Merge current set of OpenACC changes from gomp-4_0-branch.

	contrib/
	* gcc_update (files_and_dependencies): Update rules for new
	libgomp/plugin/Makefrag.am and libgomp/plugin/configfrag.ac files.
	gcc/
	* builtin-types.def (BT_FN_VOID_INT_INT_VAR)
	(BT_FN_VOID_INT_PTR_SIZE_PTR_PTR_PTR_INT_INT_VAR)
	(BT_FN_VOID_INT_OMPFN_PTR_SIZE_PTR_PTR_PTR_INT_INT_INT_INT_INT_VAR):
	New function types.
	* builtins.c: Include "gomp-constants.h".
	(expand_builtin_acc_on_device): New function.
	(expand_builtin, is_inexpensive_builtin): Handle
	BUILT_IN_ACC_ON_DEVICE.
	* builtins.def (DEF_GOACC_BUILTIN, DEF_GOACC_BUILTIN_COMPILER):
	New macros.
	* cgraph.c (cgraph_node::create): Consider flag_openacc next to
	flag_openmp.
	* config.gcc <nvptx-*> (tm_file): Add nvptx/offload.h.
	<*-intelmic-* | *-intelmicemul-*> (tm_file): Add
	i386/intelmic-offload.h.
	* gcc.c (LINK_COMMAND_SPEC, GOMP_SELF_SPECS): For -fopenacc, link
	to libgomp and its dependencies.
	* config/arc/arc.h (LINK_COMMAND_SPEC): Likewise.
	* config/darwin.h (LINK_COMMAND_SPEC_A): Likewise.
	* config/i386/mingw32.h (GOMP_SELF_SPECS): Likewise.
	* config/ia64/hpux.h (LIB_SPEC): Likewise.
	* config/pa/pa-hpux11.h (LIB_SPEC): Likewise.
	* config/pa/pa64-hpux.h (LIB_SPEC): Likewise.
	* doc/generic.texi: Update for OpenACC changes.
	* doc/gimple.texi: Likewise.
	* doc/invoke.texi: Likewise.
	* doc/sourcebuild.texi: Likewise.
	* gimple-pretty-print.c (dump_gimple_omp_for): Handle
	GF_OMP_FOR_KIND_OACC_LOOP.
	(dump_gimple_omp_target): Handle GF_OMP_TARGET_KIND_OACC_KERNELS,
	GF_OMP_TARGET_KIND_OACC_PARALLEL, GF_OMP_TARGET_KIND_OACC_DATA,
	GF_OMP_TARGET_KIND_OACC_UPDATE,
	GF_OMP_TARGET_KIND_OACC_ENTER_EXIT_DATA.
	Dump more data.
	* gimple.c: Update comments for OpenACC changes.
	* gimple.def: Likewise.
	* gimple.h: Likewise.
	(enum gf_mask): Add GF_OMP_FOR_KIND_OACC_LOOP,
	GF_OMP_TARGET_KIND_OACC_PARALLEL, GF_OMP_TARGET_KIND_OACC_KERNELS,
	GF_OMP_TARGET_KIND_OACC_DATA, GF_OMP_TARGET_KIND_OACC_UPDATE,
	GF_OMP_TARGET_KIND_OACC_ENTER_EXIT_DATA.
	(gimple_omp_for_cond, gimple_omp_for_set_cond): Sort in the
	appropriate place.
	(is_gimple_omp_oacc, is_gimple_omp_offloaded): New functions.
	* gimplify.c: Include "gomp-constants.h".
	Update comments for OpenACC changes.
	(is_gimple_stmt): Handle OACC_PARALLEL, OACC_KERNELS, OACC_DATA,
	OACC_HOST_DATA, OACC_DECLARE, OACC_UPDATE, OACC_ENTER_DATA,
	OACC_EXIT_DATA, OACC_CACHE, OACC_LOOP.
	(gimplify_scan_omp_clauses, gimplify_adjust_omp_clauses): Handle
	OMP_CLAUSE__CACHE_, OMP_CLAUSE_ASYNC, OMP_CLAUSE_WAIT,
	OMP_CLAUSE_NUM_GANGS, OMP_CLAUSE_NUM_WORKERS,
	OMP_CLAUSE_VECTOR_LENGTH, OMP_CLAUSE_GANG, OMP_CLAUSE_WORKER,
	OMP_CLAUSE_VECTOR, OMP_CLAUSE_DEVICE_RESIDENT,
	OMP_CLAUSE_USE_DEVICE, OMP_CLAUSE_INDEPENDENT, OMP_CLAUSE_AUTO,
	OMP_CLAUSE_SEQ.
	(gimplify_adjust_omp_clauses_1, gimplify_adjust_omp_clauses): Use
	GOMP_MAP_* instead of OMP_CLAUSE_MAP_*.  Use
	OMP_CLAUSE_SET_MAP_KIND.
	(gimplify_oacc_cache): New function.
	(gimplify_omp_for): Handle OACC_LOOP.
	(gimplify_omp_workshare): Handle OACC_KERNELS, OACC_PARALLEL,
	OACC_DATA.
	(gimplify_omp_target_update): Handle OACC_ENTER_DATA,
	OACC_EXIT_DATA, OACC_UPDATE.
	(gimplify_expr): Handle OACC_LOOP, OACC_CACHE, OACC_HOST_DATA,
	OACC_DECLARE, OACC_KERNELS, OACC_PARALLEL, OACC_DATA,
	OACC_ENTER_DATA, OACC_EXIT_DATA, OACC_UPDATE.
	(gimplify_body): Consider flag_openacc next to flag_openmp.
	* lto-streamer-out.c: Include "gomp-constants.h".
	* omp-builtins.def (BUILT_IN_ACC_GET_DEVICE_TYPE)
	(BUILT_IN_GOACC_DATA_START, BUILT_IN_GOACC_DATA_END)
	(BUILT_IN_GOACC_ENTER_EXIT_DATA, BUILT_IN_GOACC_PARALLEL)
	(BUILT_IN_GOACC_UPDATE, BUILT_IN_GOACC_WAIT)
	(BUILT_IN_GOACC_GET_THREAD_NUM, BUILT_IN_GOACC_GET_NUM_THREADS)
	(BUILT_IN_ACC_ON_DEVICE): New builtins.
	* omp-low.c: Include "gomp-constants.h".
	Update comments for OpenACC changes.
	(struct omp_context): Add reduction_map, gwv_below, gwv_this
	members.
	(extract_omp_for_data, use_pointer_for_field, install_var_field)
	(new_omp_context, delete_omp_context, scan_sharing_clauses)
	(create_omp_child_function, scan_omp_for, scan_omp_target)
	(check_omp_nesting_restrictions, lower_reduction_clauses)
	(build_omp_regions_1, diagnose_sb_0, make_gimple_omp_edges):
	Update for OpenACC changes.
	(scan_sharing_clauses): Handle OMP_CLAUSE_NUM_GANGS:
	OMP_CLAUSE_NUM_WORKERS: OMP_CLAUSE_VECTOR_LENGTH,
	OMP_CLAUSE_ASYNC, OMP_CLAUSE_WAIT, OMP_CLAUSE_GANG,
	OMP_CLAUSE_WORKER, OMP_CLAUSE_VECTOR, OMP_CLAUSE_DEVICE_RESIDENT,
	OMP_CLAUSE_USE_DEVICE, OMP_CLAUSE__CACHE_, OMP_CLAUSE_INDEPENDENT,
	OMP_CLAUSE_AUTO, OMP_CLAUSE_SEQ.  Use GOMP_MAP_* instead of
	OMP_CLAUSE_MAP_*.
	(expand_omp_for_static_nochunk, expand_omp_for_static_chunk):
	Handle GF_OMP_FOR_KIND_OACC_LOOP.
	(expand_omp_target, lower_omp_target): Handle
	GF_OMP_TARGET_KIND_OACC_PARALLEL, GF_OMP_TARGET_KIND_OACC_KERNELS,
	GF_OMP_TARGET_KIND_OACC_UPDATE,
	GF_OMP_TARGET_KIND_OACC_ENTER_EXIT_DATA,
	GF_OMP_TARGET_KIND_OACC_DATA.
	(pass_expand_omp::execute, execute_lower_omp)
	(pass_diagnose_omp_blocks::gate): Consider flag_openacc next to
	flag_openmp.
	(offload_symbol_decl): New variable.
	(oacc_get_reduction_array_id, oacc_max_threads)
	(get_offload_symbol_decl, get_base_type, lookup_oacc_reduction)
	(maybe_lookup_oacc_reduction, enclosing_target_ctx)
	(oacc_loop_or_target_p, oacc_lower_reduction_var_helper)
	(oacc_gimple_assign, oacc_initialize_reduction_data)
	(oacc_finalize_reduction_data, oacc_process_reduction_data): New
	functions.
	(is_targetreg_ctx): Remove function.
	* tree-core.h (enum omp_clause_code): Add OMP_CLAUSE__CACHE_,
	OMP_CLAUSE_DEVICE_RESIDENT, OMP_CLAUSE_USE_DEVICE,
	OMP_CLAUSE_GANG, OMP_CLAUSE_ASYNC, OMP_CLAUSE_WAIT,
	OMP_CLAUSE_AUTO, OMP_CLAUSE_SEQ, OMP_CLAUSE_INDEPENDENT,
	OMP_CLAUSE_WORKER, OMP_CLAUSE_VECTOR, OMP_CLAUSE_NUM_GANGS,
	OMP_CLAUSE_NUM_WORKERS, OMP_CLAUSE_VECTOR_LENGTH.
	* tree.c (omp_clause_code_name, walk_tree_1): Update accordingly.
	* tree.h (OMP_CLAUSE_GANG_EXPR, OMP_CLAUSE_GANG_STATIC_EXPR)
	(OMP_CLAUSE_ASYNC_EXPR, OMP_CLAUSE_WAIT_EXPR)
	(OMP_CLAUSE_VECTOR_EXPR, OMP_CLAUSE_WORKER_EXPR)
	(OMP_CLAUSE_NUM_GANGS_EXPR, OMP_CLAUSE_NUM_WORKERS_EXPR)
	(OMP_CLAUSE_VECTOR_LENGTH_EXPR): New macros.
	* tree-core.h: Update comments for OpenACC changes.
	(enum omp_clause_map_kind): Remove.
	(struct tree_omp_clause): Change type of map_kind member from enum
	omp_clause_map_kind to unsigned char.
	* tree-inline.c: Update comments for OpenACC changes.
	* tree-nested.c: Likewise.  Include "gomp-constants.h".
	(convert_nonlocal_reference_stmt, convert_local_reference_stmt)
	(convert_tramp_reference_stmt, convert_gimple_call): Update for
	OpenACC changes.  Use GOMP_MAP_* instead of OMP_CLAUSE_MAP_*.  Use
	OMP_CLAUSE_SET_MAP_KIND.
	* tree-pretty-print.c: Include "gomp-constants.h".
	(dump_omp_clause): Handle OMP_CLAUSE_DEVICE_RESIDENT,
	OMP_CLAUSE_USE_DEVICE, OMP_CLAUSE__CACHE_, OMP_CLAUSE_GANG,
	OMP_CLAUSE_ASYNC, OMP_CLAUSE_AUTO, OMP_CLAUSE_SEQ,
	OMP_CLAUSE_WAIT, OMP_CLAUSE_WORKER, OMP_CLAUSE_VECTOR,
	OMP_CLAUSE_NUM_GANGS, OMP_CLAUSE_NUM_WORKERS,
	OMP_CLAUSE_VECTOR_LENGTH, OMP_CLAUSE_INDEPENDENT.  Use GOMP_MAP_*
	instead of OMP_CLAUSE_MAP_*.
	(dump_generic_node): Handle OACC_PARALLEL, OACC_KERNELS,
	OACC_DATA, OACC_HOST_DATA, OACC_DECLARE, OACC_UPDATE,
	OACC_ENTER_DATA, OACC_EXIT_DATA, OACC_CACHE, OACC_LOOP.
	* tree-streamer-in.c: Include "gomp-constants.h".
	(unpack_ts_omp_clause_value_fields) Use GOMP_MAP_* instead of
	OMP_CLAUSE_MAP_*.  Use OMP_CLAUSE_SET_MAP_KIND.
	* tree-streamer-out.c: Include "gomp-constants.h".
	(pack_ts_omp_clause_value_fields): Use GOMP_MAP_* instead of
	OMP_CLAUSE_MAP_*.
	* tree.def (OACC_PARALLEL, OACC_KERNELS, OACC_DATA)
	(OACC_HOST_DATA, OACC_LOOP, OACC_CACHE, OACC_DECLARE)
	(OACC_ENTER_DATA, OACC_EXIT_DATA, OACC_UPDATE): New tree codes.
	* tree.c (omp_clause_num_ops): Update accordingly.
	* tree.h (OMP_BODY, OMP_CLAUSES, OMP_LOOP_CHECK, OMP_CLAUSE_SIZE):
	Likewise.
	(OACC_PARALLEL_BODY, OACC_PARALLEL_CLAUSES, OACC_KERNELS_BODY)
	(OACC_KERNELS_CLAUSES, OACC_DATA_BODY, OACC_DATA_CLAUSES)
	(OACC_HOST_DATA_BODY, OACC_HOST_DATA_CLAUSES, OACC_CACHE_CLAUSES)
	(OACC_DECLARE_CLAUSES, OACC_ENTER_DATA_CLAUSES)
	(OACC_EXIT_DATA_CLAUSES, OACC_UPDATE_CLAUSES)
	(OACC_KERNELS_COMBINED, OACC_PARALLEL_COMBINED): New macros.
	* tree.h (OMP_CLAUSE_MAP_KIND): Cast it to enum gomp_map_kind.
	(OMP_CLAUSE_SET_MAP_KIND): New macro.
	* varpool.c (varpool_node::get_create): Consider flag_openacc next
	to flag_openmp.
	* config/i386/intelmic-offload.h: New file.
	* config/nvptx/offload.h: Likewise.
	gcc/ada/
	* gcc-interface/utils.c (DEF_FUNCTION_TYPE_VAR_8)
	(DEF_FUNCTION_TYPE_VAR_12): New macros.
	gcc/c-family/
	* c.opt (fopenacc): New option.
	* c-cppbuiltin.c (c_cpp_builtins): Conditionally define _OPENACC.
	* c-common.c (DEF_FUNCTION_TYPE_VAR_8, DEF_FUNCTION_TYPE_VAR_12):
	New macros.
	* c-common.h (c_finish_oacc_wait): New prototype.
	* c-omp.c: Include "omp-low.h" and "gomp-constants.h".
	(c_finish_oacc_wait): New function.
	* c-pragma.c (oacc_pragmas): New variable.
	(c_pp_lookup_pragma, init_pragma): Handle it.
	* c-pragma.h (enum pragma_kind): Add PRAGMA_OACC_CACHE,
	PRAGMA_OACC_DATA, PRAGMA_OACC_ENTER_DATA, PRAGMA_OACC_EXIT_DATA,
	PRAGMA_OACC_KERNELS, PRAGMA_OACC_LOOP, PRAGMA_OACC_PARALLEL,
	PRAGMA_OACC_UPDATE, PRAGMA_OACC_WAIT.
	(enum pragma_omp_clause): Add PRAGMA_OACC_CLAUSE_ASYNC,
	PRAGMA_OACC_CLAUSE_AUTO, PRAGMA_OACC_CLAUSE_COLLAPSE,
	PRAGMA_OACC_CLAUSE_COPY, PRAGMA_OACC_CLAUSE_COPYIN,
	PRAGMA_OACC_CLAUSE_COPYOUT, PRAGMA_OACC_CLAUSE_CREATE,
	PRAGMA_OACC_CLAUSE_DELETE, PRAGMA_OACC_CLAUSE_DEVICE,
	PRAGMA_OACC_CLAUSE_DEVICEPTR, PRAGMA_OACC_CLAUSE_FIRSTPRIVATE,
	PRAGMA_OACC_CLAUSE_GANG, PRAGMA_OACC_CLAUSE_HOST,
	PRAGMA_OACC_CLAUSE_IF, PRAGMA_OACC_CLAUSE_NUM_GANGS,
	PRAGMA_OACC_CLAUSE_NUM_WORKERS, PRAGMA_OACC_CLAUSE_PRESENT,
	PRAGMA_OACC_CLAUSE_PRESENT_OR_COPY,
	PRAGMA_OACC_CLAUSE_PRESENT_OR_COPYIN,
	PRAGMA_OACC_CLAUSE_PRESENT_OR_COPYOUT,
	PRAGMA_OACC_CLAUSE_PRESENT_OR_CREATE, PRAGMA_OACC_CLAUSE_PRIVATE,
	PRAGMA_OACC_CLAUSE_REDUCTION, PRAGMA_OACC_CLAUSE_SELF,
	PRAGMA_OACC_CLAUSE_SEQ, PRAGMA_OACC_CLAUSE_VECTOR,
	PRAGMA_OACC_CLAUSE_VECTOR_LENGTH, PRAGMA_OACC_CLAUSE_WAIT,
	PRAGMA_OACC_CLAUSE_WORKER.
	gcc/c/
	* c-parser.c: Include "gomp-constants.h".
	(c_parser_omp_clause_map): Use enum gomp_map_kind instead of enum
	omp_clause_map_kind.  Use GOMP_MAP_* instead of OMP_CLAUSE_MAP_*.
	Use OMP_CLAUSE_SET_MAP_KIND.
	(c_parser_pragma): Handle PRAGMA_OACC_ENTER_DATA,
	PRAGMA_OACC_EXIT_DATA, PRAGMA_OACC_UPDATE.
	(c_parser_omp_construct): Handle PRAGMA_OACC_CACHE,
	PRAGMA_OACC_DATA, PRAGMA_OACC_KERNELS, PRAGMA_OACC_LOOP,
	PRAGMA_OACC_PARALLEL, PRAGMA_OACC_WAIT.
	(c_parser_omp_clause_name): Handle "auto", "async", "copy",
	"copyout", "create", "delete", "deviceptr", "gang", "host",
	"num_gangs", "num_workers", "present", "present_or_copy", "pcopy",
	"present_or_copyin", "pcopyin", "present_or_copyout", "pcopyout",
	"present_or_create", "pcreate", "seq", "self", "vector",
	"vector_length", "wait", "worker".
	(OACC_DATA_CLAUSE_MASK, OACC_KERNELS_CLAUSE_MASK)
	(OACC_ENTER_DATA_CLAUSE_MASK, OACC_EXIT_DATA_CLAUSE_MASK)
	(OACC_LOOP_CLAUSE_MASK, OACC_PARALLEL_CLAUSE_MASK)
	(OACC_UPDATE_CLAUSE_MASK, OACC_WAIT_CLAUSE_MASK): New macros.
	(c_parser_omp_variable_list): Handle OMP_CLAUSE__CACHE_.
	(c_parser_oacc_wait_list, c_parser_oacc_data_clause)
	(c_parser_oacc_data_clause_deviceptr)
	(c_parser_omp_clause_num_gangs, c_parser_omp_clause_num_workers)
	(c_parser_oacc_clause_async, c_parser_oacc_clause_wait)
	(c_parser_omp_clause_vector_length, c_parser_oacc_all_clauses)
	(c_parser_oacc_cache, c_parser_oacc_data, c_parser_oacc_kernels)
	(c_parser_oacc_enter_exit_data, c_parser_oacc_loop)
	(c_parser_oacc_parallel, c_parser_oacc_update)
	(c_parser_oacc_wait): New functions.
	* c-tree.h (c_finish_oacc_parallel, c_finish_oacc_kernels)
	(c_finish_oacc_data): New prototypes.
	* c-typeck.c: Include "gomp-constants.h".
	(handle_omp_array_sections): Handle GOMP_MAP_FORCE_DEVICEPTR.  Use
	GOMP_MAP_* instead of OMP_CLAUSE_MAP_*.  Use
	OMP_CLAUSE_SET_MAP_KIND.
	(c_finish_oacc_parallel, c_finish_oacc_kernels)
	(c_finish_oacc_data): New functions.
	(c_finish_omp_clauses): Handle OMP_CLAUSE__CACHE_,
	OMP_CLAUSE_NUM_GANGS, OMP_CLAUSE_NUM_WORKERS,
	OMP_CLAUSE_VECTOR_LENGTH, OMP_CLAUSE_ASYNC, OMP_CLAUSE_WAIT,
	OMP_CLAUSE_AUTO, OMP_CLAUSE_SEQ, OMP_CLAUSE_GANG,
	OMP_CLAUSE_WORKER, OMP_CLAUSE_VECTOR, and OMP_CLAUSE_MAP's
	GOMP_MAP_FORCE_DEVICEPTR.
	gcc/cp/
	* parser.c: Include "gomp-constants.h".
	(cp_parser_omp_clause_map): Use enum gomp_map_kind instead of enum
	omp_clause_map_kind.  Use GOMP_MAP_* instead of OMP_CLAUSE_MAP_*.
	Use OMP_CLAUSE_SET_MAP_KIND.
	(cp_parser_omp_construct, cp_parser_pragma): Handle
	PRAGMA_OACC_CACHE, PRAGMA_OACC_DATA, PRAGMA_OACC_ENTER_DATA,
	PRAGMA_OACC_EXIT_DATA, PRAGMA_OACC_KERNELS, PRAGMA_OACC_PARALLEL,
	PRAGMA_OACC_LOOP, PRAGMA_OACC_UPDATE, PRAGMA_OACC_WAIT.
	(cp_parser_omp_clause_name): Handle "async", "copy", "copyout",
	"create", "delete", "deviceptr", "host", "num_gangs",
	"num_workers", "present", "present_or_copy", "pcopy",
	"present_or_copyin", "pcopyin", "present_or_copyout", "pcopyout",
	"present_or_create", "pcreate", "vector_length", "wait".
	(OACC_DATA_CLAUSE_MASK, OACC_ENTER_DATA_CLAUSE_MASK)
	(OACC_EXIT_DATA_CLAUSE_MASK, OACC_KERNELS_CLAUSE_MASK)
	(OACC_LOOP_CLAUSE_MASK, OACC_PARALLEL_CLAUSE_MASK)
	(OACC_UPDATE_CLAUSE_MASK, OACC_WAIT_CLAUSE_MASK): New macros.
	(cp_parser_omp_var_list_no_open): Handle OMP_CLAUSE__CACHE_.
	(cp_parser_oacc_data_clause, cp_parser_oacc_data_clause_deviceptr)
	(cp_parser_oacc_clause_vector_length, cp_parser_oacc_wait_list)
	(cp_parser_oacc_clause_wait, cp_parser_omp_clause_num_gangs)
	(cp_parser_omp_clause_num_workers, cp_parser_oacc_clause_async)
	(cp_parser_oacc_all_clauses, cp_parser_oacc_cache)
	(cp_parser_oacc_data, cp_parser_oacc_enter_exit_data)
	(cp_parser_oacc_kernels, cp_parser_oacc_loop)
	(cp_parser_oacc_parallel, cp_parser_oacc_update)
	(cp_parser_oacc_wait): New functions.
	* cp-tree.h (finish_oacc_data, finish_oacc_kernels)
	(finish_oacc_parallel): New prototypes.
	* semantics.c: Include "gomp-constants.h".
	(handle_omp_array_sections): Handle GOMP_MAP_FORCE_DEVICEPTR.  Use
	GOMP_MAP_* instead of OMP_CLAUSE_MAP_*.  Use
	OMP_CLAUSE_SET_MAP_KIND.
	(finish_omp_clauses): Handle OMP_CLAUSE_ASYNC,
	OMP_CLAUSE_VECTOR_LENGTH, OMP_CLAUSE_WAIT, OMP_CLAUSE__CACHE_.
	Use GOMP_MAP_* instead of OMP_CLAUSE_MAP_*.
	(finish_oacc_data, finish_oacc_kernels, finish_oacc_parallel): New
	functions.
	gcc/fortran/
	* lang.opt (fopenacc): New option.
	* cpp.c (cpp_define_builtins): Conditionally define _OPENACC.
	* dump-parse-tree.c (show_omp_node): Split part of it into...
	(show_omp_clauses): ... this new function.
	(show_omp_node, show_code_node): Handle EXEC_OACC_PARALLEL_LOOP,
	EXEC_OACC_PARALLEL, EXEC_OACC_KERNELS_LOOP, EXEC_OACC_KERNELS,
	EXEC_OACC_DATA, EXEC_OACC_HOST_DATA, EXEC_OACC_LOOP,
	EXEC_OACC_UPDATE, EXEC_OACC_WAIT, EXEC_OACC_CACHE,
	EXEC_OACC_ENTER_DATA, EXEC_OACC_EXIT_DATA.
	(show_namespace): Update for OpenACC.
	* f95-lang.c (DEF_FUNCTION_TYPE_VAR_2, DEF_FUNCTION_TYPE_VAR_8)
	(DEF_FUNCTION_TYPE_VAR_12, DEF_GOACC_BUILTIN)
	(DEF_GOACC_BUILTIN_COMPILER): New macros.
	* types.def (BT_FN_VOID_INT_INT_VAR)
	(BT_FN_VOID_INT_PTR_SIZE_PTR_PTR_PTR_INT_INT_VAR)
	(BT_FN_VOID_INT_OMPFN_PTR_SIZE_PTR_PTR_PTR_INT_INT_INT_INT_INT_VAR):
	New function types.
	* gfortran.h (gfc_statement): Add ST_OACC_PARALLEL_LOOP,
	ST_OACC_END_PARALLEL_LOOP, ST_OACC_PARALLEL, ST_OACC_END_PARALLEL,
	ST_OACC_KERNELS, ST_OACC_END_KERNELS, ST_OACC_DATA,
	ST_OACC_END_DATA, ST_OACC_HOST_DATA, ST_OACC_END_HOST_DATA,
	ST_OACC_LOOP, ST_OACC_END_LOOP, ST_OACC_DECLARE, ST_OACC_UPDATE,
	ST_OACC_WAIT, ST_OACC_CACHE, ST_OACC_KERNELS_LOOP,
	ST_OACC_END_KERNELS_LOOP, ST_OACC_ENTER_DATA, ST_OACC_EXIT_DATA,
	ST_OACC_ROUTINE.
	(struct gfc_expr_list): New data type.
	(gfc_get_expr_list): New macro.
	(gfc_omp_map_op): Add OMP_MAP_FORCE_ALLOC, OMP_MAP_FORCE_DEALLOC,
	OMP_MAP_FORCE_TO, OMP_MAP_FORCE_FROM, OMP_MAP_FORCE_TOFROM,
	OMP_MAP_FORCE_PRESENT, OMP_MAP_FORCE_DEVICEPTR.
	(OMP_LIST_FIRST, OMP_LIST_DEVICE_RESIDENT, OMP_LIST_USE_DEVICE)
	(OMP_LIST_CACHE): New enumerators.
	(struct gfc_omp_clauses): Add async_expr, gang_expr, worker_expr,
	vector_expr, num_gangs_expr, num_workers_expr, vector_length_expr,
	wait_list, tile_list, async, gang, worker, vector, seq,
