocate_map->put (OMP_CLAUSE_DECL (c), val);
       }
 
   for (c = clauses; c; c = OMP_CLAUSE_CHAIN (c))
@@ -4725,6 +4728,12 @@ lower_private_allocate (tree var, tree new_var, tree &allocator,
       return false;
     }
 
+  unsigned HOST_WIDE_INT ialign = 0;
+  if (TREE_CODE (allocator) == TREE_LIST)
+    {
+      ialign = tree_to_uhwi (TREE_VALUE (allocator));
+      allocator = TREE_PURPOSE (allocator);
+    }
   if (TREE_CODE (allocator) != INTEGER_CST)
     allocator = build_outer_var_ref (allocator, ctx);
   allocator = fold_convert (pointer_sized_int_node, allocator);
@@ -4739,21 +4748,21 @@ lower_private_allocate (tree var, tree new_var, tree &allocator,
   if (TYPE_P (new_var))
     {
       ptr_type = build_pointer_type (new_var);
-      align = build_int_cst (size_type_node, TYPE_ALIGN_UNIT (new_var));
+      ialign = MAX (ialign, TYPE_ALIGN_UNIT (new_var));
     }
   else if (is_ref)
     {
       ptr_type = build_pointer_type (TREE_TYPE (TREE_TYPE (new_var)));
-      align = build_int_cst (size_type_node,
-			     TYPE_ALIGN_UNIT (TREE_TYPE (ptr_type)));
+      ialign = MAX (ialign, TYPE_ALIGN_UNIT (TREE_TYPE (ptr_type)));
     }
   else
     {
       ptr_type = build_pointer_type (TREE_TYPE (new_var));
-      align = build_int_cst (size_type_node, DECL_ALIGN_UNIT (new_var));
+      ialign = MAX (ialign, DECL_ALIGN_UNIT (new_var));
       if (sz == NULL_TREE)
 	sz = fold_convert (size_type_node, DECL_SIZE_UNIT (new_var));
     }
+  align = build_int_cst (size_type_node, ialign);
   if (TREE_CODE (sz) != INTEGER_CST)
     {
       tree szvar = create_tmp_var (size_type_node);
@@ -5707,6 +5716,8 @@ lower_rec_input_clauses (tree clauses, gimple_seq *ilist, gimple_seq *dlist,
 		    if (tree *allocatep = ctx->allocate_map->get (var))
 		      {
 			allocator = *allocatep;
+			if (TREE_CODE (allocator) == TREE_LIST)
+			  allocator = TREE_PURPOSE (allocator);
 			if (TREE_CODE (allocator) != INTEGER_CST)
 			  allocator = build_outer_var_ref (allocator, ctx);
 			allocator = fold_convert (pointer_sized_int_node,
@@ -6025,6 +6036,8 @@ lower_rec_input_clauses (tree clauses, gimple_seq *ilist, gimple_seq *dlist,
 			if (tree *allocatep = ctx->allocate_map->get (var))
 			  {
 			    allocator = *allocatep;
+			    if (TREE_CODE (allocator) == TREE_LIST)
+			      allocator = TREE_PURPOSE (allocator);
 			    if (TREE_CODE (allocator) != INTEGER_CST)
 			      allocator = build_outer_var_ref (allocator, ctx);
 			    allocator = fold_convert (pointer_sized_int_node,
@@ -12070,6 +12083,12 @@ create_task_copyfn (gomp_task *task_stmt, omp_context *ctx)
 	      if (tree *allocatorp = ctx->allocate_map->get (decl))
 		{
 		  tree allocator = *allocatorp;
+		  HOST_WIDE_INT ialign = 0;
+		  if (TREE_CODE (allocator) == TREE_LIST)
+		    {
+		      ialign = tree_to_uhwi (TREE_VALUE (allocator));
+		      allocator = TREE_PURPOSE (allocator);
+		    }
 		  if (TREE_CODE (allocator) != INTEGER_CST)
 		    {
 		      n = splay_tree_lookup (ctx->sfield_map,
@@ -12083,7 +12102,8 @@ create_task_copyfn (gomp_task *task_stmt, omp_context *ctx)
 		  allocator = fold_convert (pointer_sized_int_node, allocator);
 		  tree a = builtin_decl_explicit (BUILT_IN_GOMP_ALLOC);
 		  tree align = build_int_cst (size_type_node,
-					      DECL_ALIGN_UNIT (decl));
+					      MAX (ialign,
+						   DECL_ALIGN_UNIT (decl)));