	independent, wait, par_auto, gang_static, and loc members.
	(struct gfc_namespace): Add oacc_declare_clauses member.
	(gfc_exec_op): Add EXEC_OACC_KERNELS_LOOP,
	EXEC_OACC_PARALLEL_LOOP, EXEC_OACC_PARALLEL, EXEC_OACC_KERNELS,
	EXEC_OACC_DATA, EXEC_OACC_HOST_DATA, EXEC_OACC_LOOP,
	EXEC_OACC_UPDATE, EXEC_OACC_WAIT, EXEC_OACC_CACHE,
	EXEC_OACC_ENTER_DATA, EXEC_OACC_EXIT_DATA.
	(gfc_free_expr_list, gfc_resolve_oacc_directive)
	(gfc_resolve_oacc_declare, gfc_resolve_oacc_parallel_loop_blocks)
	(gfc_resolve_oacc_blocks): New prototypes.
	* match.c (match_exit_cycle): Handle EXEC_OACC_LOOP and
	EXEC_OACC_PARALLEL_LOOP.
	* match.h (gfc_match_oacc_cache, gfc_match_oacc_wait)
	(gfc_match_oacc_update, gfc_match_oacc_declare)
	(gfc_match_oacc_loop, gfc_match_oacc_host_data)
	(gfc_match_oacc_data, gfc_match_oacc_kernels)
	(gfc_match_oacc_kernels_loop, gfc_match_oacc_parallel)
	(gfc_match_oacc_parallel_loop, gfc_match_oacc_enter_data)
	(gfc_match_oacc_exit_data, gfc_match_oacc_routine): New
	prototypes.
	* openmp.c: Include "diagnostic.h" and "gomp-constants.h".
	(gfc_free_omp_clauses): Update for members added to struct
	gfc_omp_clauses.
	(gfc_match_omp_clauses): Change mask paramter to uint64_t.  Add
	openacc parameter.
	(resolve_omp_clauses): Add openacc parameter.  Update for OpenACC.
	(struct fortran_omp_context): Add is_openmp member.
	(gfc_resolve_omp_parallel_blocks): Initialize it.
	(gfc_resolve_do_iterator): Update for OpenACC.
	(gfc_resolve_omp_directive): Call
	resolve_omp_directive_inside_oacc_region.
	(OMP_CLAUSE_PRIVATE, OMP_CLAUSE_FIRSTPRIVATE)
	(OMP_CLAUSE_LASTPRIVATE, OMP_CLAUSE_COPYPRIVATE)
	(OMP_CLAUSE_SHARED, OMP_CLAUSE_COPYIN, OMP_CLAUSE_REDUCTION)
	(OMP_CLAUSE_IF, OMP_CLAUSE_NUM_THREADS, OMP_CLAUSE_SCHEDULE)
	(OMP_CLAUSE_DEFAULT, OMP_CLAUSE_ORDERED, OMP_CLAUSE_COLLAPSE)
	(OMP_CLAUSE_UNTIED, OMP_CLAUSE_FINAL, OMP_CLAUSE_MERGEABLE)
	(OMP_CLAUSE_ALIGNED, OMP_CLAUSE_DEPEND, OMP_CLAUSE_INBRANCH)
	(OMP_CLAUSE_LINEAR, OMP_CLAUSE_NOTINBRANCH, OMP_CLAUSE_PROC_BIND)
	(OMP_CLAUSE_SAFELEN, OMP_CLAUSE_SIMDLEN, OMP_CLAUSE_UNIFORM)
	(OMP_CLAUSE_DEVICE, OMP_CLAUSE_MAP, OMP_CLAUSE_TO)
	(OMP_CLAUSE_FROM, OMP_CLAUSE_NUM_TEAMS, OMP_CLAUSE_THREAD_LIMIT)
	(OMP_CLAUSE_DIST_SCHEDULE): Use uint64_t.
	(OMP_CLAUSE_ASYNC, OMP_CLAUSE_NUM_GANGS, OMP_CLAUSE_NUM_WORKERS)
	(OMP_CLAUSE_VECTOR_LENGTH, OMP_CLAUSE_COPY, OMP_CLAUSE_COPYOUT)
	(OMP_CLAUSE_CREATE, OMP_CLAUSE_PRESENT)
	(OMP_CLAUSE_PRESENT_OR_COPY, OMP_CLAUSE_PRESENT_OR_COPYIN)
	(OMP_CLAUSE_PRESENT_OR_COPYOUT, OMP_CLAUSE_PRESENT_OR_CREATE)
	(OMP_CLAUSE_DEVICEPTR, OMP_CLAUSE_GANG, OMP_CLAUSE_WORKER)
	(OMP_CLAUSE_VECTOR, OMP_CLAUSE_SEQ, OMP_CLAUSE_INDEPENDENT)
	(OMP_CLAUSE_USE_DEVICE, OMP_CLAUSE_DEVICE_RESIDENT)
	(OMP_CLAUSE_HOST_SELF, OMP_CLAUSE_OACC_DEVICE, OMP_CLAUSE_WAIT)
	(OMP_CLAUSE_DELETE, OMP_CLAUSE_AUTO, OMP_CLAUSE_TILE): New macros.
	(gfc_match_omp_clauses): Handle those.
	(OACC_PARALLEL_CLAUSES, OACC_KERNELS_CLAUSES, OACC_DATA_CLAUSES)
	(OACC_LOOP_CLAUSES, OACC_PARALLEL_LOOP_CLAUSES)
	(OACC_KERNELS_LOOP_CLAUSES, OACC_HOST_DATA_CLAUSES)
	(OACC_DECLARE_CLAUSES, OACC_UPDATE_CLAUSES)
	(OACC_ENTER_DATA_CLAUSES, OACC_EXIT_DATA_CLAUSES)
	(OACC_WAIT_CLAUSES): New macros.
	(gfc_free_expr_list, match_oacc_expr_list, match_oacc_clause_gang)
	(gfc_match_omp_map_clause, gfc_match_oacc_parallel_loop)
	(gfc_match_oacc_parallel, gfc_match_oacc_kernels_loop)
	(gfc_match_oacc_kernels, gfc_match_oacc_data)
	(gfc_match_oacc_host_data, gfc_match_oacc_loop)
	(gfc_match_oacc_declare, gfc_match_oacc_update)
	(gfc_match_oacc_enter_data, gfc_match_oacc_exit_data)
	(gfc_match_oacc_wait, gfc_match_oacc_cache)
	(gfc_match_oacc_routine, oacc_is_loop)
	(resolve_oacc_scalar_int_expr, resolve_oacc_positive_int_expr)
	(check_symbol_not_pointer, check_array_not_assumed)
	(resolve_oacc_data_clauses, resolve_oacc_deviceptr_clause)
	(oacc_compatible_clauses, oacc_is_parallel, oacc_is_kernels)
	(omp_code_to_statement, oacc_code_to_statement)
	(resolve_oacc_directive_inside_omp_region)
	(resolve_omp_directive_inside_oacc_region)
	(resolve_oacc_nested_loops, resolve_oacc_params_in_parallel)
	(resolve_oacc_loop_blocks, gfc_resolve_oacc_blocks)
	(resolve_oacc_loop, resolve_oacc_cache, gfc_resolve_oacc_declare)
	(gfc_resolve_oacc_directive): New functions.
	* parse.c (next_free): Update for OpenACC.  Move some code into...
	(verify_token_free): ... this new function.
	(next_fixed): Update for OpenACC.  Move some code into...
	(verify_token_fixed): ... this new function.
	(case_executable): Add ST_OACC_UPDATE, ST_OACC_WAIT,
	ST_OACC_CACHE, ST_OACC_ENTER_DATA, and ST_OACC_EXIT_DATA.
	(case_exec_markers): Add ST_OACC_PARALLEL_LOOP, ST_OACC_PARALLEL,
	ST_OACC_KERNELS, ST_OACC_DATA, ST_OACC_HOST_DATA, ST_OACC_LOOP,
	ST_OACC_KERNELS_LOOP.
	(case_decl): Add ST_OACC_ROUTINE.
	(push_state, parse_critical_block, parse_progunit): Update for
	OpenACC.
	(gfc_ascii_statement): Handle ST_OACC_PARALLEL_LOOP,
	ST_OACC_END_PARALLEL_LOOP, ST_OACC_PARALLEL, ST_OACC_END_PARALLEL,
	ST_OACC_KERNELS, ST_OACC_END_KERNELS, ST_OACC_KERNELS_LOOP,
	ST_OACC_END_KERNELS_LOOP, ST_OACC_DATA, ST_OACC_END_DATA,
	ST_OACC_HOST_DATA, ST_OACC_END_HOST_DATA, ST_OACC_LOOP,
	ST_OACC_END_LOOP, ST_OACC_DECLARE, ST_OACC_UPDATE, ST_OACC_WAIT,
	ST_OACC_CACHE, ST_OACC_ENTER_DATA, ST_OACC_EXIT_DATA,
	ST_OACC_ROUTINE.
	(verify_st_order, parse_spec): Handle ST_OACC_DECLARE.
	(parse_executable): Handle ST_OACC_PARALLEL_LOOP,
	ST_OACC_KERNELS_LOOP, ST_OACC_LOOP, ST_OACC_PARALLEL,
	ST_OACC_KERNELS, ST_OACC_DATA, ST_OACC_HOST_DATA.
	(decode_oacc_directive, parse_oacc_structured_block)
	(parse_oacc_loop, is_oacc): New functions.
	* parse.h (struct gfc_state_data): Add oacc_declare_clauses
	member.
	(is_oacc): New prototype.
	* resolve.c (gfc_resolve_blocks, gfc_resolve_code): Handle
	EXEC_OACC_PARALLEL_LOOP, EXEC_OACC_PARALLEL,
	EXEC_OACC_KERNELS_LOOP, EXEC_OACC_KERNELS, EXEC_OACC_DATA,
	EXEC_OACC_HOST_DATA, EXEC_OACC_LOOP, EXEC_OACC_UPDATE,
	EXEC_OACC_WAIT, EXEC_OACC_CACHE, EXEC_OACC_ENTER_DATA,
	EXEC_OACC_EXIT_DATA.
	(resolve_codes): Call gfc_resolve_oacc_declare.
	* scanner.c (openacc_flag, openacc_locus): New variables.
	(skip_free_comments): Update for OpenACC.  Move some code into...
	(skip_omp_attribute): ... this new function.
	(skip_oacc_attribute): New function.
	(skip_fixed_comments, gfc_next_char_literal): Update for OpenACC.
	* st.c (gfc_free_statement): Handle EXEC_OACC_PARALLEL_LOOP,
	EXEC_OACC_PARALLEL, EXEC_OACC_KERNELS_LOOP, EXEC_OACC_KERNELS,
	EXEC_OACC_DATA, EXEC_OACC_HOST_DATA, EXEC_OACC_LOOP,
	EXEC_OACC_UPDATE, EXEC_OACC_WAIT, EXEC_OACC_CACHE,
	EXEC_OACC_ENTER_DATA, EXEC_OACC_EXIT_DATA.
	* trans-decl.c (gfc_generate_function_code): Update for OpenACC.
	* trans-openmp.c: Include "gomp-constants.h".
	(gfc_omp_finish_clause, gfc_trans_omp_clauses): Use GOMP_MAP_*
	instead of OMP_CLAUSE_MAP_*.  Use OMP_CLAUSE_SET_MAP_KIND.
	(gfc_trans_omp_clauses): Handle OMP_LIST_USE_DEVICE,
	OMP_LIST_DEVICE_RESIDENT, OMP_LIST_CACHE, and OMP_MAP_FORCE_ALLOC,
	OMP_MAP_FORCE_DEALLOC, OMP_MAP_FORCE_TO, OMP_MAP_FORCE_FROM,
	OMP_MAP_FORCE_TOFROM, OMP_MAP_FORCE_PRESENT,
	OMP_MAP_FORCE_DEVICEPTR, and gfc_omp_clauses' async, seq,
	independent, wait_list, num_gangs_expr, num_workers_expr,
	vector_length_expr, vector, vector_expr, worker, worker_expr,
	gang, gang_expr members.
	(gfc_trans_omp_do): Handle EXEC_OACC_LOOP.
	(gfc_convert_expr_to_tree, gfc_trans_oacc_construct)
	(gfc_trans_oacc_executable_directive)
	(gfc_trans_oacc_wait_directive, gfc_trans_oacc_combined_directive)
	(gfc_trans_oacc_declare, gfc_trans_oacc_directive): New functions.
	* trans-stmt.c (gfc_trans_block_construct): Update for OpenACC.
	* trans-stmt.h (gfc_trans_oacc_directive, gfc_trans_oacc_declare):
	New prototypes.
	* trans.c (tranc_code): Handle EXEC_OACC_CACHE, EXEC_OACC_WAIT,
	EXEC_OACC_UPDATE, EXEC_OACC_LOOP, EXEC_OACC_HOST_DATA,
	EXEC_OACC_DATA, EXEC_OACC_KERNELS, EXEC_OACC_KERNELS_LOOP,
	EXEC_OACC_PARALLEL, EXEC_OACC_PARALLEL_LOOP, EXEC_OACC_ENTER_DATA,
	EXEC_OACC_EXIT_DATA.
	* gfortran.texi: Update for OpenACC.
	* intrinsic.texi: Likewise.
	* invoke.texi: Likewise.
	gcc/lto/
	* lto-lang.c (DEF_FUNCTION_TYPE_VAR_8, DEF_FUNCTION_TYPE_VAR_12):
	New macros.
	* lto.c: Include "gomp-constants.h".
	gcc/testsuite/
	* lib/target-supports.exp (check_effective_target_fopenacc): New
	procedure.
	* g++.dg/goacc-gomp/goacc-gomp.exp: New file.
	* g++.dg/goacc/goacc.exp: Likewise.
	* gcc.dg/goacc-gomp/goacc-gomp.exp: Likewise.
	* gcc.dg/goacc/goacc.exp: Likewise.
	* gfortran.dg/goacc/goacc.exp: Likewise.
	* c-c++-common/cpp/openacc-define-1.c: New file.
	* c-c++-common/cpp/openacc-define-2.c: Likewise.
	* c-c++-common/cpp/openacc-define-3.c: Likewise.
	* c-c++-common/goacc-gomp/nesting-1.c: Likewise.
	* c-c++-common/goacc-gomp/nesting-fail-1.c: Likewise.
	* c-c++-common/goacc/acc_on_device-2-off.c: Likewise.
	* c-c++-common/goacc/acc_on_device-2.c: Likewise.
	* c-c++-common/goacc/asyncwait-1.c: Likewise.
	* c-c++-common/goacc/cache-1.c: Likewise.
	* c-c++-common/goacc/clauses-fail.c: Likewise.
	* c-c++-common/goacc/collapse-1.c: Likewise.
	* c-c++-common/goacc/data-1.c: Likewise.
	* c-c++-common/goacc/data-2.c: Likewise.
	* c-c++-common/goacc/data-clause-duplicate-1.c: Likewise.
	* c-c++-common/goacc/deviceptr-1.c: Likewise.
	* c-c++-common/goacc/deviceptr-2.c: Likewise.
	* c-c++-common/goacc/deviceptr-3.c: Likewise.
	* c-c++-common/goacc/if-clause-1.c: Likewise.
	* c-c++-common/goacc/if-clause-2.c: Likewise.
	* c-c++-common/goacc/kernels-1.c: Likewise.
	* c-c++-common/goacc/loop-1.c: Likewise.
	* c-c++-common/goacc/loop-private-1.c: Likewise.
	* c-c++-common/goacc/nesting-1.c: Likewise.
	* c-c++-common/goacc/nesting-data-1.c: Likewise.
	* c-c++-common/goacc/nesting-fail-1.c: Likewise.
	* c-c++-common/goacc/parallel-1.c: Likewise.
	* c-c++-common/goacc/pcopy.c: Likewise.
	* c-c++-common/goacc/pcopyin.c: Likewise.
	* c-c++-common/goacc/pcopyout.c: Likewise.
	* c-c++-common/goacc/pcreate.c: Likewise.
	* c-c++-common/goacc/pragma_context.c: Likewise.
	* c-c++-common/goacc/present-1.c: Likewise.
	* c-c++-common/goacc/reduction-1.c: Likewise.
	* c-c++-common/goacc/reduction-2.c: Likewise.
	* c-c++-common/goacc/reduction-3.c: Likewise.
	* c-c++-common/goacc/reduction-4.c: Likewise.
	* c-c++-common/goacc/sb-1.c: Likewise.
	* c-c++-common/goacc/sb-2.c: Likewise.
	* c-c++-common/goacc/sb-3.c: Likewise.
	* c-c++-common/goacc/update-1.c: Likewise.
	* gcc.dg/goacc/acc_on_device-1.c: Likewise.
	* gfortran.dg/goacc/acc_on_device-1.f95: Likewise.
	* gfortran.dg/goacc/acc_on_device-2-off.f95: Likewise.
	* gfortran.dg/goacc/acc_on_device-2.f95: Likewise.
	* gfortran.dg/goacc/assumed.f95: Likewise.
	* gfortran.dg/goacc/asyncwait-1.f95: Likewise.
	* gfortran.dg/goacc/asyncwait-2.f95: Likewise.
	* gfortran.dg/goacc/asyncwait-3.f95: Likewise.
	* gfortran.dg/goacc/asyncwait-4.f95: Likewise.
	* gfortran.dg/goacc/branch.f95: Likewise.
	* gfortran.dg/goacc/cache-1.f95: Likewise.
	* gfortran.dg/goacc/coarray.f95: Likewise.
	* gfortran.dg/goacc/continuation-free-form.f95: Likewise.
	* gfortran.dg/goacc/cray.f95: Likewise.
	* gfortran.dg/goacc/critical.f95: Likewise.
	* gfortran.dg/goacc/data-clauses.f95: Likewise.
	* gfortran.dg/goacc/data-tree.f95: Likewise.
	* gfortran.dg/goacc/declare-1.f95: Likewise.
	* gfortran.dg/goacc/enter-exit-data.f95: Likewise.
	* gfortran.dg/goacc/fixed-1.f: Likewise.
	* gfortran.dg/goacc/fixed-2.f: Likewise.
	* gfortran.dg/goacc/fixed-3.f: Likewise.
	* gfortran.dg/goacc/fixed-4.f: Likewise.
	* gfortran.dg/goacc/host_data-tree.f95: Likewise.
	* gfortran.dg/goacc/if.f95: Likewise.
	* gfortran.dg/goacc/kernels-tree.f95: Likewise.
	* gfortran.dg/goacc/list.f95: Likewise.
	* gfortran.dg/goacc/literal.f95: Likewise.
	* gfortran.dg/goacc/loop-1.f95: Likewise.
	* gfortran.dg/goacc/loop-2.f95: Likewise.
	* gfortran.dg/goacc/loop-3.f95: Likewise.
	* gfortran.dg/goacc/loop-tree-1.f90: Likewise.
	* gfortran.dg/goacc/omp.f95: Likewise.
	* gfortran.dg/goacc/parallel-kernels-clauses.f95: Likewise.
	* gfortran.dg/goacc/parallel-kernels-regions.f95: Likewise.
	* gfortran.dg/goacc/parallel-tree.f95: Likewise.
	* gfortran.dg/goacc/parameter.f95: Likewise.
	* gfortran.dg/goacc/private-1.f95: Likewise.
	* gfortran.dg/goacc/private-2.f95: Likewise.
	* gfortran.dg/goacc/private-3.f95: Likewise.
	* gfortran.dg/goacc/pure-elemental-procedures.f95: Likewise.
	* gfortran.dg/goacc/reduction-2.f95: Likewise.
	* gfortran.dg/goacc/reduction.f95: Likewise.
	* gfortran.dg/goacc/routine-1.f90: Likewise.
	* gfortran.dg/goacc/routine-2.f90: Likewise.
	* gfortran.dg/goacc/sentinel-free-form.f95: Likewise.
	* gfortran.dg/goacc/several-directives.f95: Likewise.
	* gfortran.dg/goacc/sie.f95: Likewise.
	* gfortran.dg/goacc/subarrays.f95: Likewise.
	* gfortran.dg/gomp/map-1.f90: Likewise.
	* gfortran.dg/openacc-define-1.f90: Likewise.
	* gfortran.dg/openacc-define-2.f90: Likewise.
	* gfortran.dg/openacc-define-3.f90: Likewise.
	* g++.dg/gomp/block-1.C: Update for changed compiler output.
	* g++.dg/gomp/block-2.C: Likewise.
	* g++.dg/gomp/block-3.C: Likewise.
	* g++.dg/gomp/block-5.C: Likewise.
	* g++.dg/gomp/target-1.C: Likewise.
	* g++.dg/gomp/target-2.C: Likewise.
	* g++.dg/gomp/taskgroup-1.C: Likewise.
	* g++.dg/gomp/teams-1.C: Likewise.
	* gcc.dg/cilk-plus/jump-openmp.c: Likewise.
	* gcc.dg/cilk-plus/jump.c: Likewise.
	* gcc.dg/gomp/block-1.c: Likewise.
	* gcc.dg/gomp/block-10.c: Likewise.
	* gcc.dg/gomp/block-2.c: Likewise.
	* gcc.dg/gomp/block-3.c: Likewise.
	* gcc.dg/gomp/block-4.c: Likewise.
	* gcc.dg/gomp/block-5.c: Likewise.
	* gcc.dg/gomp/block-6.c: Likewise.
	* gcc.dg/gomp/block-7.c: Likewise.
	* gcc.dg/gomp/block-8.c: Likewise.
	* gcc.dg/gomp/block-9.c: Likewise.
	* gcc.dg/gomp/target-1.c: Likewise.
	* gcc.dg/gomp/target-2.c: Likewise.
	* gcc.dg/gomp/taskgroup-1.c: Likewise.
	* gcc.dg/gomp/teams-1.c: Likewise.
	include/
	* gomp-constants.h: New file.
	libgomp/
	* Makefile.am (search_path): Add $(top_srcdir)/../include.
	(libgomp_la_SOURCES): Add splay-tree.c, libgomp-plugin.c,
	oacc-parallel.c, oacc-host.c, oacc-init.c, oacc-mem.c,
	oacc-async.c, oacc-plugin.c, oacc-cuda.c.
	[USE_FORTRAN] (libgomp_la_SOURCES): Add openacc.f90.
	Include $(top_srcdir)/plugin/Makefrag.am.
	(nodist_libsubinclude_HEADERS): Add openacc.h.
	[USE_FORTRAN] (nodist_finclude_HEADERS): Add openacc_lib.h,
	openacc.f90, openacc.mod, openacc_kinds.mod.
	(omp_lib.mod): Generalize into...
	(%.mod): ... this new rule.
	(openacc_kinds.mod, openacc.mod): New rules.
	* plugin/configfrag.ac: New file.
	* configure.ac: Move plugin/offloading support into it.  Include
	it.  Instantiate testsuite/libgomp-test-support.pt.exp.
	* plugin/Makefrag.am: New file.
	* testsuite/Makefile.am (OFFLOAD_TARGETS)
	(OFFLOAD_ADDITIONAL_OPTIONS, OFFLOAD_ADDITIONAL_LIB_PATHS): Don't
	export.
	(libgomp-test-support.exp): New rule.
	(all-local): Depend on it.
	* Makefile.in: Regenerate.
	* testsuite/Makefile.in: Regenerate.
	* config.h.in: Likewise.
	* configure: Likewise.
	* configure.tgt: Harden shell syntax.
	* env.c: Include "oacc-int.h".
	(parse_acc_device_type): New function.
	(gomp_debug_var, goacc_device_type, goacc_device_num): New
	variables.
	(initialize_env): Initialize those.  Call
	goacc_runtime_initialize.
	* error.c (gomp_vdebug, gomp_debug, gomp_vfatal): New functions.
	(gomp_fatal): Call gomp_vfatal.
	* libgomp.h: Include "libgomp-plugin.h" and <stdarg.h>.
	(gomp_debug_var, goacc_device_type, goacc_device_num, gomp_vdebug)
	(gomp_debug, gomp_verror, gomp_vfatal, gomp_init_targets_once)
	(splay_tree_node, splay_tree, splay_tree_key)
	(struct target_mem_desc, struct splay_tree_key_s)
	(struct gomp_memory_mapping, struct acc_dispatch_t)
	(struct gomp_device_descr, gomp_acc_insert_pointer)
	(gomp_acc_remove_pointer, target_mem_desc, gomp_copy_from_async)
	(gomp_unmap_vars, gomp_init_device, gomp_init_tables)
	(gomp_free_memmap, gomp_fini_device): New declarations.
	(gomp_vdebug, gomp_debug): New macros.
	Include "splay-tree.h".
	* libgomp.map (OACC_2.0): New symbol version.  Use for
	acc_get_num_devices, acc_get_num_devices_h_, acc_set_device_type,
	acc_set_device_type_h_, acc_get_device_type,
	acc_get_device_type_h_, acc_set_device_num, acc_set_device_num_h_,
	acc_get_device_num, acc_get_device_num_h_, acc_async_test,
	acc_async_test_h_, acc_async_test_all, acc_async_test_all_h_,
	acc_wait, acc_wait_h_, acc_wait_async, acc_wait_async_h_,
	acc_wait_all, acc_wait_all_h_, acc_wait_all_async,
	acc_wait_all_async_h_, acc_init, acc_init_h_, acc_shutdown,
	acc_shutdown_h_, acc_on_device, acc_on_device_h_, acc_malloc,
	acc_free, acc_copyin, acc_copyin_32_h_, acc_copyin_64_h_,
	acc_copyin_array_h_, acc_present_or_copyin,
	acc_present_or_copyin_32_h_, acc_present_or_copyin_64_h_,
	acc_present_or_copyin_array_h_, acc_create, acc_create_32_h_,
	acc_create_64_h_, acc_create_array_h_, acc_present_or_create,
	acc_present_or_create_32_h_, acc_present_or_create_64_h_,
	acc_present_or_create_array_h_, acc_copyout, acc_copyout_32_h_,
	acc_copyout_64_h_, acc_copyout_array_h_, acc_delete,
	acc_delete_32_h_, acc_delete_64_h_, acc_delete_array_h_,
	acc_update_device, acc_update_device_32_h_,
	acc_update_device_64_h_, acc_update_device_array_h_,
	acc_update_self, acc_update_self_32_h_, acc_update_self_64_h_,
	acc_update_self_array_h_, acc_map_data, acc_unmap_data,
	acc_deviceptr, acc_hostptr, acc_is_present, acc_is_present_32_h_,
	acc_is_present_64_h_, acc_is_present_array_h_,
	acc_memcpy_to_device, acc_memcpy_from_device,
	acc_get_current_cuda_device, acc_get_current_cuda_context,
	acc_get_cuda_stream, acc_set_cuda_stream.
	(GOACC_2.0): New symbol version.  Use for GOACC_data_end,
	GOACC_data_start, GOACC_enter_exit_data, GOACC_parallel,
	GOACC_update, GOACC_wait, GOACC_get_thread_num,
	GOACC_get_num_threads.
	(GOMP_PLUGIN_1.0): New symbol version.  Use for
	GOMP_PLUGIN_malloc, GOMP_PLUGIN_malloc_cleared,
	GOMP_PLUGIN_realloc, GOMP_PLUGIN_debug, GOMP_PLUGIN_error,
	GOMP_PLUGIN_fatal, GOMP_PLUGIN_async_unmap_vars,
	GOMP_PLUGIN_acc_thread.
	* libgomp.texi: Update for OpenACC changes, and GOMP_DEBUG
	environment variable.
	* libgomp_g.h (GOACC_data_start, GOACC_data_end)
	(GOACC_enter_exit_data, GOACC_parallel, GOACC_update, GOACC_wait)
	(GOACC_get_num_threads, GOACC_get_thread_num): New declarations.
	* splay-tree.h (splay_tree_lookup, splay_tree_insert)
	(splay_tree_remove): New declarations.
	(rotate_left, rotate_right, splay_tree_splay, splay_tree_insert)
	(splay_tree_remove, splay_tree_lookup): Move into...
	* splay-tree.c: ... this new file.
	* target.c: Include "oacc-plugin.h", "oacc-int.h", <assert.h>.
	(splay_tree_node, splay_tree, splay_tree_key)
	(struct target_mem_desc, struct splay_tree_key_s)
	(struct gomp_device_descr): Don't declare.
	(num_devices_openmp): New variable.
	(gomp_get_num_devices ): Use it.
	(gomp_init_targets_once): New function.
	(gomp_get_num_devices ): Use it.
	(get_kind, gomp_copy_from_async, gomp_free_memmap)
	(gomp_fini_device, gomp_register_image_for_device): New functions.
	(gomp_map_vars): Add devaddrs parameter.
	(gomp_update): Add mm parameter.
	(gomp_init_device): Move most of it into...
	(gomp_init_tables): ... this new function.
	(gomp_register_images_for_device): Remove function.
	(splay_compare, gomp_map_vars, gomp_unmap_vars, gomp_init_device):
	Make them hidden instead of static.
	(gomp_map_vars_existing, gomp_map_vars, gomp_unmap_vars)
	(gomp_update, gomp_init_device, GOMP_target, GOMP_target_data)
	(GOMP_target_end_data, GOMP_target_update)
	(gomp_load_plugin_for_device, gomp_target_init): Update for
	OpenACC changes.
	* oacc-async.c: New file.
	* oacc-cuda.c: Likewise.
	* oacc-host.c: Likewise.
	* oacc-init.c: Likewise.
	* oacc-int.h: Likewise.
	* oacc-mem.c: Likewise.
	* oacc-parallel.c: Likewise.
	* oacc-plugin.c: Likewise.
	* oacc-plugin.h: Likewise.
	* oacc-ptx.h: Likewise.
	* openacc.f90: Likewise.
	* openacc.h: Likewise.
	* openacc_lib.h: Likewise.
	* plugin/plugin-host.c: Likewise.
	* plugin/plugin-nvptx.c: Likewise.
	* libgomp-plugin.c: Likewise.
	* libgomp-plugin.h: Likewise.
	* libgomp_target.h: Remove file after merging content into the
	former file.  Update all users.
	* testsuite/lib/libgomp.exp: Load libgomp-test-support.exp.
	(offload_targets_s, offload_targets_s_openacc): New variables.
	(check_effective_target_openacc_nvidia_accel_present)
	(check_effective_target_openacc_nvidia_accel_selected): New
	procedures.
	(libgomp_init): Update for OpenACC changes.
	* testsuite/libgomp-test-support.exp.in: New file.
	* testsuite/libgomp.oacc-c++/c++.exp: Likewise.
	* testsuite/libgomp.oacc-c/c.exp: Likewise.
	* testsuite/libgomp.oacc-fortran/fortran.exp: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/abort-1.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/abort-2.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/abort-3.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/abort-4.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/acc_on_device-1.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/asyncwait-1.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/cache-1.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/clauses-1.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/clauses-2.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/collapse-1.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/collapse-2.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/collapse-3.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/collapse-4.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/context-1.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/context-2.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/context-3.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/context-4.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/data-1.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/data-2.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/data-3.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/data-already-1.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/data-already-2.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/data-already-3.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/data-already-4.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/data-already-5.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/data-already-6.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/data-already-7.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/data-already-8.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/deviceptr-1.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/if-1.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/kernels-1.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/kernels-empty.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-1.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-10.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-11.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-12.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-13.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-14.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-15.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-16.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-17.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-18.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-19.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-2.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-20.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-21.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-22.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-23.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-24.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-25.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-26.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-27.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-28.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-29.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-3.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-30.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-31.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-32.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-33.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-34.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-35.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-36.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-37.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-38.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-39.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-4.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-40.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-41.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-42.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-43.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-44.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-45.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-46.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-47.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-48.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-49.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-5.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-50.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-51.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-52.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-53.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-54.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-55.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-56.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-57.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-58.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-59.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-6.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-60.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-61.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-62.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-63.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-64.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-65.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-66.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-67.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-68.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-69.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-7.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-70.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-71.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-72.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-73.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-74.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-75.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-76.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-77.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-78.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-79.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-80.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-81.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-82.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-83.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-84.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-85.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-86.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-87.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-88.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-89.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-9.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-90.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-91.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/lib-92.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/nested-1.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/nested-2.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/offset-1.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/parallel-1.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/parallel-empty.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/pointer-align-1.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/present-1.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/present-2.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/reduction-1.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/reduction-2.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/reduction-3.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/reduction-4.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/reduction-5.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/reduction-initial-1.c:
	Likewise.
	* testsuite/libgomp.oacc-c-c++-common/subr.h: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/subr.ptx: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/timer.h: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/update-1-2.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/update-1.c: Likewise.
	* testsuite/libgomp.oacc-fortran/abort-1.f90: Likewise.
	* testsuite/libgomp.oacc-fortran/abort-2.f90: Likewise.
	* testsuite/libgomp.oacc-fortran/acc_on_device-1-1.f90: Likewise.
	* testsuite/libgomp.oacc-fortran/acc_on_device-1-2.f: Likewise.
	* testsuite/libgomp.oacc-fortran/acc_on_device-1-3.f: Likewise.
	* testsuite/libgomp.oacc-fortran/asyncwait-1.f90: Likewise.
	* testsuite/libgomp.oacc-fortran/asyncwait-2.f90: Likewise.
	* testsuite/libgomp.oacc-fortran/asyncwait-3.f90: Likewise.
	* testsuite/libgomp.oacc-fortran/collapse-1.f90: Likewise.
	* testsuite/libgomp.oacc-fortran/collapse-2.f90: Likewise.
	* testsuite/libgomp.oacc-fortran/collapse-3.f90: Likewise.
	* testsuite/libgomp.oacc-fortran/collapse-4.f90: Likewise.
	* testsuite/libgomp.oacc-fortran/collapse-5.f90: Likewise.
	* testsuite/libgomp.oacc-fortran/collapse-6.f90: Likewise.
	* testsuite/libgomp.oacc-fortran/collapse-7.f90: Likewise.
	* testsuite/libgomp.oacc-fortran/collapse-8.f90: Likewise.
	* testsuite/libgomp.oacc-fortran/data-1.f90: Likewise.
	* testsuite/libgomp.oacc-fortran/data-2.f90: Likewise.
	* testsuite/libgomp.oacc-fortran/data-3.f90: Likewise.
	* testsuite/libgomp.oacc-fortran/data-4-2.f90: Likewise.
	* testsuite/libgomp.oacc-fortran/data-4.f90: Likewise.
	* testsuite/libgomp.oacc-fortran/data-already-1.f: Likewise.
	* testsuite/libgomp.oacc-fortran/data-already-2.f: Likewise.
	* testsuite/libgomp.oacc-fortran/data-already-3.f: Likewise.
	* testsuite/libgomp.oacc-fortran/data-already-4.f: Likewise.
	* testsuite/libgomp.oacc-fortran/data-already-5.f: Likewise.
	* testsuite/libgomp.oacc-fortran/data-already-6.f: Likewise.
	* testsuite/libgomp.oacc-fortran/data-already-7.f: Likewise.
	* testsuite/libgomp.oacc-fortran/data-already-8.f: Likewise.
	* testsuite/libgomp.oacc-fortran/lib-1.f90: Likewise.
	* testsuite/libgomp.oacc-fortran/lib-10.f90: Likewise.
	* testsuite/libgomp.oacc-fortran/lib-2.f: Likewise.
	* testsuite/libgomp.oacc-fortran/lib-3.f: Likewise.
	* testsuite/libgomp.oacc-fortran/lib-4.f90: Likewise.
	* testsuite/libgomp.oacc-fortran/lib-5.f90: Likewise.
	* testsuite/libgomp.oacc-fortran/lib-6.f90: Likewise.
	* testsuite/libgomp.oacc-fortran/lib-7.f90: Likewise.
	* testsuite/libgomp.oacc-fortran/lib-8.f90: Likewise.
	* testsuite/libgomp.oacc-fortran/map-1.f90: Likewise.
	* testsuite/libgomp.oacc-fortran/openacc_version-1.f: Likewise.
	* testsuite/libgomp.oacc-fortran/openacc_version-2.f90: Likewise.
	* testsuite/libgomp.oacc-fortran/pointer-align-1.f90: Likewise.
	* testsuite/libgomp.oacc-fortran/pset-1.f90: Likewise.
	* testsuite/libgomp.oacc-fortran/reduction-1.f90: Likewise.
	* testsuite/libgomp.oacc-fortran/reduction-2.f90: Likewise.
	* testsuite/libgomp.oacc-fortran/reduction-3.f90: Likewise.
	* testsuite/libgomp.oacc-fortran/reduction-4.f90: Likewise.
	* testsuite/libgomp.oacc-fortran/reduction-5.f90: Likewise.
	* testsuite/libgomp.oacc-fortran/reduction-6.f90: Likewise.
	* testsuite/libgomp.oacc-fortran/routine-1.f90: Likewise.
	* testsuite/libgomp.oacc-fortran/routine-2.f90: Likewise.
	* testsuite/libgomp.oacc-fortran/routine-3.f90: Likewise.
	* testsuite/libgomp.oacc-fortran/routine-4.f90: Likewise.
	* testsuite/libgomp.oacc-fortran/subarrays-1.f90: Likewise.
	* testsuite/libgomp.oacc-fortran/subarrays-2.f90: Likewise.
	liboffloadmic/
	* plugin/libgomp-plugin-intelmic.cpp (GOMP_OFFLOAD_get_name)
	(GOMP_OFFLOAD_get_caps, GOMP_OFFLOAD_fini_device): New functions.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@219682 138bc75d-0d04-0410-961f-82ee72b054a4