 		  tree sz = TYPE_SIZE_UNIT (TREE_TYPE (TREE_TYPE (dst)));
 		  tree ptr = build_call_expr_loc (loc, a, 3, align, sz,
 						  allocator);
diff --git a/gcc/testsuite/c-c++-common/gomp/allocate-6.c b/gcc/testsuite/c-c++-common/gomp/allocate-6.c
new file mode 100644
index 0000000..059fd72
--- /dev/null
+++ b/gcc/testsuite/c-c++-common/gomp/allocate-6.c
@@ -0,0 +1,84 @@
+typedef enum omp_allocator_handle_t
+#if __cplusplus >= 201103L
+: __UINTPTR_TYPE__
+#endif
+{
+  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,
+  __omp_allocator_handle_t_max__ = __UINTPTR_MAX__
+} omp_allocator_handle_t;
+
+int bar (int, int *, int);
+omp_allocator_handle_t baz (void);
+
+void
+foo (int x, int z)
+{
+  int y[16] = { 0 }, r = 0, i;
+  omp_allocator_handle_t h = baz ();
+  #pragma omp parallel allocate (align (sizeof (int)) : x) allocate (allocator (omp_default_mem_alloc) : y) \
+	      allocate (align (8), allocator ((omp_allocator_handle_t) omp_default_mem_alloc):z) firstprivate (x, y, z)
+  bar (x, y, z);
+  #pragma omp task private (x) firstprivate (z) allocate (allocator (omp_low_lat_mem_alloc) :x,z)
+  bar (0, &x, z);
+  #pragma omp taskwait
+  #pragma omp target teams distribute parallel for private (x) firstprivate (y) \
+	      allocate (allocator ((omp_allocator_handle_t)(omp_default_mem_alloc + 0)), align (16) : z) \
+	      allocate (allocator (omp_default_mem_alloc) : x, y) allocate (align (32), allocator (omp_low_lat_mem_alloc): r) \
+	      lastprivate (z) reduction(+:r)
+  for (i = 0; i < 64; i++)
+    {
+      z = bar (0, &x, 0);
+      r += bar (1, y, 0);
+    }
+  #pragma omp single private (x) allocate (allocator (h):x)
+  ;
+  #pragma omp single allocate (align (2 * sizeof (int)), allocator (*&h) : x) private (x)
+  ;
+  #pragma omp parallel shared (r, x, z)
+  #pragma omp single firstprivate (r) allocate (align (4) : x, r, z) private (x, z)
+  ;
+  #pragma omp for allocate (align (2 * 2 * 2) : x) private (x)
+  for (i = 0; i < 64; i++)
+    x = 1;
+  #pragma omp sections private (x) allocate (allocator (omp_low_lat_mem_alloc), align (8): x)
+  {
+    x = 1;
+    #pragma omp section
+    x = 2;
+    #pragma omp section
+    x = 3;
+  }
+  #pragma omp taskgroup task_reduction(+:r) allocate (allocator (omp_default_mem_alloc), align (__alignof (r)) : r)
+  #pragma omp task in_reduction(+:r) allocate (align (2 * sizeof (r)), allocator (omp_default_mem_alloc) : r)
+  r += bar (r, &r, 0);
+  #pragma omp teams private (x) firstprivate (y) allocate (allocator (h), align (8) : x, y)
+  bar (x, y, 0);
+  #pragma omp taskloop lastprivate (x) reduction (+:r) allocate (align (16), allocator (h) : x, r)
+  for (i = 0; i < 16; i++)
+    {
+      r += bar (0, &r, 0);
+      x = i;
+    }
+  #pragma omp taskgroup task_reduction(+:r) allocate (allocator (omp_default_mem_alloc), align (64) : r)
+  #pragma omp taskloop firstprivate (x) in_reduction (+:r) \
+		       allocate (allocator (omp_default_mem_alloc), align (128) : x, r)
+  for (i = 0; i < 16; i++)
+    r += bar (x, &r, 0);
+  #pragma omp taskwait
+}
+
+void
+qux (const omp_allocator_handle_t h)
+{
+  int x = 0;
+  #pragma omp parallel firstprivate (x) allocate (align (16), allocator (h): x)
+  x = 1;
+}
diff --git a/gcc/testsuite/c-c++-common/gomp/allocate-7.c b/gcc/testsuite/c-c++-common/gomp/allocate-7.c
new file mode 100644
index 0000000..d125d92
--- /dev/null