239 files changed, 23335 insertions, 548 deletions

diff --git a/libgomp/ChangeLog b/libgomp/ChangeLog
index 6e1e14169b5..9b003cbb027 100644
--- a/libgomp/ChangeLog
+++ b/libgomp/ChangeLog
@@ -1,3 +1,354 @@
+2015-01-15  Thomas Schwinge  <thomas@codesourcery.com>
+	    James Norris  <jnorris@codesourcery.com>
+	    Tom de Vries  <tom@codesourcery.com>
+	    Julian Brown  <julian@codesourcery.com>
+	    Cesar Philippidis  <cesar@codesourcery.com>
+	    Nathan Sidwell  <nathan@codesourcery.com>
+	    Tobias Burnus  <burnus@net-b.de>
+
+	* Makefile.am (search_path): Add $(top_srcdir)/../include.
+	(libgomp_la_SOURCES): Add splay-tree.c, libgomp-plugin.c,
+	oacc-parallel.c, oacc-host.c, oacc-init.c, oacc-mem.c,
+	oacc-async.c, oacc-plugin.c, oacc-cuda.c.
+	[USE_FORTRAN] (libgomp_la_SOURCES): Add openacc.f90.
+	Include $(top_srcdir)/plugin/Makefrag.am.
+	(nodist_libsubinclude_HEADERS): Add openacc.h.
+	[USE_FORTRAN] (nodist_finclude_HEADERS): Add openacc_lib.h,
+	openacc.f90, openacc.mod, openacc_kinds.mod.
+	(omp_lib.mod): Generalize into...
+	(%.mod): ... this new rule.
+	(openacc_kinds.mod, openacc.mod): New rules.
+	* plugin/configfrag.ac: New file.
+	* configure.ac: Move plugin/offloading support into it.  Include
+	it.  Instantiate testsuite/libgomp-test-support.pt.exp.
+	* plugin/Makefrag.am: New file.
+	* testsuite/Makefile.am (OFFLOAD_TARGETS)
+	(OFFLOAD_ADDITIONAL_OPTIONS, OFFLOAD_ADDITIONAL_LIB_PATHS): Don't
+	export.
+	(libgomp-test-support.exp): New rule.
+	(all-local): Depend on it.
+	* Makefile.in: Regenerate.
+	* testsuite/Makefile.in: Regenerate.
+	* config.h.in: Likewise.
+	* configure: Likewise.
+	* configure.tgt: Harden shell syntax.
+	* env.c: Include "oacc-int.h".
+	(parse_acc_device_type): New function.
+	(gomp_debug_var, goacc_device_type, goacc_device_num): New
+	variables.
+	(initialize_env): Initialize those.  Call
+	goacc_runtime_initialize.
+	* error.c (gomp_vdebug, gomp_debug, gomp_vfatal): New functions.
+	(gomp_fatal): Call gomp_vfatal.
+	* libgomp.h: Include "libgomp-plugin.h" and <stdarg.h>.
+	(gomp_debug_var, goacc_device_type, goacc_device_num, gomp_vdebug)
+	(gomp_debug, gomp_verror, gomp_vfatal, gomp_init_targets_once)
+	(splay_tree_node, splay_tree, splay_tree_key)
+	(struct target_mem_desc, struct splay_tree_key_s)
+	(struct gomp_memory_mapping, struct acc_dispatch_t)
+	(struct gomp_device_descr, gomp_acc_insert_pointer)
+	(gomp_acc_remove_pointer, target_mem_desc, gomp_copy_from_async)
+	(gomp_unmap_vars, gomp_init_device, gomp_init_tables)
+	(gomp_free_memmap, gomp_fini_device): New declarations.
+	(gomp_vdebug, gomp_debug): New macros.
+	Include "splay-tree.h".
+	* libgomp.map (OACC_2.0): New symbol version.  Use for
+	acc_get_num_devices, acc_get_num_devices_h_, acc_set_device_type,
+	acc_set_device_type_h_, acc_get_device_type,
+	acc_get_device_type_h_, acc_set_device_num, acc_set_device_num_h_,
+	acc_get_device_num, acc_get_device_num_h_, acc_async_test,
+	acc_async_test_h_, acc_async_test_all, acc_async_test_all_h_,
+	acc_wait, acc_wait_h_, acc_wait_async, acc_wait_async_h_,
+	acc_wait_all, acc_wait_all_h_, acc_wait_all_async,
+	acc_wait_all_async_h_, acc_init, acc_init_h_, acc_shutdown,
+	acc_shutdown_h_, acc_on_device, acc_on_device_h_, acc_malloc,
+	acc_free, acc_copyin, acc_copyin_32_h_, acc_copyin_64_h_,
+	acc_copyin_array_h_, acc_present_or_copyin,
+	acc_present_or_copyin_32_h_, acc_present_or_copyin_64_h_,
+	acc_present_or_copyin_array_h_, acc_create, acc_create_32_h_,
+	acc_create_64_h_, acc_create_array_h_, acc_present_or_create,
+	acc_present_or_create_32_h_, acc_present_or_create_64_h_,
+	acc_present_or_create_array_h_, acc_copyout, acc_copyout_32_h_,
+	acc_copyout_64_h_, acc_copyout_array_h_, acc_delete,
+	acc_delete_32_h_, acc_delete_64_h_, acc_delete_array_h_,
+	acc_update_device, acc_update_device_32_h_,
+	acc_update_device_64_h_, acc_update_device_array_h_,
+	acc_update_self, acc_update_self_32_h_, acc_update_self_64_h_,
+	acc_update_self_array_h_, acc_map_data, acc_unmap_data,
+	acc_deviceptr, acc_hostptr, acc_is_present, acc_is_present_32_h_,
+	acc_is_present_64_h_, acc_is_present_array_h_,
+	acc_memcpy_to_device, acc_memcpy_from_device,
+	acc_get_current_cuda_device, acc_get_current_cuda_context,
+	acc_get_cuda_stream, acc_set_cuda_stream.
+	(GOACC_2.0): New symbol version.  Use for GOACC_data_end,
+	GOACC_data_start, GOACC_enter_exit_data, GOACC_parallel,
+	GOACC_update, GOACC_wait, GOACC_get_thread_num,
+	GOACC_get_num_threads.
+	(GOMP_PLUGIN_1.0): New symbol version.  Use for
+	GOMP_PLUGIN_malloc, GOMP_PLUGIN_malloc_cleared,
+	GOMP_PLUGIN_realloc, GOMP_PLUGIN_debug, GOMP_PLUGIN_error,
+	GOMP_PLUGIN_fatal, GOMP_PLUGIN_async_unmap_vars,
+	GOMP_PLUGIN_acc_thread.
+	* libgomp.texi: Update for OpenACC changes, and GOMP_DEBUG
+	environment variable.
+	* libgomp_g.h (GOACC_data_start, GOACC_data_end)
+	(GOACC_enter_exit_data, GOACC_parallel, GOACC_update, GOACC_wait)
+	(GOACC_get_num_threads, GOACC_get_thread_num): New declarations.
+	* splay-tree.h (splay_tree_lookup, splay_tree_insert)
+	(splay_tree_remove): New declarations.
+	(rotate_left, rotate_right, splay_tree_splay, splay_tree_insert)
+	(splay_tree_remove, splay_tree_lookup): Move into...
+	* splay-tree.c: ... this new file.
+	* target.c: Include "oacc-plugin.h", "oacc-int.h", <assert.h>.
+	(splay_tree_node, splay_tree, splay_tree_key)
+	(struct target_mem_desc, struct splay_tree_key_s)
+	(struct gomp_device_descr): Don't declare.
+	(num_devices_openmp): New variable.
+	(gomp_get_num_devices ): Use it.
+	(gomp_init_targets_once): New function.
+	(gomp_get_num_devices ): Use it.
+	(get_kind, gomp_copy_from_async, gomp_free_memmap)
+	(gomp_fini_device, gomp_register_image_for_device): New functions.
+	(gomp_map_vars): Add devaddrs parameter.
+	(gomp_update): Add mm parameter.
+	(gomp_init_device): Move most of it into...
+	(gomp_init_tables): ... this new function.
+	(gomp_register_images_for_device): Remove function.
+	(splay_compare, gomp_map_vars, gomp_unmap_vars, gomp_init_device):
+	Make them hidden instead of static.
+	(gomp_map_vars_existing, gomp_map_vars, gomp_unmap_vars)
+	(gomp_update, gomp_init_device, GOMP_target, GOMP_target_data)
+	(GOMP_target_end_data, GOMP_target_update)
+	(gomp_load_plugin_for_device, gomp_target_init): Update for
+	OpenACC changes.
+	* oacc-async.c: New file.
+	* oacc-cuda.c: Likewise.
+	* oacc-host.c: Likewise.
+	* oacc-init.c: Likewise.
+	* oacc-int.h: Likewise.
+	* oacc-mem.c: Likewise.
+	* oacc-parallel.c: Likewise.
+	* oacc-plugin.c: Likewise.
+	* oacc-plugin.h: Likewise.
+	* oacc-ptx.h: Likewise.
+	* openacc.f90: Likewise.
+	* openacc.h: Likewise.
+	* openacc_lib.h: Likewise.
+	* plugin/plugin-host.c: Likewise.
+	* plugin/plugin-nvptx.c: Likewise.
+	* libgomp-plugin.c: Likewise.
+	* libgomp-plugin.h: Likewise.
+	* libgomp_target.h: Remove file after merging content into the
+	former file.  Update all users.
+	* testsuite/lib/libgomp.exp: Load libgomp-test-support.exp.
+	(offload_targets_s, offload_targets_s_openacc): New variables.
+	(check_effective_target_openacc_nvidia_accel_present)
+	(check_effective_target_openacc_nvidia_accel_selected): New
+	procedures.
+	(libgomp_init): Update for OpenACC changes.
+	* testsuite/libgomp-test-support.exp.in: New file.
+	* testsuite/libgomp.oacc-c++/c++.exp: Likewise.
+	* testsuite/libgomp.oacc-c/c.exp: Likewise.
+	* testsuite/libgomp.oacc-fortran/fortran.exp: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/abort-1.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/abort-2.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/abort-3.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/abort-4.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/acc_on_device-1.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/asyncwait-1.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/cache-1.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/clauses-1.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/clauses-2.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/collapse-1.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/collapse-2.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/collapse-3.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/collapse-4.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/context-1.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/context-2.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/context-3.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/context-4.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/data-1.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/data-2.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/data-3.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/data-already-1.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/data-already-2.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/data-already-3.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/data-already-4.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/data-already-5.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/data-already-6.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/data-already-7.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/data-already-8.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/deviceptr-1.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/if-1.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/kernels-1.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/kernels-empty.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-1.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-10.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-11.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-12.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-13.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-14.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-15.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-16.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-17.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-18.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-19.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-2.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-20.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-21.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-22.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-23.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-24.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-25.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-26.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-27.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-28.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-29.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-3.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-30.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-31.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-32.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-33.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-34.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-35.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-36.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-37.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-38.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-39.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-4.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-40.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-41.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-42.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-43.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-44.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-45.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-46.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-47.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-48.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-49.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-5.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-50.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-51.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-52.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-53.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-54.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-55.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-56.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-57.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-58.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-59.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-6.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-60.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-61.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-62.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-63.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-64.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-65.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-66.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-67.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-68.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-69.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-7.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-70.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-71.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-72.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-73.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-74.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-75.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-76.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-77.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-78.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-79.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-80.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-81.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-82.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-83.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-84.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-85.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-86.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-87.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-88.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-89.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-9.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-90.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-91.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/lib-92.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/nested-1.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/nested-2.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/offset-1.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/parallel-1.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/parallel-empty.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/pointer-align-1.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/present-1.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/present-2.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/reduction-1.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/reduction-2.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/reduction-3.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/reduction-4.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/reduction-5.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/reduction-initial-1.c:
+	Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/subr.h: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/subr.ptx: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/timer.h: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/update-1-2.c: Likewise.
+	* testsuite/libgomp.oacc-c-c++-common/update-1.c: Likewise.
+	* testsuite/libgomp.oacc-fortran/abort-1.f90: Likewise.
+	* testsuite/libgomp.oacc-fortran/abort-2.f90: Likewise.
+	* testsuite/libgomp.oacc-fortran/acc_on_device-1-1.f90: Likewise.
+	* testsuite/libgomp.oacc-fortran/acc_on_device-1-2.f: Likewise.
+	* testsuite/libgomp.oacc-fortran/acc_on_device-1-3.f: Likewise.
+	* testsuite/libgomp.oacc-fortran/asyncwait-1.f90: Likewise.
+	* testsuite/libgomp.oacc-fortran/asyncwait-2.f90: Likewise.
+	* testsuite/libgomp.oacc-fortran/asyncwait-3.f90: Likewise.
+	* testsuite/libgomp.oacc-fortran/collapse-1.f90: Likewise.
+	* testsuite/libgomp.oacc-fortran/collapse-2.f90: Likewise.
+	* testsuite/libgomp.oacc-fortran/collapse-3.f90: Likewise.
+	* testsuite/libgomp.oacc-fortran/collapse-4.f90: Likewise.
+	* testsuite/libgomp.oacc-fortran/collapse-5.f90: Likewise.
+	* testsuite/libgomp.oacc-fortran/collapse-6.f90: Likewise.
+	* testsuite/libgomp.oacc-fortran/collapse-7.f90: Likewise.
+	* testsuite/libgomp.oacc-fortran/collapse-8.f90: Likewise.
+	* testsuite/libgomp.oacc-fortran/data-1.f90: Likewise.
+	* testsuite/libgomp.oacc-fortran/data-2.f90: Likewise.
+	* testsuite/libgomp.oacc-fortran/data-3.f90: Likewise.
+	* testsuite/libgomp.oacc-fortran/data-4-2.f90: Likewise.
+	* testsuite/libgomp.oacc-fortran/data-4.f90: Likewise.
+	* testsuite/libgomp.oacc-fortran/data-already-1.f: Likewise.
+	* testsuite/libgomp.oacc-fortran/data-already-2.f: Likewise.
+	* testsuite/libgomp.oacc-fortran/data-already-3.f: Likewise.
+	* testsuite/libgomp.oacc-fortran/data-already-4.f: Likewise.
+	* testsuite/libgomp.oacc-fortran/data-already-5.f: Likewise.
+	* testsuite/libgomp.oacc-fortran/data-already-6.f: Likewise.
+	* testsuite/libgomp.oacc-fortran/data-already-7.f: Likewise.
+	* testsuite/libgomp.oacc-fortran/data-already-8.f: Likewise.
+	* testsuite/libgomp.oacc-fortran/lib-1.f90: Likewise.
+	* testsuite/libgomp.oacc-fortran/lib-10.f90: Likewise.
+	* testsuite/libgomp.oacc-fortran/lib-2.f: Likewise.
+	* testsuite/libgomp.oacc-fortran/lib-3.f: Likewise.
+	* testsuite/libgomp.oacc-fortran/lib-4.f90: Likewise.
+	* testsuite/libgomp.oacc-fortran/lib-5.f90: Likewise.
+	* testsuite/libgomp.oacc-fortran/lib-6.f90: Likewise.
+	* testsuite/libgomp.oacc-fortran/lib-7.f90: Likewise.
+	* testsuite/libgomp.oacc-fortran/lib-8.f90: Likewise.
+	* testsuite/libgomp.oacc-fortran/map-1.f90: Likewise.
+	* testsuite/libgomp.oacc-fortran/openacc_version-1.f: Likewise.
+	* testsuite/libgomp.oacc-fortran/openacc_version-2.f90: Likewise.
+	* testsuite/libgomp.oacc-fortran/pointer-align-1.f90: Likewise.
+	* testsuite/libgomp.oacc-fortran/pset-1.f90: Likewise.
+	* testsuite/libgomp.oacc-fortran/reduction-1.f90: Likewise.
+	* testsuite/libgomp.oacc-fortran/reduction-2.f90: Likewise.
+	* testsuite/libgomp.oacc-fortran/reduction-3.f90: Likewise.
+	* testsuite/libgomp.oacc-fortran/reduction-4.f90: Likewise.
+	* testsuite/libgomp.oacc-fortran/reduction-5.f90: Likewise.
+	* testsuite/libgomp.oacc-fortran/reduction-6.f90: Likewise.
+	* testsuite/libgomp.oacc-fortran/routine-1.f90: Likewise.
+	* testsuite/libgomp.oacc-fortran/routine-2.f90: Likewise.
+	* testsuite/libgomp.oacc-fortran/routine-3.f90: Likewise.
+	* testsuite/libgomp.oacc-fortran/routine-4.f90: Likewise.
+	* testsuite/libgomp.oacc-fortran/subarrays-1.f90: Likewise.
+	* testsuite/libgomp.oacc-fortran/subarrays-2.f90: Likewise.
+
 2015-01-10  Thomas Schwinge  <thomas@codesourcery.com>
 	    Julian Brown  <julian@codesourcery.com>
 	    David Malcolm  <dmalcolm@redhat.com>
diff --git a/libgomp/Makefile.am b/libgomp/Makefile.am
index 427415ee084..5411278b018 100644
--- a/libgomp/Makefile.am
+++ b/libgomp/Makefile.am
@@ -7,7 +7,8 @@ SUBDIRS = testsuite
 gcc_version := $(shell cat $(top_srcdir)/../gcc/BASE-VER)
 
 config_path = @config_path@
-search_path = $(addprefix $(top_srcdir)/config/, $(config_path)) $(top_srcdir)
+search_path = $(addprefix $(top_srcdir)/config/, $(config_path)) $(top_srcdir) \
+	      $(top_srcdir)/../include
 
 fincludedir = $(libdir)/gcc/$(target_alias)/$(gcc_version)/finclude
 libsubincludedir = $(libdir)/gcc/$(target_alias)/$(gcc_version)/include
@@ -60,12 +61,21 @@ libgomp_la_LINK = $(LINK) $(libgomp_la_LDFLAGS)
 libgomp_la_SOURCES = alloc.c barrier.c critical.c env.c error.c iter.c \
 	iter_ull.c loop.c loop_ull.c ordered.c parallel.c sections.c single.c \
 	task.c team.c work.c lock.c mutex.c proc.c sem.c bar.c ptrlock.c \
-	time.c fortran.c affinity.c target.c
+	time.c fortran.c affinity.c target.c splay-tree.c libgomp-plugin.c \
+	oacc-parallel.c oacc-host.c oacc-init.c oacc-mem.c oacc-async.c \
+	oacc-plugin.c oacc-cuda.c
+
+include $(top_srcdir)/plugin/Makefrag.am
+
+if USE_FORTRAN
+libgomp_la_SOURCES += openacc.f90
+endif
 
 nodist_noinst_HEADERS = libgomp_f.h
-nodist_libsubinclude_HEADERS = omp.h
+nodist_libsubinclude_HEADERS = omp.h openacc.h
 if USE_FORTRAN
-nodist_finclude_HEADERS = omp_lib.h omp_lib.f90 omp_lib.mod omp_lib_kinds.mod
+nodist_finclude_HEADERS = omp_lib.h omp_lib.f90 omp_lib.mod omp_lib_kinds.mod \
+	openacc_lib.h openacc.f90 openacc.mod openacc_kinds.mod
 endif
 
 LTLDFLAGS = $(shell $(SHELL) $(top_srcdir)/../libtool-ldflags $(LDFLAGS))
@@ -75,7 +85,11 @@ LINK = $(LIBTOOL) --tag CC $(AM_LIBTOOLFLAGS) $(LIBTOOLFLAGS) --mode=link \
 
 omp_lib_kinds.mod: omp_lib.mod
 	:
-omp_lib.mod: omp_lib.f90
+openacc_kinds.mod: openacc.mod
+	:
+openacc.mod: openacc.lo
+	:
+%.mod: %.f90
 	$(FC) $(FCFLAGS) -fsyntax-only $<
 fortran.lo: libgomp_f.h
 fortran.o: libgomp_f.h
diff --git a/libgomp/Makefile.in b/libgomp/Makefile.in
index 8e4774f6c0e..b61b108e36c 100644
--- a/libgomp/Makefile.in
+++ b/libgomp/Makefile.in
@@ -15,6 +15,34 @@
 
 @SET_MAKE@
 
+# Plugins for offload execution, Makefile.am fragment.
+#
+# Copyright (C) 2014-2015 Free Software Foundation, Inc.
+#
+# Contributed by Mentor Embedded.
+#
+# This file is part of the GNU Offloading and Multi Processing Library
+# (libgomp).
+#
+# Libgomp is free software; you can redistribute it and/or modify it
+# under the terms of the GNU General Public License as published by
+# the Free Software Foundation; either version 3, or (at your option)
+# any later version.
+#
+# Libgomp is distributed in the hope that it will be useful, but WITHOUT ANY
+# WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
+# FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+# more details.
+#
+# Under Section 7 of GPL version 3, you are granted additional
+# permissions described in the GCC Runtime Library Exception, version
+# 3.1, as published by the Free Software Foundation.
+#
+# You should have received a copy of the GNU General Public License and
+# a copy of the GCC Runtime Library Exception along with this program;
+# see the files COPYING3 and COPYING.RUNTIME respectively.  If not, see
+# <http://www.gnu.org/licenses/>.
+
 
 VPATH = @srcdir@
 pkgdatadir = $(datadir)/@PACKAGE@
@@ -36,13 +64,16 @@ POST_UNINSTALL = :
 build_triplet = @build@
 host_triplet = @host@
 target_triplet = @target@
-subdir = .
-DIST_COMMON = ChangeLog $(srcdir)/Makefile.in $(srcdir)/Makefile.am \
+DIST_COMMON = $(top_srcdir)/plugin/Makefrag.am ChangeLog \
+	$(srcdir)/Makefile.in $(srcdir)/Makefile.am \
 	$(top_srcdir)/configure $(am__configure_deps) \
 	$(srcdir)/config.h.in $(srcdir)/../mkinstalldirs \
 	$(srcdir)/omp.h.in $(srcdir)/omp_lib.h.in \
 	$(srcdir)/omp_lib.f90.in $(srcdir)/libgomp_f.h.in \
 	$(srcdir)/libgomp.spec.in $(srcdir)/../depcomp
+@PLUGIN_NVPTX_TRUE@am__append_1 = libgomp-plugin-nvptx.la
+@USE_FORTRAN_TRUE@am__append_2 = openacc.f90
+subdir = .
 ACLOCAL_M4 = $(top_srcdir)/aclocal.m4
 am__aclocal_m4_deps = $(top_srcdir)/../config/acx.m4 \
 	$(top_srcdir)/../config/depstand.m4 \
@@ -56,7 +87,8 @@ am__aclocal_m4_deps = $(top_srcdir)/../config/acx.m4 \
 	$(top_srcdir)/../config/tls.m4 $(top_srcdir)/../ltoptions.m4 \
 	$(top_srcdir)/../ltsugar.m4 $(top_srcdir)/../ltversion.m4 \
 	$(top_srcdir)/../lt~obsolete.m4 $(top_srcdir)/acinclude.m4 \
-	$(top_srcdir)/../libtool.m4 $(top_srcdir)/configure.ac
+	$(top_srcdir)/../libtool.m4 $(top_srcdir)/plugin/configfrag.ac \
+	$(top_srcdir)/configure.ac
 am__configure_deps = $(am__aclocal_m4_deps) $(CONFIGURE_DEPENDENCIES) \
 	$(ACLOCAL_M4)
 am__CONFIG_DISTCLEAN_FILES = config.status config.cache config.log \
@@ -91,12 +123,38 @@ am__installdirs = "$(DESTDIR)$(toolexeclibdir)" "$(DESTDIR)$(infodir)" \
 	"$(DESTDIR)$(fincludedir)" "$(DESTDIR)$(libsubincludedir)" \
 	"$(DESTDIR)$(toolexeclibdir)"
 LTLIBRARIES = $(toolexeclib_LTLIBRARIES)
+libgomp_plugin_host_nonshm_la_DEPENDENCIES = libgomp.la
+am_libgomp_plugin_host_nonshm_la_OBJECTS =  \
+	libgomp_plugin_host_nonshm_la-plugin-host.lo
+libgomp_plugin_host_nonshm_la_OBJECTS =  \
+	$(am_libgomp_plugin_host_nonshm_la_OBJECTS)
+libgomp_plugin_host_nonshm_la_LINK = $(LIBTOOL) --tag=CC \
+	$(libgomp_plugin_host_nonshm_la_LIBTOOLFLAGS) $(LIBTOOLFLAGS) \
+	--mode=link $(CCLD) $(AM_CFLAGS) $(CFLAGS) \
+	$(libgomp_plugin_host_nonshm_la_LDFLAGS) $(LDFLAGS) -o $@
+am__DEPENDENCIES_1 =
+@PLUGIN_NVPTX_TRUE@libgomp_plugin_nvptx_la_DEPENDENCIES = libgomp.la \
+@PLUGIN_NVPTX_TRUE@	$(am__DEPENDENCIES_1)
+@PLUGIN_NVPTX_TRUE@am_libgomp_plugin_nvptx_la_OBJECTS =  \
+@PLUGIN_NVPTX_TRUE@	libgomp_plugin_nvptx_la-plugin-nvptx.lo
+libgomp_plugin_nvptx_la_OBJECTS =  \
+	$(am_libgomp_plugin_nvptx_la_OBJECTS)
+libgomp_plugin_nvptx_la_LINK = $(LIBTOOL) --tag=CC \
+	$(libgomp_plugin_nvptx_la_LIBTOOLFLAGS) $(LIBTOOLFLAGS) \
+	--mode=link $(CCLD) $(AM_CFLAGS) $(CFLAGS) \
+	$(libgomp_plugin_nvptx_la_LDFLAGS) $(LDFLAGS) -o $@
+@PLUGIN_NVPTX_TRUE@am_libgomp_plugin_nvptx_la_rpath = -rpath \
+@PLUGIN_NVPTX_TRUE@	$(toolexeclibdir)
 libgomp_la_LIBADD =
+@USE_FORTRAN_TRUE@am__objects_1 = openacc.lo
 am_libgomp_la_OBJECTS = alloc.lo barrier.lo critical.lo env.lo \
 	error.lo iter.lo iter_ull.lo loop.lo loop_ull.lo ordered.lo \
 	parallel.lo sections.lo single.lo task.lo team.lo work.lo \
 	lock.lo mutex.lo proc.lo sem.lo bar.lo ptrlock.lo time.lo \
-	fortran.lo affinity.lo target.lo
+	fortran.lo affinity.lo target.lo splay-tree.lo \
+	libgomp-plugin.lo oacc-parallel.lo oacc-host.lo oacc-init.lo \
+	oacc-mem.lo oacc-async.lo oacc-plugin.lo oacc-cuda.lo \
+	$(am__objects_1)
 libgomp_la_OBJECTS = $(am_libgomp_la_OBJECTS)
 DEFAULT_INCLUDES = -I.@am__isrc@
 depcomp = $(SHELL) $(top_srcdir)/../depcomp
@@ -108,7 +166,15 @@ LTCOMPILE = $(LIBTOOL) --tag=CC $(AM_LIBTOOLFLAGS) $(LIBTOOLFLAGS) \
 	--mode=compile $(CC) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) \
 	$(AM_CPPFLAGS) $(CPPFLAGS) $(AM_CFLAGS) $(CFLAGS)
 CCLD = $(CC)
-SOURCES = $(libgomp_la_SOURCES)
+FCCOMPILE = $(FC) $(AM_FCFLAGS) $(FCFLAGS)
+LTFCCOMPILE = $(LIBTOOL) --tag=FC $(AM_LIBTOOLFLAGS) $(LIBTOOLFLAGS) \
+	--mode=compile $(FC) $(AM_FCFLAGS) $(FCFLAGS)
+FCLD = $(FC)
+FCLINK = $(LIBTOOL) --tag=FC $(AM_LIBTOOLFLAGS) $(LIBTOOLFLAGS) \
+	--mode=link $(FCLD) $(AM_FCFLAGS) $(FCFLAGS) $(AM_LDFLAGS) \
+	$(LDFLAGS) -o $@
+SOURCES = $(libgomp_plugin_host_nonshm_la_SOURCES) \
+	$(libgomp_plugin_nvptx_la_SOURCES) $(libgomp_la_SOURCES)
 MULTISRCTOP = 
 MULTIBUILDTOP = 
 MULTIDIRS = 
@@ -155,6 +221,8 @@ CCDEPMODE = @CCDEPMODE@
 CFLAGS = @CFLAGS@
 CPP = @CPP@
 CPPFLAGS = @CPPFLAGS@
+CUDA_DRIVER_INCLUDE = @CUDA_DRIVER_INCLUDE@
+CUDA_DRIVER_LIB = @CUDA_DRIVER_LIB@
 CYGPATH_W = @CYGPATH_W@
 DEFS = @DEFS@
 DEPDIR = @DEPDIR@
@@ -213,6 +281,10 @@ PACKAGE_URL = @PACKAGE_URL@
 PACKAGE_VERSION = @PACKAGE_VERSION@
 PATH_SEPARATOR = @PATH_SEPARATOR@
 PERL = @PERL@
+PLUGIN_NVPTX = @PLUGIN_NVPTX@
+PLUGIN_NVPTX_CPPFLAGS = @PLUGIN_NVPTX_CPPFLAGS@
+PLUGIN_NVPTX_LDFLAGS = @PLUGIN_NVPTX_LDFLAGS@
+PLUGIN_NVPTX_LIBS = @PLUGIN_NVPTX_LIBS@
 RANLIB = @RANLIB@
 SECTION_LDFLAGS = @SECTION_LDFLAGS@
 SED = @SED@
@@ -293,13 +365,16 @@ top_srcdir = @top_srcdir@
 ACLOCAL_AMFLAGS = -I .. -I ../config
 SUBDIRS = testsuite
 gcc_version := $(shell cat $(top_srcdir)/../gcc/BASE-VER)
-search_path = $(addprefix $(top_srcdir)/config/, $(config_path)) $(top_srcdir)
+search_path = $(addprefix $(top_srcdir)/config/, $(config_path)) $(top_srcdir) \
+	      $(top_srcdir)/../include
+
 fincludedir = $(libdir)/gcc/$(target_alias)/$(gcc_version)/finclude
 libsubincludedir = $(libdir)/gcc/$(target_alias)/$(gcc_version)/include
 AM_CPPFLAGS = $(addprefix -I, $(search_path))
 AM_CFLAGS = $(XCFLAGS)
 AM_LDFLAGS = $(XLDFLAGS) $(SECTION_LDFLAGS) $(OPT_LDFLAGS)
-toolexeclib_LTLIBRARIES = libgomp.la
+toolexeclib_LTLIBRARIES = libgomp.la $(am__append_1) \
+	libgomp-plugin-host_nonshm.la
 nodist_toolexeclib_HEADERS = libgomp.spec
 