+++ b/gcc/testsuite/c-c++-common/gomp/allocate-7.c
@@ -0,0 +1,41 @@
+typedef enum omp_allocator_handle_t
+#if __cplusplus >= 201103L
+: __UINTPTR_TYPE__
+#endif
+{
+  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,
+  __omp_allocator_handle_t_max__ = __UINTPTR_MAX__
+} omp_allocator_handle_t;
+
+int bar (int *);
+omp_allocator_handle_t baz (void);
+
+void
+foo (int x, int z)
+{
+  int i;
+  #pragma omp parallel private (x) allocate (allocator (0.0) : x)	/* { dg-error "'allocate' clause allocator expression has type 'double' rather than 'omp_allocator_handle_t'" } */
+  bar (&x);
+  #pragma omp parallel private (x) allocate (allocator (0) : x)	/* { dg-error "'allocate' clause allocator expression has type 'int' rather than 'omp_allocator_handle_t'" } */
+  bar (&x);
+  #pragma omp parallel private (x) allocate (align (z) : x)	/* { dg-error "'allocate' clause 'align' modifier argument needs to be positive constant power of two integer expression" } */
+  bar (&x);
+  #pragma omp parallel private (x) allocate (align (16.0) : x)	/* { dg-error "'allocate' clause 'align' modifier argument needs to be positive constant power of two integer expression" } */
+  bar (&x);
+  #pragma omp parallel private (x) allocate (align (14) : x)	/* { dg-error "'allocate' clause 'align' modifier argument needs to be positive constant power of two integer expression" } */
+  bar (&x);
+  #pragma omp parallel private (x) allocate (align (0) : x)	/* { dg-error "'allocate' clause 'align' modifier argument needs to be positive constant power of two integer expression" } */
+  bar (&x);
+  #pragma omp parallel private (x) allocate (align (16), align (16) : x)	/* { dg-error "expected|duplicate|declared|specified" } */
+  bar (&x);									/* { dg-warning "more than once" "" { target c++ } .-1 } */
+  #pragma omp parallel private (x) allocate (allocator (omp_default_mem_alloc), allocator (omp_default_mem_alloc) : x)	/* { dg-error "expected|duplicate|declared|specified" } */
+  bar (&x);									/* { dg-warning "more than once" "" { target c++ } .-1 } */
+}
diff --git a/gcc/testsuite/g++.dg/gomp/allocate-4.C b/gcc/testsuite/g++.dg/gomp/allocate-4.C
new file mode 100644
index 0000000..dd5e5f0
--- /dev/null
+++ b/gcc/testsuite/g++.dg/gomp/allocate-4.C
@@ -0,0 +1,108 @@
+// { dg-do compile }
+// { dg-additional-options "-std=c++11" }
+
+typedef enum omp_allocator_handle_t
+#if __cplusplus >= 201103L
+: __UINTPTR_TYPE__
+#endif
+{
+  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,
+  __omp_allocator_handle_t_max__ = __UINTPTR_MAX__
+} omp_allocator_handle_t;
+
+namespace N1
+{
+  using ::omp_allocator_handle_t;
+  void
+  foo (const omp_allocator_handle_t h)
+  {
+    int x = 0;
+    #pragma omp parallel allocate (allocator (h): x) private (x)
+    x = 1;
+  }
+}
+
+namespace N2
+{
+  typedef enum omp_allocator_handle_t { my = 0 } omp_allocator_handle_t;
+  void
+  foo (omp_allocator_handle_t h)
+  {
+    int x = 0;
+    #pragma omp parallel allocate (allocator (h): x) private (x) // { dg-error "'allocate' clause allocator expression has type 'N2::omp_allocator_handle_t' rather than 'omp_allocator_handle_t'" }
+    x = 1;
+  }
+}
+
+struct S
+{
+  void foo ()
+  {
+    #pragma omp parallel allocate (allocator(omp_default_mem_alloc):s) firstprivate (s)
+    s++;
+  }
+  int s;
+};
+
+template <typename T>
+struct U
+{
+  int foo ()
+  {