 # -Wc is only a libtool option.
@@ -318,13 +393,35 @@ libgomp_la_LDFLAGS = $(libgomp_version_info) $(libgomp_version_script) \
 libgomp_la_DEPENDENCIES = $(libgomp_version_dep)
 libgomp_la_LINK = $(LINK) $(libgomp_la_LDFLAGS)
 libgomp_la_SOURCES = alloc.c barrier.c critical.c env.c error.c iter.c \
-	iter_ull.c loop.c loop_ull.c ordered.c parallel.c sections.c single.c \
-	task.c team.c work.c lock.c mutex.c proc.c sem.c bar.c ptrlock.c \
-	time.c fortran.c affinity.c target.c
-
+	iter_ull.c loop.c loop_ull.c ordered.c parallel.c sections.c \
+	single.c task.c team.c work.c lock.c mutex.c proc.c sem.c \
+	bar.c ptrlock.c time.c fortran.c affinity.c target.c \
+	splay-tree.c libgomp-plugin.c oacc-parallel.c oacc-host.c \
+	oacc-init.c oacc-mem.c oacc-async.c oacc-plugin.c oacc-cuda.c \
+	$(am__append_2)
+
+# Nvidia PTX OpenACC plugin.
+@PLUGIN_NVPTX_TRUE@libgomp_plugin_nvptx_version_info = -version-info $(libtool_VERSION)
+@PLUGIN_NVPTX_TRUE@libgomp_plugin_nvptx_la_SOURCES = plugin/plugin-nvptx.c
+@PLUGIN_NVPTX_TRUE@libgomp_plugin_nvptx_la_CPPFLAGS = $(AM_CPPFLAGS) $(PLUGIN_NVPTX_CPPFLAGS)
+@PLUGIN_NVPTX_TRUE@libgomp_plugin_nvptx_la_LDFLAGS =  \
+@PLUGIN_NVPTX_TRUE@	$(libgomp_plugin_nvptx_version_info) \
+@PLUGIN_NVPTX_TRUE@	$(lt_host_flags) $(PLUGIN_NVPTX_LDFLAGS)
+@PLUGIN_NVPTX_TRUE@libgomp_plugin_nvptx_la_LIBADD = libgomp.la $(PLUGIN_NVPTX_LIBS)
+@PLUGIN_NVPTX_TRUE@libgomp_plugin_nvptx_la_LIBTOOLFLAGS = --tag=disable-static
+libgomp_plugin_host_nonshm_version_info = -version-info $(libtool_VERSION)
+libgomp_plugin_host_nonshm_la_SOURCES = plugin/plugin-host.c
+libgomp_plugin_host_nonshm_la_CPPFLAGS = $(AM_CPPFLAGS) -DHOST_NONSHM_PLUGIN
+libgomp_plugin_host_nonshm_la_LDFLAGS = \
+	$(libgomp_plugin_host_nonshm_version_info) $(lt_host_flags)
+
+libgomp_plugin_host_nonshm_la_LIBADD = libgomp.la
+libgomp_plugin_host_nonshm_la_LIBTOOLFLAGS = --tag=disable-static
 nodist_noinst_HEADERS = libgomp_f.h
-nodist_libsubinclude_HEADERS = omp.h
-@USE_FORTRAN_TRUE@nodist_finclude_HEADERS = omp_lib.h omp_lib.f90 omp_lib.mod omp_lib_kinds.mod
+nodist_libsubinclude_HEADERS = omp.h openacc.h
+@USE_FORTRAN_TRUE@nodist_finclude_HEADERS = omp_lib.h omp_lib.f90 omp_lib.mod omp_lib_kinds.mod \
+@USE_FORTRAN_TRUE@	openacc_lib.h openacc.f90 openacc.mod openacc_kinds.mod
+
 LTLDFLAGS = $(shell $(SHELL) $(top_srcdir)/../libtool-ldflags $(LDFLAGS))
 LINK = $(LIBTOOL) --tag CC $(AM_LIBTOOLFLAGS) $(LIBTOOLFLAGS) --mode=link \
 	$(CCLD) $(AM_CFLAGS) $(CFLAGS) $(AM_LDFLAGS) $(LTLDFLAGS) -o $@
@@ -354,10 +451,10 @@ all: config.h
 	$(MAKE) $(AM_MAKEFLAGS) all-recursive
 
 .SUFFIXES:
-.SUFFIXES: .c .dvi .lo .o .obj .ps
+.SUFFIXES: .c .dvi .f90 .lo .o .obj .ps
 am--refresh:
 	@:
-$(srcdir)/Makefile.in: @MAINTAINER_MODE_TRUE@ $(srcdir)/Makefile.am  $(am__configure_deps)
+$(srcdir)/Makefile.in: @MAINTAINER_MODE_TRUE@ $(srcdir)/Makefile.am $(top_srcdir)/plugin/Makefrag.am $(am__configure_deps)
 	@for dep in $?; do \
 	  case '$(am__configure_deps)' in \
 	    *$$dep*) \
@@ -447,6 +544,10 @@ clean-toolexeclibLTLIBRARIES:
 	  echo "rm -f \"$${dir}/so_locations\""; \
 	  rm -f "$${dir}/so_locations"; \
 	done
+libgomp-plugin-host_nonshm.la: $(libgomp_plugin_host_nonshm_la_OBJECTS) $(libgomp_plugin_host_nonshm_la_DEPENDENCIES) 
+	$(libgomp_plugin_host_nonshm_la_LINK) -rpath $(toolexeclibdir) $(libgomp_plugin_host_nonshm_la_OBJECTS) $(libgomp_plugin_host_nonshm_la_LIBADD) $(LIBS)
+libgomp-plugin-nvptx.la: $(libgomp_plugin_nvptx_la_OBJECTS) $(libgomp_plugin_nvptx_la_DEPENDENCIES) 
+	$(libgomp_plugin_nvptx_la_LINK) $(am_libgomp_plugin_nvptx_la_rpath) $(libgomp_plugin_nvptx_la_OBJECTS) $(libgomp_plugin_nvptx_la_LIBADD) $(LIBS)
 libgomp.la: $(libgomp_la_OBJECTS) $(libgomp_la_DEPENDENCIES) 
 	$(libgomp_la_LINK) -rpath $(toolexeclibdir) $(libgomp_la_OBJECTS) $(libgomp_la_LIBADD) $(LIBS)
 
@@ -466,10 +567,20 @@ distclean-compile:
 @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/fortran.Plo@am__quote@
 @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/iter.Plo@am__quote@
 @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/iter_ull.Plo@am__quote@
+@AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/libgomp-plugin.Plo@am__quote@
+@AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/libgomp_plugin_host_nonshm_la-plugin-host.Plo@am__quote@
+@AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/libgomp_plugin_nvptx_la-plugin-nvptx.Plo@am__quote@
 @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/lock.Plo@am__quote@
 @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/loop.Plo@am__quote@
 @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/loop_ull.Plo@am__quote@
 @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/mutex.Plo@am__quote@
+@AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/oacc-async.Plo@am__quote@
+@AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/oacc-cuda.Plo@am__quote@
+@AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/oacc-host.Plo@am__quote@
+@AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/oacc-init.Plo@am__quote@
+@AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/oacc-mem.Plo@am__quote@
+@AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/oacc-parallel.Plo@am__quote@
+@AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/oacc-plugin.Plo@am__quote@
 @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/ordered.Plo@am__quote@
 @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/parallel.Plo@am__quote@
 @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/proc.Plo@am__quote@
@@ -477,6 +588,7 @@ distclean-compile:
 @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/sections.Plo@am__quote@
 @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/sem.Plo@am__quote@
 @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/single.Plo@am__quote@
+@AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/splay-tree.Plo@am__quote@
 @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/target.Plo@am__quote@
 @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/task.Plo@am__quote@
 @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/team.Plo@am__quote@
@@ -504,6 +616,29 @@ distclean-compile:
 @AMDEP_TRUE@@am__fastdepCC_FALSE@	DEPDIR=$(DEPDIR) $(CCDEPMODE) $(depcomp) @AMDEPBACKSLASH@
 @am__fastdepCC_FALSE@	$(LTCOMPILE) -c -o $@ $<
 
+libgomp_plugin_host_nonshm_la-plugin-host.lo: plugin/plugin-host.c
+@am__fastdepCC_TRUE@	$(LIBTOOL)  --tag=CC $(libgomp_plugin_host_nonshm_la_LIBTOOLFLAGS) $(LIBTOOLFLAGS) --mode=compile $(CC) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) $(libgomp_plugin_host_nonshm_la_CPPFLAGS) $(CPPFLAGS) $(AM_CFLAGS) $(CFLAGS) -MT libgomp_plugin_host_nonshm_la-plugin-host.lo -MD -MP -MF $(DEPDIR)/libgomp_plugin_host_nonshm_la-plugin-host.Tpo -c -o libgomp_plugin_host_nonshm_la-plugin-host.lo `test -f 'plugin/plugin-host.c' || echo '$(srcdir)/'`plugin/plugin-host.c
+@am__fastdepCC_TRUE@	$(am__mv) $(DEPDIR)/libgomp_plugin_host_nonshm_la-plugin-host.Tpo $(DEPDIR)/libgomp_plugin_host_nonshm_la-plugin-host.Plo
+@AMDEP_TRUE@@am__fastdepCC_FALSE@	source='plugin/plugin-host.c' object='libgomp_plugin_host_nonshm_la-plugin-host.lo' libtool=yes @AMDEPBACKSLASH@
+@AMDEP_TRUE@@am__fastdepCC_FALSE@	DEPDIR=$(DEPDIR) $(CCDEPMODE) $(depcomp) @AMDEPBACKSLASH@
+@am__fastdepCC_FALSE@	$(LIBTOOL)  --tag=CC $(libgomp_plugin_host_nonshm_la_LIBTOOLFLAGS) $(LIBTOOLFLAGS) --mode=compile $(CC) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) $(libgomp_plugin_host_nonshm_la_CPPFLAGS) $(CPPFLAGS) $(AM_CFLAGS) $(CFLAGS) -c -o libgomp_plugin_host_nonshm_la-plugin-host.lo `test -f 'plugin/plugin-host.c' || echo '$(srcdir)/'`plugin/plugin-host.c
+
+libgomp_plugin_nvptx_la-plugin-nvptx.lo: plugin/plugin-nvptx.c
+@am__fastdepCC_TRUE@	$(LIBTOOL)  --tag=CC $(libgomp_plugin_nvptx_la_LIBTOOLFLAGS) $(LIBTOOLFLAGS) --mode=compile $(CC) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) $(libgomp_plugin_nvptx_la_CPPFLAGS) $(CPPFLAGS) $(AM_CFLAGS) $(CFLAGS) -MT libgomp_plugin_nvptx_la-plugin-nvptx.lo -MD -MP -MF $(DEPDIR)/libgomp_plugin_nvptx_la-plugin-nvptx.Tpo -c -o libgomp_plugin_nvptx_la-plugin-nvptx.lo `test -f 'plugin/plugin-nvptx.c' || echo '$(srcdir)/'`plugin/plugin-nvptx.c
+@am__fastdepCC_TRUE@	$(am__mv) $(DEPDIR)/libgomp_plugin_nvptx_la-plugin-nvptx.Tpo $(DEPDIR)/libgomp_plugin_nvptx_la-plugin-nvptx.Plo
+@AMDEP_TRUE@@am__fastdepCC_FALSE@	source='plugin/plugin-nvptx.c' object='libgomp_plugin_nvptx_la-plugin-nvptx.lo' libtool=yes @AMDEPBACKSLASH@
+@AMDEP_TRUE@@am__fastdepCC_FALSE@	DEPDIR=$(DEPDIR) $(CCDEPMODE) $(depcomp) @AMDEPBACKSLASH@
+@am__fastdepCC_FALSE@	$(LIBTOOL)  --tag=CC $(libgomp_plugin_nvptx_la_LIBTOOLFLAGS) $(LIBTOOLFLAGS) --mode=compile $(CC) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) $(libgomp_plugin_nvptx_la_CPPFLAGS) $(CPPFLAGS) $(AM_CFLAGS) $(CFLAGS) -c -o libgomp_plugin_nvptx_la-plugin-nvptx.lo `test -f 'plugin/plugin-nvptx.c' || echo '$(srcdir)/'`plugin/plugin-nvptx.c
+
+.f90.o:
+	$(FCCOMPILE) -c -o $@ $<
+
+.f90.obj:
+	$(FCCOMPILE) -c -o $@ `$(CYGPATH_W) '$<'`
+
+.f90.lo:
+	$(LTFCCOMPILE) -c -o $@ $<
+
 mostlyclean-libtool:
 	-rm -f *.lo
 
@@ -1088,7 +1223,11 @@ vpath % $(strip $(search_path))
 
 omp_lib_kinds.mod: omp_lib.mod
 	:
-omp_lib.mod: omp_lib.f90
+openacc_kinds.mod: openacc.mod
+	:
+openacc.mod: openacc.lo
+	:
+%.mod: %.f90
 	$(FC) $(FCFLAGS) -fsyntax-only $<
 fortran.lo: libgomp_f.h
 fortran.o: libgomp_f.h
diff --git a/libgomp/config.h.in b/libgomp/config.h.in
index a5e27ca4eaf..02547b16341 100644
--- a/libgomp/config.h.in
+++ b/libgomp/config.h.in
@@ -79,9 +79,6 @@
 /* Define to 1 if the target supports thread-local storage. */
 #undef HAVE_TLS
 
-/* Define to 1 if the target use emutls for thread-local storage. */
-#undef USE_EMUTLS
-
 /* Define to 1 if you have the <unistd.h> header file. */
 #undef HAVE_UNISTD_H
 
@@ -116,6 +113,9 @@
 /* Define to the version of this package. */
 #undef PACKAGE_VERSION
 
+/* Define to 1 if the NVIDIA plugin is built, 0 if not. */
+#undef PLUGIN_NVPTX
+
 /* Define if all infrastructure, needed for plugins, is supported. */
 #undef PLUGIN_SUPPORT
 
@@ -143,5 +143,8 @@
 /* Define to 1 if you can safely include both <sys/time.h> and <time.h>. */
 #undef TIME_WITH_SYS_TIME
 
+/* Define to 1 if the target use emutls for thread-local storage. */
+#undef USE_EMUTLS
+
 /* Version number of package */
 #undef VERSION
diff --git a/libgomp/configure b/libgomp/configure
index 3214e9dd84f..081870728e7 100755
--- a/libgomp/configure
+++ b/libgomp/configure
@@ -616,9 +616,6 @@ OMP_LOCK_SIZE
 USE_FORTRAN_FALSE
 USE_FORTRAN_TRUE
 link_gomp
-offload_additional_lib_paths
-offload_additional_options
-offload_targets
 XLDFLAGS
 XCFLAGS
 config_path
@@ -630,6 +627,17 @@ LIBGOMP_BUILD_VERSIONED_SHLIB_FALSE
 LIBGOMP_BUILD_VERSIONED_SHLIB_TRUE
 OPT_LDFLAGS
 SECTION_LDFLAGS
+PLUGIN_NVPTX_FALSE
+PLUGIN_NVPTX_TRUE
+offload_additional_lib_paths
+offload_additional_options
+PLUGIN_NVPTX_LIBS
+PLUGIN_NVPTX_LDFLAGS
+PLUGIN_NVPTX_CPPFLAGS
+PLUGIN_NVPTX
+CUDA_DRIVER_LIB
+CUDA_DRIVER_INCLUDE
+offload_targets
 libtool_VERSION
 ac_ct_FC
 FCFLAGS
@@ -770,6 +778,9 @@ enable_fast_install
 with_gnu_ld
 enable_libtool_lock
 enable_maintainer_mode
+with_cuda_driver
+with_cuda_driver_include
+with_cuda_driver_lib
 enable_linux_futex
 enable_tls
 enable_symvers
@@ -1431,6 +1442,16 @@ Optional Packages:
   --with-pic              try to use only PIC/non-PIC objects [default=use
                           both]
   --with-gnu-ld           assume the C compiler uses GNU ld [default=no]
+  --with-cuda-driver=PATH specify prefix directory for installed CUDA driver
+                          package. Equivalent to
+                          --with-cuda-driver-include=PATH/include plus
+                          --with-cuda-driver-lib=PATH/lib
+  --with-cuda-driver-include=PATH
+                          specify directory for installed CUDA driver include
+                          files
+  --with-cuda-driver-lib=PATH
+                          specify directory for the installed CUDA driver
+                          library
 
 Some influential environment variables:
   CC          C compiler command
@@ -11097,7 +11118,7 @@ else
   lt_dlunknown=0; lt_dlno_uscore=1; lt_dlneed_uscore=2
   lt_status=$lt_dlunknown
   cat > conftest.$ac_ext <<_LT_EOF
-#line 11100 "configure"
+#line 11121 "configure"
 #include "confdefs.h"
 
 #if HAVE_DLFCN_H
@@ -11203,7 +11224,7 @@ else
   lt_dlunknown=0; lt_dlno_uscore=1; lt_dlneed_uscore=2
   lt_status=$lt_dlunknown
   cat > conftest.$ac_ext <<_LT_EOF
-#line 11206 "configure"
+#line 11227 "configure"
 #include "confdefs.h"
 
 #if HAVE_DLFCN_H
@@ -15055,6 +15076,36 @@ fi
 rm -f core conftest.err conftest.$ac_objext \
     conftest$ac_exeext conftest.$ac_ext
 
+# Plugins for offload execution, configure.ac fragment.  -*- mode: autoconf -*-
+#
+# Copyright (C) 2014-2015 Free Software Foundation, Inc.
+#
+# Contributed by Mentor Embedded.
+#
+# This file is part of the GNU Offloading and Multi Processing Library
+# (libgomp).
+#
+# Libgomp is free software; you can redistribute it and/or modify it
+# under the terms of the GNU General Public License as published by
+# the Free Software Foundation; either version 3, or (at your option)
+# any later version.
+#
+# Libgomp is distributed in the hope that it will be useful, but WITHOUT ANY
+# WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
+# FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+# more details.
+#
+# Under Section 7 of GPL version 3, you are granted additional
+# permissions described in the GCC Runtime Library Exception, version
+# 3.1, as published by the Free Software Foundation.
+#
+# You should have received a copy of the GNU General Public License and
+# a copy of the GCC Runtime Library Exception along with this program;
+# see the files COPYING3 and COPYING.RUNTIME respectively.  If not, see
+# <http://www.gnu.org/licenses/>.
+
+offload_targets=
+
 plugin_support=yes
 { $as_echo "$as_me:${as_lineno-$LINENO}: checking for dlsym in -ldl" >&5
 $as_echo_n "checking for dlsym in -ldl... " >&6; }
@@ -15107,8 +15158,152 @@ if test x"$plugin_support" = xyes; then
 
 $as_echo "#define PLUGIN_SUPPORT 1" >>confdefs.h
 
+  offload_targets=host_nonshm
+elif test "x${enable_offload_targets-no}" != xno; then
+  as_fn_error "Can't support offloading without support for plugins" "$LINENO" 5
+fi
+
+# Look for the CUDA driver package.
+CUDA_DRIVER_INCLUDE=
+CUDA_DRIVER_LIB=
+
+
+CUDA_DRIVER_CPPFLAGS=
+CUDA_DRIVER_LDFLAGS=
+
+# Check whether --with-cuda-driver was given.
+if test "${with_cuda_driver+set}" = set; then :
+  withval=$with_cuda_driver;
 fi
 
+
+# Check whether --with-cuda-driver-include was given.
+if test "${with_cuda_driver_include+set}" = set; then :
+  withval=$with_cuda_driver_include;
+fi
+
+
+# Check whether --with-cuda-driver-lib was given.
+if test "${with_cuda_driver_lib+set}" = set; then :
+  withval=$with_cuda_driver_lib;
+fi
+
+if test "x$with_cuda_driver" != x; then
+  CUDA_DRIVER_INCLUDE=$with_cuda_driver/include
+  CUDA_DRIVER_LIB=$with_cuda_driver/lib
+fi
+if test "x$with_cuda_driver_include" != x; then
+  CUDA_DRIVER_INCLUDE=$with_cuda_driver_include
+fi
+if test "x$with_cuda_driver_lib" != x; then
+  CUDA_DRIVER_LIB=$with_cuda_driver_lib
+fi
+if test "x$CUDA_DRIVER_INCLUDE" != x; then
+  CUDA_DRIVER_CPPFLAGS=-I$CUDA_DRIVER_INCLUDE
+fi
+if test "x$CUDA_DRIVER_LIB" != x; then
+  CUDA_DRIVER_LDFLAGS=-L$CUDA_DRIVER_LIB
+fi
+
+PLUGIN_NVPTX=0
+PLUGIN_NVPTX_CPPFLAGS=
+PLUGIN_NVPTX_LDFLAGS=
+PLUGIN_NVPTX_LIBS=
+
+
+
+
+
+# Get offload targets and path to install tree of offloading compiler.
+offload_additional_options=
+offload_additional_lib_paths=
+
+
+if test x"$enable_offload_targets" != x; then
+  for tgt in `echo $enable_offload_targets | sed -e 's#,# #g'`; do
+    tgt_dir=`echo $tgt | grep '=' | sed 's/.*=//'`
+    tgt=`echo $tgt | sed 's/=.*//'`
+    case $tgt in
+      *-intelmic-* | *-intelmicemul-*)
+	tgt_name=intelmic
+	;;
+      nvptx*)
+        tgt_name=nvptx
+	PLUGIN_NVPTX=$tgt
+	PLUGIN_NVPTX_CPPFLAGS=$CUDA_DRIVER_CPPFLAGS
+	PLUGIN_NVPTX_LDFLAGS=$CUDA_DRIVER_LDFLAGS
+	PLUGIN_NVPTX_LIBS='-lcuda'
+
+	PLUGIN_NVPTX_save_CPPFLAGS=$CPPFLAGS
+	CPPFLAGS="$PLUGIN_NVPTX_CPPFLAGS $CPPFLAGS"
+	PLUGIN_NVPTX_save_LDFLAGS=$LDFLAGS
+	LDFLAGS="$PLUGIN_NVPTX_LDFLAGS $LDFLAGS"
+	PLUGIN_NVPTX_save_LIBS=$LIBS
+	LIBS="$PLUGIN_NVPTX_LIBS $LIBS"
+	cat confdefs.h - <<_ACEOF >conftest.$ac_ext
+/* end confdefs.h.  */
+#include "cuda.h"
+int
+main ()
+{
+CUresult r = cuCtxPushCurrent (NULL);
+  ;
+  return 0;
+}
+_ACEOF
+if ac_fn_c_try_link "$LINENO"; then :
+  PLUGIN_NVPTX=1
+fi
+rm -f core conftest.err conftest.$ac_objext \
+    conftest$ac_exeext conftest.$ac_ext
+	CPPFLAGS=$PLUGIN_NVPTX_save_CPPFLAGS
+	LDFLAGS=$PLUGIN_NVPTX_save_LDFLAGS
+	LIBS=$PLUGIN_NVPTX_save_LIBS
+	case $PLUGIN_NVPTX in
+	  nvptx*)
+	    PLUGIN_NVPTX=0
+	    as_fn_error "CUDA driver package required for nvptx support" "$LINENO" 5
+	    ;;
+	esac
+	;;
+      *)
+	as_fn_error "unknown offload target specified" "$LINENO" 5
+	;;
+    esac
+    if test x"$offload_targets" = x; then
+      offload_targets=$tgt_name
+    else
+      offload_targets=$offload_targets,$tgt_name
+    fi
+    if test x"$tgt_dir" != x; then
+      offload_additional_options="$offload_additional_options -B$tgt_dir/libexec/gcc/\$(target_alias)/\$(gcc_version) -B$tgt_dir/bin"
+      offload_additional_lib_paths="$offload_additional_lib_paths:$tgt_dir/lib64:$tgt_dir/lib:$tgt_dir/lib32"
+    else
+      offload_additional_options="$offload_additional_options -B\$(libexecdir)/gcc/\$(target_alias)/\$(gcc_version) -B\$(bindir)"
+      offload_additional_lib_paths="$offload_additional_lib_paths:$toolexeclibdir"
+    fi
+  done
+fi
+
+cat >>confdefs.h <<_ACEOF
+#define OFFLOAD_TARGETS "$offload_targets"
+_ACEOF
+
+ if test $PLUGIN_NVPTX = 1; then
+  PLUGIN_NVPTX_TRUE=
+  PLUGIN_NVPTX_FALSE='#'
+else
+  PLUGIN_NVPTX_TRUE='#'
+  PLUGIN_NVPTX_FALSE=
+fi
+
+
+cat >>confdefs.h <<_ACEOF
+#define PLUGIN_NVPTX $PLUGIN_NVPTX
+_ACEOF
+
+
+
 # Check for functions needed.
 for ac_func in getloadavg clock_gettime strtoull
 do :
@@ -16241,43 +16436,6 @@ else
   multilib_arg=
 fi
 
-# Get accel target and path to install tree of accel compiler
-offload_additional_options=
-offload_additional_lib_paths=
-offload_targets=
-if test x"$enable_offload_targets" != x; then
-  for tgt in `echo $enable_offload_targets | sed -e 's#,# #g'`; do
-    tgt_dir=`echo $tgt | grep '=' | sed 's/.*=//'`
-    tgt=`echo $tgt | sed 's/=.*//'`
-    case $tgt in
-      *-intelmic-* | *-intelmicemul-*)
-	tgt_name="intelmic" ;;
-      *)
-	as_fn_error "unknown offload target specified" "$LINENO" 5 ;;
-    esac
-    if test x"$offload_targets" = x; then
-      offload_targets=$tgt_name
-    else
-      offload_targets=$offload_targets,$tgt_name
-    fi
-    if test x"$tgt_dir" != x; then
-      offload_additional_options="$offload_additional_options -B$tgt_dir/libexec/gcc/\$(target_alias)/\$(gcc_version) -B$tgt_dir/bin"
-      offload_additional_lib_paths="$offload_additional_lib_paths:$tgt_dir/lib64:$tgt_dir/lib:$tgt_dir/lib32"
-    else
-      offload_additional_options="$offload_additional_options -B\$(libexecdir)/gcc/\$(target_alias)/\$(gcc_version) -B\$(bindir)"
-      offload_additional_lib_paths="$offload_additional_lib_paths:$toolexeclibdir"
-    fi
-  done
-fi
-
-cat >>confdefs.h <<_ACEOF
-#define OFFLOAD_TARGETS "$offload_targets"
-_ACEOF
-
-
-
-
-
 # Set up the set of libraries that we need to link against for libgomp.
 # Note that the GOMP_SELF_SPEC in gcc.c may force -pthread,
 # which will force linkage against -lpthread (or equivalent for the system).
@@ -16395,6 +16553,8 @@ ac_config_files="$ac_config_files omp.h omp_lib.h omp_lib.f90 libgomp_f.h"
 
 ac_config_files="$ac_config_files Makefile testsuite/Makefile libgomp.spec"
 
+ac_config_files="$ac_config_files testsuite/libgomp-test-support.pt.exp:testsuite/libgomp-test-support.exp.in"
+
 cat >confcache <<\_ACEOF
 # This file is a shell script that caches the results of configure
 # tests run on this system so they can be shared between configure
@@ -16520,6 +16680,10 @@ if test -z "${MAINTAINER_MODE_TRUE}" && test -z "${MAINTAINER_MODE_FALSE}"; then
   as_fn_error "conditional \"MAINTAINER_MODE\" was never defined.
 Usually this means the macro was only invoked conditionally." "$LINENO" 5
 fi
+if test -z "${PLUGIN_NVPTX_TRUE}" && test -z "${PLUGIN_NVPTX_FALSE}"; then
+  as_fn_error "conditional \"PLUGIN_NVPTX\" was never defined.
+Usually this means the macro was only invoked conditionally." "$LINENO" 5
+fi
 if test -z "${LIBGOMP_BUILD_VERSIONED_SHLIB_TRUE}" && test -z "${LIBGOMP_BUILD_VERSIONED_SHLIB_FALSE}"; then
   as_fn_error "conditional \"LIBGOMP_BUILD_VERSIONED_SHLIB\" was never defined.
 Usually this means the macro was only invoked conditionally." "$LINENO" 5
@@ -17535,6 +17699,7 @@ do
     "Makefile") CONFIG_FILES="$CONFIG_FILES Makefile" ;;
     "testsuite/Makefile") CONFIG_FILES="$CONFIG_FILES testsuite/Makefile" ;;
     "libgomp.spec") CONFIG_FILES="$CONFIG_FILES libgomp.spec" ;;
+    "testsuite/libgomp-test-support.pt.exp") CONFIG_FILES="$CONFIG_FILES testsuite/libgomp-test-support.pt.exp:testsuite/libgomp-test-support.exp.in" ;;
 
   *) as_fn_error "invalid argument: \`$ac_config_target'" "$LINENO" 5;;
   esac
diff --git a/libgomp/configure.ac b/libgomp/configure.ac
index 8ed1baebcb8..4687b01d5cc 100644
--- a/libgomp/configure.ac
+++ b/libgomp/configure.ac
@@ -193,12 +193,7 @@ AC_LINK_IFELSE(
    [],
    [AC_MSG_ERROR([Pthreads are required to build libgomp])])])
 
-plugin_support=yes
-AC_CHECK_LIB(dl, dlsym, , [plugin_support=no])
-if test x"$plugin_support" = xyes; then
-  AC_DEFINE(PLUGIN_SUPPORT, 1,
-    [Define if all infrastructure, needed for plugins, is supported.])
-fi
+m4_include([plugin/configfrag.ac])
 
 # Check for functions needed.
 AC_CHECK_FUNCS(getloadavg clock_gettime strtoull)
@@ -283,40 +278,6 @@ else
   multilib_arg=
 fi
 
-# Get accel target and path to install tree of accel compiler
-offload_additional_options=
-offload_additional_lib_paths=
-offload_targets=
-if test x"$enable_offload_targets" != x; then
-  for tgt in `echo $enable_offload_targets | sed -e 's#,# #g'`; do
-    tgt_dir=`echo $tgt | grep '=' | sed 's/.*=//'`
-    tgt=`echo $tgt | sed 's/=.*//'`
-    case $tgt in
-      *-intelmic-* | *-intelmicemul-*)
-	tgt_name="intelmic" ;;
-      *)
-	AC_MSG_ERROR([unknown offload target specified]) ;;
-    esac
-    if test x"$offload_targets" = x; then
-      offload_targets=$tgt_name
-    else
-      offload_targets=$offload_targets,$tgt_name
-    fi
-    if test x"$tgt_dir" != x; then
-      offload_additional_options="$offload_additional_options -B$tgt_dir/libexec/gcc/\$(target_alias)/\$(gcc_version) -B$tgt_dir/bin"
-      offload_additional_lib_paths="$offload_additional_lib_paths:$tgt_dir/lib64:$tgt_dir/lib:$tgt_dir/lib32"
-    else
-      offload_additional_options="$offload_additional_options -B\$(libexecdir)/gcc/\$(target_alias)/\$(gcc_version) -B\$(bindir)"
-      offload_additional_lib_paths="$offload_additional_lib_paths:$toolexeclibdir"
-    fi
-  done
-fi
-AC_DEFINE_UNQUOTED(OFFLOAD_TARGETS, "$offload_targets",
-  [Define to hold the list of target names suitable for offloading.])
-AC_SUBST(offload_targets)
-AC_SUBST(offload_additional_options)
-AC_SUBST(offload_additional_lib_paths)
-
 # Set up the set of libraries that we need to link against for libgomp.
 # Note that the GOMP_SELF_SPEC in gcc.c may force -pthread,
 # which will force linkage against -lpthread (or equivalent for the system).
@@ -391,4 +352,5 @@ CFLAGS="$save_CFLAGS"
 
 AC_CONFIG_FILES(omp.h omp_lib.h omp_lib.f90 libgomp_f.h)
 AC_CONFIG_FILES(Makefile testsuite/Makefile libgomp.spec)
+AC_CONFIG_FILES([testsuite/libgomp-test-support.pt.exp:testsuite/libgomp-test-support.exp.in])
 AC_OUTPUT
diff --git a/libgomp/configure.tgt b/libgomp/configure.tgt
index ebd9be97d08..2ef49264e54 100644
--- a/libgomp/configure.tgt
+++ b/libgomp/configure.tgt
@@ -27,7 +27,7 @@ fi
 config_path="posix"
 
 # Check for futex enabled all at once.
-if test $enable_linux_futex = yes; then
+if test x$enable_linux_futex = xyes; then
   case "${target}" in
 
     aarch64*-*-linux*)
diff --git a/libgomp/env.c b/libgomp/env.c
index b05b73a6e20..6b5e963c4ea 100644
--- a/libgomp/env.c
+++ b/libgomp/env.c
@@ -28,6 +28,7 @@
 
 #include "libgomp.h"
 #include "libgomp_f.h"
+#include "oacc-int.h"
 #include <ctype.h>
 #include <stdlib.h>
 #include <stdio.h>
@@ -77,6 +78,9 @@ char *gomp_bind_var_list;
 unsigned long gomp_bind_var_list_len;
 void **gomp_places_list;
 unsigned long gomp_places_list_len;
+int gomp_debug_var;
+char *goacc_device_type;
+int goacc_device_num;
 
 /* Parse the OMP_SCHEDULE environment variable.  */
 
@@ -1012,6 +1016,16 @@ parse_affinity (bool ignore)
   return false;
 }
 
+static void
+parse_acc_device_type (void)
+{
+  const char *env = getenv ("ACC_DEVICE_TYPE");
+
+  if (env && *env != '\0')
+    goacc_device_type = strdup (env);
+  else
+    goacc_device_type = NULL;
+}
 
 static void
 handle_omp_display_env (unsigned long stacksize, int wait_policy)
@@ -1182,6 +1196,7 @@ initialize_env (void)
       gomp_global_icv.thread_limit_var
 	= thread_limit_var > INT_MAX ? UINT_MAX : thread_limit_var;
     }
+  parse_int ("GOMP_DEBUG", &gomp_debug_var, true);
 #ifndef HAVE_SYNC_BUILTINS
   gomp_mutex_init (&gomp_managed_threads_lock);
 #endif
@@ -1272,6 +1287,15 @@ initialize_env (void)
     }
 
   handle_omp_display_env (stacksize, wait_policy);
+
+  /* OpenACC.  */
+
+  if (!parse_int ("ACC_DEVICE_NUM", &goacc_device_num, true))
+    goacc_device_num = 0;
+
+  parse_acc_device_type ();
+
+  goacc_runtime_initialize ();
 }
 
 
diff --git a/libgomp/error.c b/libgomp/error.c
index e61d82f53d9..094c24a38c0 100644
--- a/libgomp/error.c
+++ b/libgomp/error.c
@@ -36,7 +36,26 @@
 #include <stdlib.h>
 
 
-static void
+#undef gomp_vdebug
+void
+gomp_vdebug (int kind __attribute__ ((unused)), const char *msg, va_list list)
+{
+  if (gomp_debug_var)
+    vfprintf (stderr, msg, list);
+}
+
+#undef gomp_debug
+void
+gomp_debug (int kind, const char *msg, ...)
+{
+  va_list list;
+
+  va_start (list, msg);
+  gomp_vdebug (kind, msg, list);
+  va_end (list);
+}
+
+void
 gomp_verror (const char *fmt, va_list list)
 {
   fputs ("\nlibgomp: ", stderr);
@@ -55,13 +74,18 @@ gomp_error (const char *fmt, ...)
 }
 
 void
+gomp_vfatal (const char *fmt, va_list list)
+{
+  gomp_verror (fmt, list);
+  exit (EXIT_FAILURE);
+}
+
+void
 gomp_fatal (const char *fmt, ...)
 {
   va_list list;
 
   va_start (list, fmt);
-  gomp_verror (fmt, list);
+  gomp_vfatal (fmt, list);
   va_end (list);
-
-  exit (EXIT_FAILURE);
 }
diff --git a/libgomp/libgomp-plugin.c b/libgomp/libgomp-plugin.c
new file mode 100644
index 00000000000..ffb22e93862
--- /dev/null
+++ b/libgomp/libgomp-plugin.c
@@ -0,0 +1,80 @@
+/* Copyright (C) 2014-2015 Free Software Foundation, Inc.
+
+   Contributed by Mentor Embedded.
+
+   This file is part of the GNU Offloading and Multi Processing Library
+   (libgomp).
+
+   Libgomp is free software; you can redistribute it and/or modify it
+   under the terms of the GNU General Public License as published by
+   the Free Software Foundation; either version 3, or (at your option)
+   any later version.
+
+   Libgomp is distributed in the hope that it will be useful, but WITHOUT ANY
+   WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
+   FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+   more details.
+
+   Under Section 7 of GPL version 3, you are granted additional
+   permissions described in the GCC Runtime Library Exception, version
+   3.1, as published by the Free Software Foundation.
+
+   You should have received a copy of the GNU General Public License and
+   a copy of the GCC Runtime Library Exception along with this program;
+   see the files COPYING3 and COPYING.RUNTIME respectively.  If not, see
+   <http://www.gnu.org/licenses/>.  */
+
+/* Exported (non-hidden) functions exposing libgomp interface for plugins.  */
+
+#include <stdlib.h>
+
+#include "libgomp.h"
+#include "libgomp-plugin.h"
+
+void *
+GOMP_PLUGIN_malloc (size_t size)
+{
+  return gomp_malloc (size);
+}
+
+void *
+GOMP_PLUGIN_malloc_cleared (size_t size)
+{
+  return gomp_malloc_cleared (size);
+}
+
+void *
+GOMP_PLUGIN_realloc (void *ptr, size_t size)
+{
+  return gomp_realloc (ptr, size);
+}
+
+void
+GOMP_PLUGIN_debug (int kind, const char *msg, ...)
+{
+  va_list ap;
+
+  va_start (ap, msg);
+  gomp_debug (kind, msg, ap);
+  va_end (ap);
+}
+
+void
+GOMP_PLUGIN_error (const char *msg, ...)
+{
+  va_list ap;
+
+  va_start (ap, msg);
+  gomp_verror (msg, ap);
+  va_end (ap);
+}
+
+void
+GOMP_PLUGIN_fatal (const char *msg, ...)
+{
+  va_list ap;
+
+  va_start (ap, msg);
+  gomp_vfatal (msg, ap);
+  va_end (ap);
+}
diff --git a/libgomp/libgomp-plugin.h b/libgomp/libgomp-plugin.h
new file mode 100644
index 00000000000..d9cbff5fe71
--- /dev/null
+++ b/libgomp/libgomp-plugin.h
@@ -0,0 +1,80 @@
+/* Copyright (C) 2014-2015 Free Software Foundation, Inc.
+
+   Contributed by Mentor Embedded.
+
+   This file is part of the GNU Offloading and Multi Processing Library
+   (libgomp).
+
+   Libgomp is free software; you can redistribute it and/or modify it
+   under the terms of the GNU General Public License as published by
+   the Free Software Foundation; either version 3, or (at your option)
+   any later version.
+
+   Libgomp is distributed in the hope that it will be useful, but WITHOUT ANY
+   WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
+   FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+   more details.
+
+   Under Section 7 of GPL version 3, you are granted additional
+   permissions described in the GCC Runtime Library Exception, version
+   3.1, as published by the Free Software Foundation.
+
+   You should have received a copy of the GNU General Public License and
+   a copy of the GCC Runtime Library Exception along with this program;
+   see the files COPYING3 and COPYING.RUNTIME respectively.  If not, see
+   <http://www.gnu.org/licenses/>.  */
+
+/* An interface to various libgomp-internal functions for use by plugins.  */
+
+#ifndef LIBGOMP_PLUGIN_H
+#define LIBGOMP_PLUGIN_H 1
+
+#include <stddef.h>
+#include <stdint.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/* Capabilities of offloading devices.  */
+#define GOMP_OFFLOAD_CAP_SHARED_MEM	(1 << 0)
+#define GOMP_OFFLOAD_CAP_NATIVE_EXEC	(1 << 1)
+#define GOMP_OFFLOAD_CAP_OPENMP_400	(1 << 2)
+#define GOMP_OFFLOAD_CAP_OPENACC_200	(1 << 3)
+
+/* Type of offload target device.  Keep in sync with include/gomp-constants.h.  */
+enum offload_target_type
+{
+  OFFLOAD_TARGET_TYPE_HOST = 2,
+  OFFLOAD_TARGET_TYPE_HOST_NONSHM = 3,
+  OFFLOAD_TARGET_TYPE_NVIDIA_PTX = 5,
+  OFFLOAD_TARGET_TYPE_INTEL_MIC = 6
+};
+
+/* Auxiliary struct, used for transferring a host-target address range mapping
+   from plugin to libgomp.  */
+struct mapping_table
+{
+  uintptr_t host_start;
+  uintptr_t host_end;
+  uintptr_t tgt_start;
+  uintptr_t tgt_end;
+};
+
+/* Miscellaneous functions.  */
+extern void *GOMP_PLUGIN_malloc (size_t) __attribute__ ((malloc));
+extern void *GOMP_PLUGIN_malloc_cleared (size_t) __attribute__ ((malloc));
+extern void *GOMP_PLUGIN_realloc (void *, size_t);
+
+extern void GOMP_PLUGIN_debug (int, const char *, ...)
+	__attribute__ ((format (printf, 2, 3)));
+extern void GOMP_PLUGIN_error (const char *, ...)
+	__attribute__ ((format (printf, 1, 2)));
+extern void GOMP_PLUGIN_fatal (const char *, ...)
+	__attribute__ ((noreturn, format (printf, 1, 2)));
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif
diff --git a/libgomp/libgomp.h b/libgomp/libgomp.h
index 05f34967242..3089401c47f 100644
--- a/libgomp/libgomp.h
+++ b/libgomp/libgomp.h
@@ -24,9 +24,10 @@
    <http://www.gnu.org/licenses/>.  */
 
 /* This file contains data types and function declarations that are not
-   part of the official OpenMP user interface.  There are declarations
-   in here that are part of the GNU OpenMP ABI, in that the compiler is
-   required to know about them and use them.
+   part of the official OpenACC or OpenMP user interfaces.  There are
+   declarations in here that are part of the GNU Offloading and Multi
+   Processing ABI, in that the compiler is required to know about them
+   and use them.
 