+    #pragma omp parallel allocate (allocator(omp_default_mem_alloc):s) firstprivate (s)
+    s++;
+    return 1;
+  }
+  T s;
+};
+
+template <typename T, int N>
+int foo (T t)
+{
+  int x = 0;
+  #pragma omp parallel firstprivate (x) allocate (allocator(t), align (N): x)
+  x = 1;
+  return 0;
+}
+
+template <typename T>
+int bar (T t)
+{
+  int x = 0;
+  #pragma omp parallel firstprivate (x) allocate (allocator(t): x)	// { dg-error "'allocate' clause allocator expression has type 'int' rather than 'omp_allocator_handle_t'" }
+  x = 1;
+  return 0;
+}
+
+template <typename T, int N>
+int baz (T t)
+{
+  int x = 0;
+  #pragma omp parallel firstprivate (x) allocate (allocator(t), align (N): x) // { dg-error "'allocate' clause 'align' modifier argument needs to be positive constant power of two integer expression" }
+  x = 1;
+  return 0;
+}
+
+template <typename T>
+int qux ()
+{
+  int x = 0;
+  #pragma omp parallel firstprivate (x) allocate (align ((T) 16): x) // { dg-error "'allocate' clause 'align' modifier argument needs to be positive constant power of two integer expression" }
+  x = 1;
+  return 0;
+}
+
+omp_allocator_handle_t h;
+int a = foo<omp_allocator_handle_t, 16> (h);
+int b = bar (0);
+int c = U<int> ().foo ();
+int d = baz<omp_allocator_handle_t, 13> (h);
+int e = qux<long double> ();
diff --git a/gcc/tree-pretty-print.c b/gcc/tree-pretty-print.c
index 7de12f3..35e567c 100644
--- a/gcc/tree-pretty-print.c
+++ b/gcc/tree-pretty-print.c
@@ -735,10 +735,23 @@ dump_omp_clause (pretty_printer *pp, tree clause, int spc, dump_flags_t flags)
       pp_string (pp, "allocate(");
       if (OMP_CLAUSE_ALLOCATE_ALLOCATOR (clause))
 	{
+	  pp_string (pp, "allocator(");
 	  dump_generic_node (pp, OMP_CLAUSE_ALLOCATE_ALLOCATOR (clause),
 			     spc, flags, false);
-	  pp_colon (pp);
+	  pp_right_paren (pp);
+	}
+      if (OMP_CLAUSE_ALLOCATE_ALIGN (clause))
+	{
+	  if (OMP_CLAUSE_ALLOCATE_ALLOCATOR (clause))
+	    pp_comma (pp);
+	  pp_string (pp, "align(");
+	  dump_generic_node (pp, OMP_CLAUSE_ALLOCATE_ALIGN (clause),
+			     spc, flags, false);
+	  pp_right_paren (pp);
 	}
+      if (OMP_CLAUSE_ALLOCATE_ALLOCATOR (clause)
+	  || OMP_CLAUSE_ALLOCATE_ALIGN (clause))
+	pp_colon (pp);
       dump_generic_node (pp, OMP_CLAUSE_DECL (clause),
 			 spc, flags, false);
       pp_right_paren (pp);
diff --git a/gcc/tree.c b/gcc/tree.c
index 994775d..561b9cd5 100644
--- a/gcc/tree.c
+++ b/gcc/tree.c
@@ -291,7 +291,7 @@ unsigned const char omp_clause_num_ops[] =
   3, /* OMP_CLAUSE_LINEAR  */
   1, /* OMP_CLAUSE_AFFINITY  */
   2, /* OMP_CLAUSE_ALIGNED  */
-  2, /* OMP_CLAUSE_ALLOCATE  */
+  3, /* OMP_CLAUSE_ALLOCATE  */
   1, /* OMP_CLAUSE_DEPEND  */
   1, /* OMP_CLAUSE_NONTEMPORAL  */
   1, /* OMP_CLAUSE_UNIFORM  */
diff --git a/gcc/tree.h b/gcc/tree.h
index 091ad3d..7d1257b 100644
--- a/gcc/tree.h
+++ b/gcc/tree.h
@@ -1790,6 +1790,9 @@ class auto_suppress_location_wrappers
 #define OMP_CLAUSE_ALLOCATE_ALLOCATOR(NODE) \
   OMP_CLAUSE_OPERAND (OMP_CLAUSE_SUBCODE_CHECK (NODE, OMP_CLAUSE_ALLOCATE), 1)
 
+#define OMP_CLAUSE_ALLOCATE_ALIGN(NODE) \
+  OMP_CLAUSE_OPERAND (OMP_CLAUSE_SUBCODE_CHECK (NODE, OMP_CLAUSE_ALLOCATE), 2)
+
 /* True if an ALLOCATE clause was present on a combined or composite
    construct and the code for splitting the clauses has already performed
    checking if the listed variable has explicit privatization on the