    The convention is that the all caps prefix "GOMP" is used group items
    that are part of the external ABI, and the lower case prefix "gomp"
@@ -37,10 +38,12 @@
 
 #include "config.h"
 #include "gstdint.h"
+#include "libgomp-plugin.h"
 
 #include <pthread.h>
 #include <stdbool.h>
 #include <stdlib.h>
+#include <stdarg.h>
 
 #ifdef HAVE_ATTRIBUTE_VISIBILITY
 # pragma GCC visibility push(hidden)
@@ -221,6 +224,7 @@ struct gomp_team_state
 };
 
 struct target_mem_desc;
+struct gomp_memory_mapping;
 
 /* These are the OpenMP 4.0 Internal Control Variables described in
    section 2.3.1.  Those described as having one copy per task are
@@ -254,6 +258,9 @@ extern char *gomp_bind_var_list;
 extern unsigned long gomp_bind_var_list_len;
 extern void **gomp_places_list;
 extern unsigned long gomp_places_list_len;
+extern int gomp_debug_var;
+extern int goacc_device_num;
+extern char *goacc_device_type;
 
 enum gomp_task_kind
 {
@@ -533,10 +540,26 @@ extern void *gomp_realloc (void *, size_t);
 
 /* error.c */
 
+extern void gomp_vdebug (int, const char *, va_list);
+extern void gomp_debug (int, const char *, ...)
+	__attribute__ ((format (printf, 2, 3)));
+#define gomp_vdebug(KIND, FMT, VALIST) \
+  do { \
+    if (__builtin_expect (gomp_debug_var, 0)) \
+      (gomp_vdebug) ((KIND), (FMT), (VALIST)); \
+  } while (0)
+#define gomp_debug(KIND, ...) \
+  do { \
+    if (__builtin_expect (gomp_debug_var, 0)) \
+      (gomp_debug) ((KIND), __VA_ARGS__); \
+  } while (0)
+extern void gomp_verror (const char *, va_list);
 extern void gomp_error (const char *, ...)
-	__attribute__((format (printf, 1, 2)));
+	__attribute__ ((format (printf, 1, 2)));
+extern void gomp_vfatal (const char *, va_list)
+	__attribute__ ((noreturn));
 extern void gomp_fatal (const char *, ...)
-	__attribute__((noreturn, format (printf, 1, 2)));
+	__attribute__ ((noreturn, format (printf, 1, 2)));
 
 /* iter.c */
 
@@ -607,8 +630,192 @@ extern void gomp_free_thread (void *);
 
 /* target.c */
 
+extern void gomp_init_targets_once (void);
 extern int gomp_get_num_devices (void);
 
+typedef struct splay_tree_node_s *splay_tree_node;
+typedef struct splay_tree_s *splay_tree;
+typedef struct splay_tree_key_s *splay_tree_key;
+
+struct target_mem_desc {
+  /* Reference count.  */
+  uintptr_t refcount;
+  /* All the splay nodes allocated together.  */
+  splay_tree_node array;
+  /* Start of the target region.  */
+  uintptr_t tgt_start;
+  /* End of the targer region.  */
+  uintptr_t tgt_end;
+  /* Handle to free.  */
+  void *to_free;
+  /* Previous target_mem_desc.  */
+  struct target_mem_desc *prev;
+  /* Number of items in following list.  */
+  size_t list_count;
+
+  /* Corresponding target device descriptor.  */
+  struct gomp_device_descr *device_descr;
+
+  /* Memory mapping info for the thread that created this descriptor.  */
+  struct gomp_memory_mapping *mem_map;
+
+  /* List of splay keys to remove (or decrease refcount)
+     at the end of region.  */
+  splay_tree_key list[];
+};
+
+struct splay_tree_key_s {
+  /* Address of the host object.  */
+  uintptr_t host_start;
+  /* Address immediately after the host object.  */
+  uintptr_t host_end;
+  /* Descriptor of the target memory.  */
+  struct target_mem_desc *tgt;
+  /* Offset from tgt->tgt_start to the start of the target object.  */
+  uintptr_t tgt_offset;
+  /* Reference count.  */
+  uintptr_t refcount;
+  /* Asynchronous reference count.  */
+  uintptr_t async_refcount;
+  /* True if data should be copied from device to host at the end.  */
+  bool copy_from;
+};
+
+#include "splay-tree.h"
+
+/* Information about mapped memory regions (per device/context).  */
+
+struct gomp_memory_mapping
+{
+  /* Mutex for operating with the splay tree and other shared structures.  */
+  gomp_mutex_t lock;
+
+  /* True when tables have been added to this memory map.  */
+  bool is_initialized;
+
+  /* Splay tree containing information about mapped memory regions.  */
+  struct splay_tree_s splay_tree;
+};
+
+typedef struct acc_dispatch_t
+{
+  /* This is a linked list of data mapped using the
+     acc_map_data/acc_unmap_data or "acc enter data"/"acc exit data" pragmas.
+     Unlike mapped_data in the goacc_thread struct, unmapping can
+     happen out-of-order with respect to mapping.  */
+  /* This is guarded by the lock in the "outer" struct gomp_device_descr.  */
+  struct target_mem_desc *data_environ;
+
+  /* Extra information required for a device instance by a given target.  */
+  /* This is guarded by the lock in the "outer" struct gomp_device_descr.  */
+  void *target_data;
+
+  /* Open or close a device instance.  */
+  void *(*open_device_func) (int n);
+  int (*close_device_func) (void *h);
+
+  /* Set or get the device number.  */
+  int (*get_device_num_func) (void);
+  void (*set_device_num_func) (int);
+
+  /* Execute.  */
+  void (*exec_func) (void (*) (void *), size_t, void **, void **, size_t *,
+		     unsigned short *, int, int, int, int, void *);
+
+  /* Async cleanup callback registration.  */
+  void (*register_async_cleanup_func) (void *);
+
+  /* Asynchronous routines.  */
+  int (*async_test_func) (int);
+  int (*async_test_all_func) (void);
+  void (*async_wait_func) (int);
+  void (*async_wait_async_func) (int, int);
+  void (*async_wait_all_func) (void);
+  void (*async_wait_all_async_func) (int);
+  void (*async_set_async_func) (int);
+
+  /* Create/destroy TLS data.  */
+  void *(*create_thread_data_func) (void *);
+  void (*destroy_thread_data_func) (void *);
+
+  /* NVIDIA target specific routines.  */
+  struct {
+    void *(*get_current_device_func) (void);
+    void *(*get_current_context_func) (void);
+    void *(*get_stream_func) (int);
+    int (*set_stream_func) (int, void *);
+  } cuda;
+} acc_dispatch_t;
+
+/* This structure describes accelerator device.
+   It contains name of the corresponding libgomp plugin, function handlers for
+   interaction with the device, ID-number of the device, and information about
+   mapped memory.  */
+struct gomp_device_descr
+{
+  /* Immutable data, which is only set during initialization, and which is not
+     guarded by the lock.  */
+
+  /* The name of the device.  */
+  const char *name;
+
+  /* Capabilities of device (supports OpenACC, OpenMP).  */
+  unsigned int capabilities;
+
+  /* This is the ID number of device among devices of the same type.  */
+  int target_id;
+
+  /* This is the TYPE of device.  */
+  enum offload_target_type type;
+
+  /* Function handlers.  */
+  const char *(*get_name_func) (void);
+  unsigned int (*get_caps_func) (void);
+  int (*get_type_func) (void);
+  int (*get_num_devices_func) (void);
+  void (*register_image_func) (void *, void *);
+  void (*init_device_func) (int);
+  void (*fini_device_func) (int);
+  int (*get_table_func) (int, struct mapping_table **);
+  void *(*alloc_func) (int, size_t);
+  void (*free_func) (int, void *);
+  void *(*dev2host_func) (int, void *, const void *, size_t);
+  void *(*host2dev_func) (int, void *, const void *, size_t);
+  void (*run_func) (int, void *, void *);
+
+  /* Memory-mapping info for this device instance.  */
+  /* Uses a separate lock.  */
+  struct gomp_memory_mapping mem_map;
+
+  /* Mutex for the mutable data.  */
+  gomp_mutex_t lock;
+
+  /* Set to true when device is initialized.  */
+  bool is_initialized;
+
+  /* True when offload regions have been registered with this device.  */
+  bool offload_regions_registered;
+
+  /* OpenACC-specific data and functions.  */
+  /* This is mutable because of its mutable data_environ and target_data
+     members.  */
+  acc_dispatch_t openacc;
+};
+
+extern void gomp_acc_insert_pointer (size_t, void **, size_t *, void *);
+extern void gomp_acc_remove_pointer (void *, bool, int, int);
+
+extern struct target_mem_desc *gomp_map_vars (struct gomp_device_descr *,
+					      size_t, void **, void **,
+					      size_t *, void *, bool, bool);
+extern void gomp_copy_from_async (struct target_mem_desc *);
+extern void gomp_unmap_vars (struct target_mem_desc *, bool);
+extern void gomp_init_device (struct gomp_device_descr *);
+extern void gomp_init_tables (struct gomp_device_descr *,
+			      struct gomp_memory_mapping *);
+extern void gomp_free_memmap (struct gomp_memory_mapping *);
+extern void gomp_fini_device (struct gomp_device_descr *);
+
 /* work.c */
 
 extern void gomp_init_work_share (struct gomp_work_share *, bool, unsigned);
diff --git a/libgomp/libgomp.map b/libgomp/libgomp.map
index f36df23e795..f44174e83b2 100644
--- a/libgomp/libgomp.map
+++ b/libgomp/libgomp.map
@@ -232,3 +232,107 @@ GOMP_4.0.1 {
   global:
 	GOMP_offload_register;
 } GOMP_4.0;
+
+OACC_2.0 {
+  global:
+	acc_get_num_devices;
+	acc_get_num_devices_h_;
+	acc_set_device_type;
+	acc_set_device_type_h_;
+	acc_get_device_type;
+	acc_get_device_type_h_;
+	acc_set_device_num;
+	acc_set_device_num_h_;
+	acc_get_device_num;
+	acc_get_device_num_h_;
+	acc_async_test;
+	acc_async_test_h_;
+	acc_async_test_all;
+	acc_async_test_all_h_;
+	acc_wait;
+	acc_wait_h_;
+	acc_wait_async;
+	acc_wait_async_h_;
+	acc_wait_all;
+	acc_wait_all_h_;
+	acc_wait_all_async;
+	acc_wait_all_async_h_;
+	acc_init;
+	acc_init_h_;
+	acc_shutdown;
+	acc_shutdown_h_;
+	acc_on_device;
+	acc_on_device_h_;
+	acc_malloc;
+	acc_free;
+	acc_copyin;
+	acc_copyin_32_h_;
+	acc_copyin_64_h_;
+	acc_copyin_array_h_;
+	acc_present_or_copyin;
+	acc_present_or_copyin_32_h_;
+	acc_present_or_copyin_64_h_;
+	acc_present_or_copyin_array_h_;
+	acc_create;
+	acc_create_32_h_;
+	acc_create_64_h_;
+	acc_create_array_h_;
+	acc_present_or_create;
+	acc_present_or_create_32_h_;
+	acc_present_or_create_64_h_;
+	acc_present_or_create_array_h_;
+	acc_copyout;
+	acc_copyout_32_h_;
+	acc_copyout_64_h_;
+	acc_copyout_array_h_;
+	acc_delete;
+	acc_delete_32_h_;
+	acc_delete_64_h_;
+	acc_delete_array_h_;
+	acc_update_device;
+	acc_update_device_32_h_;
+	acc_update_device_64_h_;
+	acc_update_device_array_h_;
+	acc_update_self;
+	acc_update_self_32_h_;
+	acc_update_self_64_h_;
+	acc_update_self_array_h_;
+	acc_map_data;
+	acc_unmap_data;
+	acc_deviceptr;
+	acc_hostptr;
+	acc_is_present;
+	acc_is_present_32_h_;
+	acc_is_present_64_h_;
+	acc_is_present_array_h_;
+	acc_memcpy_to_device;
+	acc_memcpy_from_device;
+	acc_get_current_cuda_device;
+	acc_get_current_cuda_context;
+	acc_get_cuda_stream;
+	acc_set_cuda_stream;
+};
+
+GOACC_2.0 {
+  global:
+	GOACC_data_end;
+	GOACC_data_start;
+	GOACC_enter_exit_data;
+	GOACC_parallel;
+	GOACC_update;
+	GOACC_wait;
+	GOACC_get_thread_num;
+	GOACC_get_num_threads;
+};
+
+GOMP_PLUGIN_1.0 {
+  global:
+	GOMP_PLUGIN_malloc;
+	GOMP_PLUGIN_malloc_cleared;
+	GOMP_PLUGIN_realloc;
+	GOMP_PLUGIN_debug;
+	GOMP_PLUGIN_error;
+	GOMP_PLUGIN_fatal;
+	GOMP_PLUGIN_async_unmap_vars;
+	GOMP_PLUGIN_acc_thread;
+};
diff --git a/libgomp/libgomp.texi b/libgomp/libgomp.texi
index b7306f14c6f..6c7f1aed641 100644
--- a/libgomp/libgomp.texi
+++ b/libgomp/libgomp.texi
@@ -35,8 +35,9 @@ texts being (a) (see below), and with the Back-Cover Texts being (b)
 @end direntry
 
 This manual documents libgomp, the GNU Offloading and Multi Processing
-Runtime library.  This is the GNU implementation of the OpenMP API for
-multi-platform shared-memory parallel programming in C/C++ and Fortran.
+Runtime library.  This is the GNU implementation of the OpenMP and
+OpenACC APIs for parallel and accelerator programming in C/C++ and
+Fortran.
 
 Published by the Free Software Foundation
 51 Franklin Street, Fifth Floor
@@ -50,7 +51,7 @@ Boston, MA 02110-1301 USA
 
 @titlepage
 @title GNU Offloading and Multi Processing Runtime Library
-@subtitle The GNU OpenMP Implementation
+@subtitle The GNU OpenMP and OpenACC Implementation
 @page
 @vskip 0pt plus 1filll
 @comment For the @value{version-GCC} Version*
@@ -72,11 +73,18 @@ Boston, MA 02110-1301, USA@*
 @cindex Introduction
 
 This manual documents the usage of libgomp, the GNU Offloading and
-Multi Processing Runtime Library.  This is the GNU implementation of the
-@uref{http://www.openmp.org, OpenMP} Application Programming Interface (API)
-for multi-platform shared-memory parallel programming in C/C++ and Fortran.
+Multi Processing Runtime Library.  This includes the GNU
+implementation of the @uref{http://www.openmp.org, OpenMP} Application
+Programming Interface (API) for multi-platform shared-memory parallel
+programming in C/C++ and Fortran, and the GNU implementation of the
+@uref{http://www.openacc.org/, OpenACC} Application Programming
+Interface (API) for offloading of code to accelerator devices in C/C++
+and Fortran.
 
-Originally, libgomp was known as the GNU OpenMP Runtime Library.
+Originally, libgomp implemented the GNU OpenMP Runtime Library.  Based
+on this, support for OpenACC and offloading (both OpenACC and OpenMP
+4's target construct) has been added later on, and the library's name
+changed to GNU Offloading and Multi Processing Runtime Library.
 
 
 
@@ -1312,6 +1320,7 @@ beginning with @env{GOMP_} are GNU extensions.
 * OMP_THREAD_LIMIT::      Set the maximum number of threads
 * OMP_WAIT_POLICY::       How waiting threads are handled
 * GOMP_CPU_AFFINITY::     Bind threads to specific CPUs
+* GOMP_DEBUG::            Enable debugging output
 * GOMP_STACKSIZE::        Set default thread stack size
 * GOMP_SPINCOUNT::        Set the busy-wait spin count
 @end menu
@@ -1631,6 +1640,20 @@ If both @env{GOMP_CPU_AFFINITY} and @env{OMP_PROC_BIND} are set,
 
 
 
+@node GOMP_DEBUG
+@section @env{GOMP_DEBUG} -- Enable debugging output
+@cindex Environment Variable
+@table @asis
+@item @emph{Description}:
+Enable debugging output.  The variable should be set to @code{0}
+(disabled, also the default if not set), or @code{1} (enabled).
+
+If enabled, some debugging output will be printed during execution.
+This is currently not specified in more detail, and subject to change.
+@end table
+
+
+
 @node GOMP_STACKSIZE
 @section @env{GOMP_STACKSIZE} -- Set default thread stack size
 @cindex Environment Variable
@@ -2080,7 +2103,8 @@ becomes
 
 Bugs in the GNU Offloading and Multi Processing Runtime Library should
 be reported via @uref{http://gcc.gnu.org/bugzilla/, Bugzilla}.  Please add
-"openmp" to the keywords field in the bug report.
+"openacc", or "openmp", or both to the keywords field in the bug
+report, as appropriate.
 
 
 
diff --git a/libgomp/libgomp_g.h b/libgomp/libgomp_g.h
index 56a4a97d1d5..c1e4e6367ab 100644
--- a/libgomp/libgomp_g.h
+++ b/libgomp/libgomp_g.h
@@ -215,4 +215,20 @@ extern void GOMP_target_update (int, const void *,
 				size_t, void **, size_t *, unsigned char *);
 extern void GOMP_teams (unsigned int, unsigned int);
 
+/* oacc-parallel.c */
+
+extern void GOACC_data_start (int, const void *,
+			      size_t, void **, size_t *, unsigned short *);
+extern void GOACC_data_end (void);
+extern void GOACC_enter_exit_data (int, const void *, size_t, void **,
+				   size_t *, unsigned short *, int, int, ...);
+extern void GOACC_parallel (int, void (*) (void *), const void *, size_t,
+			    void **, size_t *, unsigned short *, int, int, int,
+			    int, int, ...);
+extern void GOACC_update (int, const void *, size_t, void **, size_t *,
+			  unsigned short *, int, int, ...);
+extern void GOACC_wait (int, int, ...);
+extern int GOACC_get_num_threads (void);
+extern int GOACC_get_thread_num (void);
+
 #endif /* LIBGOMP_G_H */
diff --git a/libgomp/oacc-async.c b/libgomp/oacc-async.c
new file mode 100644
index 00000000000..08b7c5e1945
--- /dev/null
+++ b/libgomp/oacc-async.c
@@ -0,0 +1,77 @@
+/* OpenACC Runtime Library Definitions.
+
+   Copyright (C) 2013-2015 Free Software Foundation, Inc.
+
+   Contributed by Mentor Embedded.
+
+   This file is part of the GNU Offloading and Multi Processing Library
+   (libgomp).
+
+   Libgomp is free software; you can redistribute it and/or modify it
+   under the terms of the GNU General Public License as published by
+   the Free Software Foundation; either version 3, or (at your option)
+   any later version.
+
+   Libgomp is distributed in the hope that it will be useful, but WITHOUT ANY
+   WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
+   FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+   more details.
+
+   Under Section 7 of GPL version 3, you are granted additional
+   permissions described in the GCC Runtime Library Exception, version
+   3.1, as published by the Free Software Foundation.
+
+   You should have received a copy of the GNU General Public License and
+   a copy of the GCC Runtime Library Exception along with this program;
+   see the files COPYING3 and COPYING.RUNTIME respectively.  If not, see
+   <http://www.gnu.org/licenses/>.  */
+
+
+#include "openacc.h"
+#include "libgomp.h"
+#include "oacc-int.h"
+
+int
+acc_async_test (int async)
+{
+  if (async < acc_async_sync)
+    gomp_fatal ("invalid async argument: %d", async);
+
+  return base_dev->openacc.async_test_func (async);
+}
+
+int
+acc_async_test_all (void)
+{
+  return base_dev->openacc.async_test_all_func ();
+}
+
+void
+acc_wait (int async)
+{
+  if (async < acc_async_sync)
+    gomp_fatal ("invalid async argument: %d", async);
+
+  base_dev->openacc.async_wait_func (async);
+}
+
+void
+acc_wait_async (int async1, int async2)
+{
+  base_dev->openacc.async_wait_async_func (async1, async2);
+}
+
+void
+acc_wait_all (void)
+{
+  base_dev->openacc.async_wait_all_func ();
+}
+
+void
+acc_wait_all_async (int async)
+{
+  if (async < acc_async_sync)
+    gomp_fatal ("invalid async argument: %d", async);
+
+  base_dev->openacc.async_wait_all_async_func (async);
+}
diff --git a/libgomp/oacc-cuda.c b/libgomp/oacc-cuda.c
new file mode 100644
index 00000000000..c8ef376e3a2
--- /dev/null
+++ b/libgomp/oacc-cuda.c
@@ -0,0 +1,84 @@
+/* OpenACC Runtime Library: CUDA support glue.
+
+   Copyright (C) 2014-2015 Free Software Foundation, Inc.
+
+   Contributed by Mentor Embedded.
+
+   This file is part of the GNU Offloading and Multi Processing Library
+   (libgomp).
+
+   Libgomp is free software; you can redistribute it and/or modify it
+   under the terms of the GNU General Public License as published by
+   the Free Software Foundation; either version 3, or (at your option)
+   any later version.
+
+   Libgomp is distributed in the hope that it will be useful, but WITHOUT ANY
+   WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
+   FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+   more details.
+
+   Under Section 7 of GPL version 3, you are granted additional
+   permissions described in the GCC Runtime Library Exception, version
+   3.1, as published by the Free Software Foundation.
+
+   You should have received a copy of the GNU General Public License and
+   a copy of the GCC Runtime Library Exception along with this program;
+   see the files COPYING3 and COPYING.RUNTIME respectively.  If not, see
+   <http://www.gnu.org/licenses/>.  */
+
+#include "openacc.h"
+#include "config.h"
+#include "libgomp.h"
+#include "oacc-int.h"
+
+void *
+acc_get_current_cuda_device (void)
+{
+  void *p = NULL;
+
+  if (base_dev && base_dev->openacc.cuda.get_current_device_func)
+    p = base_dev->openacc.cuda.get_current_device_func ();
+
+  return p;
+}
+
+void *
+acc_get_current_cuda_context (void)
+{
+  void *p = NULL;
+
+  if (base_dev && base_dev->openacc.cuda.get_current_context_func)
+    p = base_dev->openacc.cuda.get_current_context_func ();
+
+  return p;
+}
+
+void *
+acc_get_cuda_stream (int async)
+{
+  void *p = NULL;
+
+  if (async < 0)
+    return p;
+
+  if (base_dev && base_dev->openacc.cuda.get_stream_func)
+    p = base_dev->openacc.cuda.get_stream_func (async);
+
+  return p;
+}
+
+int
+acc_set_cuda_stream (int async, void *stream)
+{
+  int s = -1;
+
+  if (async < 0 || stream == NULL)
+    return 0;
+
+  goacc_lazy_initialize ();
+
+  if (base_dev && base_dev->openacc.cuda.set_stream_func)
+    s = base_dev->openacc.cuda.set_stream_func (async, stream);
+
+  return s;
+}
diff --git a/libgomp/oacc-host.c b/libgomp/oacc-host.c
new file mode 100644
index 00000000000..6aeb1e765d5
--- /dev/null
+++ b/libgomp/oacc-host.c
@@ -0,0 +1,100 @@
+/* OpenACC Runtime Library: acc_device_host.
+
+   Copyright (C) 2013-2015 Free Software Foundation, Inc.
+
+   Contributed by Mentor Embedded.
+
+   This file is part of the GNU Offloading and Multi Processing Library
+   (libgomp).
+
+   Libgomp is free software; you can redistribute it and/or modify it
+   under the terms of the GNU General Public License as published by
+   the Free Software Foundation; either version 3, or (at your option)
+   any later version.
+
+   Libgomp is distributed in the hope that it will be useful, but WITHOUT ANY
+   WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
+   FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+   more details.
+
+   Under Section 7 of GPL version 3, you are granted additional
+   permissions described in the GCC Runtime Library Exception, version
+   3.1, as published by the Free Software Foundation.
+
+   You should have received a copy of the GNU General Public License and
+   a copy of the GCC Runtime Library Exception along with this program;
+   see the files COPYING3 and COPYING.RUNTIME respectively.  If not, see
+   <http://www.gnu.org/licenses/>.  */
+
+/* This shares much of the implementation of the plugin-host.c "host_nonshm"
+   plugin.  */
+#include "plugin/plugin-host.c"
+
+static struct gomp_device_descr host_dispatch =
+  {
+    .name = "host",
+    .capabilities = (GOMP_OFFLOAD_CAP_OPENACC_200
+		     | GOMP_OFFLOAD_CAP_NATIVE_EXEC
+		     | GOMP_OFFLOAD_CAP_SHARED_MEM),
+    .target_id = 0,
+    .type = OFFLOAD_TARGET_TYPE_HOST,
+
+    .get_name_func = GOMP_OFFLOAD_get_name,
+    .get_caps_func = GOMP_OFFLOAD_get_caps,
+    .get_type_func = GOMP_OFFLOAD_get_type,
+    .get_num_devices_func = GOMP_OFFLOAD_get_num_devices,
+    .register_image_func = GOMP_OFFLOAD_register_image,
+    .init_device_func = GOMP_OFFLOAD_init_device,
+    .fini_device_func = GOMP_OFFLOAD_fini_device,
+    .get_table_func = GOMP_OFFLOAD_get_table,
+    .alloc_func = GOMP_OFFLOAD_alloc,
+    .free_func = GOMP_OFFLOAD_free,
+    .dev2host_func = GOMP_OFFLOAD_dev2host,
+    .host2dev_func = GOMP_OFFLOAD_host2dev,
+    .run_func = GOMP_OFFLOAD_run,
+
+    .mem_map.is_initialized = false,
+    .mem_map.splay_tree.root = NULL,
+    .is_initialized = false,
+    .offload_regions_registered = false,
+
+    .openacc = {
+      .open_device_func = GOMP_OFFLOAD_openacc_open_device,
+      .close_device_func = GOMP_OFFLOAD_openacc_close_device,
+
+      .get_device_num_func = GOMP_OFFLOAD_openacc_get_device_num,
+      .set_device_num_func = GOMP_OFFLOAD_openacc_set_device_num,
+
+      .exec_func = GOMP_OFFLOAD_openacc_parallel,
+
+      .register_async_cleanup_func
+        = GOMP_OFFLOAD_openacc_register_async_cleanup,
+
+      .async_set_async_func = GOMP_OFFLOAD_openacc_async_set_async,
+      .async_test_func = GOMP_OFFLOAD_openacc_async_test,
+      .async_test_all_func = GOMP_OFFLOAD_openacc_async_test_all,
+      .async_wait_func = GOMP_OFFLOAD_openacc_async_wait,
+      .async_wait_async_func = GOMP_OFFLOAD_openacc_async_wait_async,
+      .async_wait_all_func = GOMP_OFFLOAD_openacc_async_wait_all,
+      .async_wait_all_async_func = GOMP_OFFLOAD_openacc_async_wait_all_async,
+
+      .create_thread_data_func = GOMP_OFFLOAD_openacc_create_thread_data,
+      .destroy_thread_data_func = GOMP_OFFLOAD_openacc_destroy_thread_data,
+
+      .cuda = {
+	.get_current_device_func = NULL,
+	.get_current_context_func = NULL,
+	.get_stream_func = NULL,
+	.set_stream_func = NULL,
+      }
+    }
+  };
+
+/* Register this device type.  */
+static __attribute__ ((constructor))
+void goacc_host_init (void)
+{
+  gomp_mutex_init (&host_dispatch.mem_map.lock);
+  gomp_mutex_init (&host_dispatch.lock);
+  goacc_register (&host_dispatch);
+}
diff --git a/libgomp/oacc-init.c b/libgomp/oacc-init.c
new file mode 100644
index 00000000000..166eb553a6c
--- /dev/null
+++ b/libgomp/oacc-init.c
@@ -0,0 +1,636 @@
+/* OpenACC Runtime initialization routines
+
+   Copyright (C) 2013-2015 Free Software Foundation, Inc.
+
+   Contributed by Mentor Embedded.
+
+   This file is part of the GNU Offloading and Multi Processing Library
+   (libgomp).
+
+   Libgomp is free software; you can redistribute it and/or modify it
+   under the terms of the GNU General Public License as published by
+   the Free Software Foundation; either version 3, or (at your option)
+   any later version.
+
+   Libgomp is distributed in the hope that it will be useful, but WITHOUT ANY
+   WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
+   FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+   more details.
+
+   Under Section 7 of GPL version 3, you are granted additional
+   permissions described in the GCC Runtime Library Exception, version
+   3.1, as published by the Free Software Foundation.
+
+   You should have received a copy of the GNU General Public License and
+   a copy of the GCC Runtime Library Exception along with this program;
+   see the files COPYING3 and COPYING.RUNTIME respectively.  If not, see
+   <http://www.gnu.org/licenses/>.  */
+
+#include "libgomp.h"
+#include "oacc-int.h"
+#include "openacc.h"
+#include <assert.h>
+#include <stdlib.h>
+#include <strings.h>
+#include <stdbool.h>
+#include <string.h>
+
+static gomp_mutex_t acc_device_lock;
+
+/* The dispatch table for the current accelerator device.  This is global, so
+   you can only have one type of device open at any given time in a program.
+   This is the "base" device in that several devices that use the same
+   dispatch table may be active concurrently: this one (the "zeroth") is used
+   for overall initialisation/shutdown, and other instances -- not necessarily
+   including this one -- may be opened and closed once the base device has
+   been initialized.  */
+struct gomp_device_descr *base_dev;
+
+#if defined HAVE_TLS || defined USE_EMUTLS
+__thread struct goacc_thread *goacc_tls_data;
+#else
+pthread_key_t goacc_tls_key;
+#endif
+static pthread_key_t goacc_cleanup_key;
+
+/* Current dispatcher, and how it was initialized */
+static acc_device_t init_key = _ACC_device_hwm;
+
+static struct goacc_thread *goacc_threads;
+static gomp_mutex_t goacc_thread_lock;
+
+/* An array of dispatchers for device types, indexed by the type.  This array
+   only references "base" devices, and other instances of the same type are
+   found by simply indexing from each such device (which are stored linearly,
+   grouped by device in target.c:devices).  */
+static struct gomp_device_descr *dispatchers[_ACC_device_hwm] = { 0 };
+
+attribute_hidden void
+goacc_register (struct gomp_device_descr *disp)
+{
+  /* Only register the 0th device here.  */
+  if (disp->target_id != 0)
+    return;
+
+  gomp_mutex_lock (&acc_device_lock);
+
+  assert (acc_device_type (disp->type) != acc_device_none
+	  && acc_device_type (disp->type) != acc_device_default
+	  && acc_device_type (disp->type) != acc_device_not_host);
+  assert (!dispatchers[disp->type]);
+  dispatchers[disp->type] = disp;
+
+  gomp_mutex_unlock (&acc_device_lock);
+}
+
+/* OpenACC names some things a little differently.  */
+
+static const char *
+get_openacc_name (const char *name)
+{
+  if (strcmp (name, "nvptx") == 0)
+    return "nvidia";
+  else
+    return name;
+}
+
+static struct gomp_device_descr *
+resolve_device (acc_device_t d)
+{
+  acc_device_t d_arg = d;
+
+  switch (d)
+    {
+    case acc_device_default:
+      {
+	if (goacc_device_type)
+	  {
+	    /* Lookup the named device.  */
+	    while (++d != _ACC_device_hwm)
+	      if (dispatchers[d]
+		  && !strcasecmp (goacc_device_type,
+				  get_openacc_name (dispatchers[d]->name))
+		  && dispatchers[d]->get_num_devices_func () > 0)
+		goto found;
+
+	    gomp_fatal ("device type %s not supported", goacc_device_type);
+	  }
+
+	/* No default device specified, so start scanning for any non-host
+	   device that is available.  */
+	d = acc_device_not_host;
+      }
+      /* FALLTHROUGH */
+
+    case acc_device_not_host:
+      /* Find the first available device after acc_device_not_host.  */
+      while (++d != _ACC_device_hwm)
+	if (dispatchers[d] && dispatchers[d]->get_num_devices_func () > 0)
+	  goto found;
+      if (d_arg == acc_device_default)
+	{
+	  d = acc_device_host;
+	  goto found;
+	}
+      gomp_fatal ("no device found");
+      break;
+
+    case acc_device_host:
+      break;
+
+    default:
+      if (d > _ACC_device_hwm)
+	gomp_fatal ("device %u out of range", (unsigned)d);
+      break;
+    }
+ found:
+
+  assert (d != acc_device_none
+	  && d != acc_device_default
+	  && d != acc_device_not_host);
+
+  return dispatchers[d];
+}
+
+/* This is called when plugins have been initialized, and serves to call
+   (indirectly) the target's device_init hook.  Calling multiple times without
+   an intervening acc_shutdown_1 call is an error.  */
+
+static struct gomp_device_descr *
+acc_init_1 (acc_device_t d)
+{
+  struct gomp_device_descr *acc_dev;
+
+  acc_dev = resolve_device (d);
+
+  if (!acc_dev || acc_dev->get_num_devices_func () <= 0)
+    gomp_fatal ("device %u not supported", (unsigned)d);
+
+  if (acc_dev->is_initialized)
+    gomp_fatal ("device already active");
+
+  /* We need to remember what we were intialized as, to check shutdown etc.  */
+  init_key = d;
+
+  gomp_init_device (acc_dev);
+
+  return acc_dev;
+}
+
+static struct goacc_thread *
+goacc_new_thread (void)
+{
+  struct goacc_thread *thr = gomp_malloc (sizeof (struct gomp_thread));
+
+#if defined HAVE_TLS || defined USE_EMUTLS
+  goacc_tls_data = thr;
+#else
+  pthread_setspecific (goacc_tls_key, thr);
+#endif
+
+  pthread_setspecific (goacc_cleanup_key, thr);
+
+  gomp_mutex_lock (&goacc_thread_lock);
+  thr->next = goacc_threads;
+  goacc_threads = thr;
+  gomp_mutex_unlock (&goacc_thread_lock);
+
+  return thr;
+}
+
+static void
+goacc_destroy_thread (void *data)
+{
+  struct goacc_thread *thr = data, *walk, *prev;
+
+  gomp_mutex_lock (&goacc_thread_lock);
+
+  if (thr)
+    {
+      if (base_dev && thr->target_tls)
+	{
+	  base_dev->openacc.destroy_thread_data_func (thr->target_tls);
+	  thr->target_tls = NULL;
+	}
+
+      assert (!thr->mapped_data);
+
+      /* Remove from thread list.  */
+      for (prev = NULL, walk = goacc_threads; walk;
+	   prev = walk, walk = walk->next)
+	if (walk == thr)
+	  {
+	    if (prev == NULL)
+	      goacc_threads = walk->next;
+	    else
+	      prev->next = walk->next;
+
+	    free (thr);
+
+	    break;
+	  }
+
+      assert (walk);
+    }
+
+  gomp_mutex_unlock (&goacc_thread_lock);
+}
+
+/* Open the ORD'th device of the currently-active type (base_dev must be
+   initialised before calling).  If ORD is < 0, open the default-numbered
+   device (set by the ACC_DEVICE_NUM environment variable or a call to
+   acc_set_device_num), or leave any currently-opened device as is.  "Opening"
+   consists of calling the device's open_device_func hook, and setting up
+   thread-local data (maybe allocating, then initializing with information
+   pertaining to the newly-opened or previously-opened device).  */
+
+static void
+lazy_open (int ord)
+{
+  struct goacc_thread *thr = goacc_thread ();
+  struct gomp_device_descr *acc_dev;
+
+  if (thr && thr->dev)
+    {
+      assert (ord < 0 || ord == thr->dev->target_id);
+      return;
+    }
+
+  assert (base_dev);
+
+  if (ord < 0)
+    ord = goacc_device_num;
+
+  /* The OpenACC 2.0 spec leaves the runtime's behaviour when an out-of-range
+     device is requested as implementation-defined (4.2 ACC_DEVICE_NUM).
+     We choose to raise an error in such a case.  */
+  if (ord >= base_dev->get_num_devices_func ())
+    gomp_fatal ("device %u does not exist", ord);
+
+  if (!thr)
+    thr = goacc_new_thread ();
+
+  acc_dev = thr->dev = &base_dev[ord];
+
+  assert (acc_dev->target_id == ord);
+
+  thr->saved_bound_dev = NULL;
+  thr->mapped_data = NULL;
+
+  if (!acc_dev->openacc.target_data)
+    acc_dev->openacc.target_data = acc_dev->openacc.open_device_func (ord);
+
+  thr->target_tls
+    = acc_dev->openacc.create_thread_data_func (acc_dev->openacc.target_data);
+
+  acc_dev->openacc.async_set_async_func (acc_async_sync);
+
+  struct gomp_memory_mapping *mem_map = &acc_dev->mem_map;
+  gomp_mutex_lock (&mem_map->lock);
+  if (!mem_map->is_initialized)
+    gomp_init_tables (acc_dev, mem_map);
+  gomp_mutex_unlock (&mem_map->lock);
+}
+
+/* OpenACC 2.0a (3.2.12, 3.2.13) doesn't specify whether the serialization of
+   init/shutdown is per-process or per-thread.  We choose per-process.  */
+
+void
+acc_init (acc_device_t d)
+{
+  if (!base_dev)
+    gomp_init_targets_once ();
+
+  gomp_mutex_lock (&acc_device_lock);
+
+  base_dev = acc_init_1 (d);
+
+  lazy_open (-1);
+
+  gomp_mutex_unlock (&acc_device_lock);
+}
+
+ialias (acc_init)
+
+static void
+acc_shutdown_1 (acc_device_t d)
+{
+  struct goacc_thread *walk;
+
+  /* We don't check whether d matches the actual device found, because
+     OpenACC 2.0 (3.2.12) says the parameters to the init and this
+     call must match (for the shutdown call anyway, it's silent on
+     others).  */
+
+  if (!base_dev)
+    gomp_fatal ("no device initialized");
+  if (d != init_key)
+    gomp_fatal ("device %u(%u) is initialized",
+		(unsigned) init_key, (unsigned) base_dev->type);
+
+  gomp_mutex_lock (&goacc_thread_lock);
+
+  /* Free target-specific TLS data and close all devices.  */
+  for (walk = goacc_threads; walk != NULL; walk = walk->next)
+    {
+      if (walk->target_tls)
+	base_dev->openacc.destroy_thread_data_func (walk->target_tls);
+
+      walk->target_tls = NULL;
+
+      /* This would mean the user is shutting down OpenACC in the middle of an
+         "acc data" pragma.  Likely not intentional.  */
+      if (walk->mapped_data)
+	gomp_fatal ("shutdown in 'acc data' region");
+
+      if (walk->dev)
+	{
+	  void *target_data = walk->dev->openacc.target_data;
+	  if (walk->dev->openacc.close_device_func (target_data) < 0)
+	    gomp_fatal ("failed to close device");
+
+	  walk->dev->openacc.target_data = target_data = NULL;
+
+	  struct gomp_memory_mapping *mem_map = &walk->dev->mem_map;
+	  gomp_mutex_lock (&mem_map->lock);
+	  gomp_free_memmap (mem_map);
+	  gomp_mutex_unlock (&mem_map->lock);
+
+	  walk->dev = NULL;
+	}
+    }
+
+  gomp_mutex_unlock (&goacc_thread_lock);
+
+  gomp_fini_device (base_dev);
+
+  base_dev = NULL;
+}
+
+void
+acc_shutdown (acc_device_t d)
+{
+  gomp_mutex_lock (&acc_device_lock);
+
+  acc_shutdown_1 (d);
+
+  gomp_mutex_unlock (&acc_device_lock);
+}
+
+ialias (acc_shutdown)
+
+/* This function is called after plugins have been initialized.  It deals with
+   the "base" device, and is used to prepare the runtime for dealing with a
+   number of such devices (as implemented by some particular plugin).  If the
+   argument device type D matches a previous call to the function, return the
+   current base device, else shut the old device down and re-initialize with
+   the new device type.  */
+
+static struct gomp_device_descr *
+lazy_init (acc_device_t d)
+{
+  if (base_dev)
+    {
+      /* Re-initializing the same device, do nothing.  */
+      if (d == init_key)
+	return base_dev;
+
+      acc_shutdown_1 (init_key);
+    }
+
+  assert (!base_dev);
+
+  return acc_init_1 (d);
+}
+
+/* Ensure that plugins are loaded, initialize and open the (default-numbered)
+   device.  */
+
+static void
+lazy_init_and_open (acc_device_t d)
+{
+  if (!base_dev)
+    gomp_init_targets_once ();
+
+  gomp_mutex_lock (&acc_device_lock);
+
+  base_dev = lazy_init (d);
+
+  lazy_open (-1);
+
+  gomp_mutex_unlock (&acc_device_lock);
+}
+
+int
+acc_get_num_devices (acc_device_t d)
+{
+  int n = 0;
+  const struct gomp_device_descr *acc_dev;
+
+  if (d == acc_device_none)
+    return 0;
+
+  if (!base_dev)
+    gomp_init_targets_once ();
+
+  acc_dev = resolve_device (d);
+  if (!acc_dev)
+    return 0;
+
+  n = acc_dev->get_num_devices_func ();
+  if (n < 0)
+    n = 0;
+
+  return n;
+}
+
+ialias (acc_get_num_devices)
+
+void
+acc_set_device_type (acc_device_t d)
+{
+  lazy_init_and_open (d);
+}
+
+ialias (acc_set_device_type)
+
+acc_device_t
+acc_get_device_type (void)
+{
+  acc_device_t res = acc_device_none;
+  const struct gomp_device_descr *dev;
+
+  if (base_dev)
+    res = acc_device_type (base_dev->type);
+  else
+    {
+      gomp_init_targets_once ();
+
+      dev = resolve_device (acc_device_default);
+      res = acc_device_type (dev->type);
+    }
+
+  assert (res != acc_device_default
+	  && res != acc_device_not_host);
+
+  return res;
+}
+
+ialias (acc_get_device_type)
+
+int
+acc_get_device_num (acc_device_t d)
+{
+  const struct gomp_device_descr *dev;
+  int num;
+
+  if (d >= _ACC_device_hwm)
+    gomp_fatal ("device %u out of range", (unsigned)d);
+
+  if (!base_dev)
+    gomp_init_targets_once ();
+
+  dev = resolve_device (d);
+  if (!dev)
+    gomp_fatal ("no devices of type %u", d);
+
+  /* We might not have called lazy_open for this host thread yet, in which case
+     the get_device_num_func hook will return -1.  */
+  num = dev->openacc.get_device_num_func ();
+  if (num < 0)
+    num = goacc_device_num;
+
+  return num;
+}
+
+ialias (acc_get_device_num)
+
+void
+acc_set_device_num (int n, acc_device_t d)
+{
+  const struct gomp_device_descr *dev;
+  int num_devices;
+
+  if (!base_dev)
+    gomp_init_targets_once ();
+
+  if ((int) d == 0)
+    {
+      int i;
+
+      /* A device setting of zero sets all device types on the system to use
+         the Nth instance of that device type.  Only attempt it for initialized
+	 devices though.  */
+      for (i = acc_device_not_host + 1; i < _ACC_device_hwm; i++)
+        {
+	  dev = resolve_device (d);
+	  if (dev && dev->is_initialized)
+	    dev->openacc.set_device_num_func (n);
+	}
+
+      /* ...and for future calls to acc_init/acc_set_device_type, etc.  */
+      goacc_device_num = n;
+    }
+  else
+    {
+      struct goacc_thread *thr = goacc_thread ();
+
+      gomp_mutex_lock (&acc_device_lock);
+
+      base_dev = lazy_init (d);
+
+      num_devices = base_dev->get_num_devices_func ();
+
+      if (n >= num_devices)
+        gomp_fatal ("device %u out of range", n);
+
+      /* If we're changing the device number, de-associate this thread with
+	 the device (but don't close the device, since it may be in use by
+	 other threads).  */
+      if (thr && thr->dev && n != thr->dev->target_id)
+	thr->dev = NULL;
+
+      lazy_open (n);
+
+      gomp_mutex_unlock (&acc_device_lock);
+    }
+}
+
+ialias (acc_set_device_num)
+
+int
+acc_on_device (acc_device_t dev)
+{
+  struct goacc_thread *thr = goacc_thread ();
+
+  if (thr && thr->dev
+      && acc_device_type (thr->dev->type) == acc_device_host_nonshm)
+    return dev == acc_device_host_nonshm || dev == acc_device_not_host;
+
+  /* Just rely on the compiler builtin.  */
+  return __builtin_acc_on_device (dev);
+}
+
+ialias (acc_on_device)
+
+attribute_hidden void
+goacc_runtime_initialize (void)
+{
+  gomp_mutex_init (&acc_device_lock);
+
+#if !(defined HAVE_TLS || defined USE_EMUTLS)
+  pthread_key_create (&goacc_tls_key, NULL);
+#endif
+
+  pthread_key_create (&goacc_cleanup_key, goacc_destroy_thread);
+
+  base_dev = NULL;
+
+  goacc_threads = NULL;
+  gomp_mutex_init (&goacc_thread_lock);
+}
+
+/* Compiler helper functions */
+
+attribute_hidden void
+goacc_save_and_set_bind (acc_device_t d)
+{
+  struct goacc_thread *thr = goacc_thread ();
+
+  assert (!thr->saved_bound_dev);
+
+  thr->saved_bound_dev = thr->dev;
+  thr->dev = dispatchers[d];
+}
+
+attribute_hidden void
+goacc_restore_bind (void)
+{
+  struct goacc_thread *thr = goacc_thread ();
+
+  thr->dev = thr->saved_bound_dev;
+  thr->saved_bound_dev = NULL;
+}
+
+/* This is called from any OpenACC support function that may need to implicitly
+   initialize the libgomp runtime.  On exit all such initialization will have
+   been done, and both the global ACC_dev and the per-host-thread ACC_memmap
+   pointers will be valid.  */
+
+attribute_hidden void
+goacc_lazy_initialize (void)
+{
+  struct goacc_thread *thr = goacc_thread ();
+
+  if (thr && thr->dev)
+    return;
+
+  if (!base_dev)
+    lazy_init_and_open (acc_device_default);
+  else
+    {
+      gomp_mutex_lock (&acc_device_lock);
+      lazy_open (-1);
+      gomp_mutex_unlock (&acc_device_lock);
+    }
+}
diff --git a/libgomp/oacc-int.h b/libgomp/oacc-int.h
new file mode 100644
index 00000000000..85619c8d10a
--- /dev/null
+++ b/libgomp/oacc-int.h
@@ -0,0 +1,105 @@
+/* OpenACC Runtime - internal declarations
+
+   Copyright (C) 2013-2015 Free Software Foundation, Inc.
+
+   Contributed by Mentor Embedded.
+
+   This file is part of the GNU Offloading and Multi Processing Library
+   (libgomp).
+
+   Libgomp is free software; you can redistribute it and/or modify it
+   under the terms of the GNU General Public License as published by
+   the Free Software Foundation; either version 3, or (at your option)
+   any later version.
+
+   Libgomp is distributed in the hope that it will be useful, but WITHOUT ANY
+   WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
+   FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+   more details.
+
+   Under Section 7 of GPL version 3, you are granted additional
+   permissions described in the GCC Runtime Library Exception, version
+   3.1, as published by the Free Software Foundation.
+
+   You should have received a copy of the GNU General Public License and
+   a copy of the GCC Runtime Library Exception along with this program;
+   see the files COPYING3 and COPYING.RUNTIME respectively.  If not, see
+   <http://www.gnu.org/licenses/>.  */
+
+/* This file contains data types and function declarations that are not
+   part of the official OpenACC user interface.  There are declarations
+   in here that are part of the GNU OpenACC ABI, in that the compiler is
+   required to know about them and use them.
+
+   The convention is that the all caps prefix "GOACC" is used group items
+   that are part of the external ABI, and the lower case prefix "goacc"
+   is used group items that are completely private to the library.  */
+
+#ifndef OACC_INT_H
+#define OACC_INT_H 1
+
+#include "openacc.h"
+#include "config.h"
+#include <stddef.h>
+#include <stdbool.h>
+#include <stdarg.h>
+
+#ifdef HAVE_ATTRIBUTE_VISIBILITY
+# pragma GCC visibility push(hidden)
+#endif
+
+static inline enum acc_device_t
+acc_device_type (enum offload_target_type type)
+{
+  return (enum acc_device_t) type;
+}
+
+struct goacc_thread
+{
+  /* The device for the current thread.  */
+  struct gomp_device_descr *dev;
+
+  struct gomp_device_descr *saved_bound_dev;
+
+  /* This is a linked list of data mapped by the "acc data" pragma, following
+     strictly push/pop semantics according to lexical scope.  */
+  struct target_mem_desc *mapped_data;
+
+  /* These structures form a list: this is the next thread in that list.  */
+  struct goacc_thread *next;
+
+  /* Target-specific data (used by plugin).  */
+  void *target_tls;
+};
+
+#if defined HAVE_TLS || defined USE_EMUTLS
+extern __thread struct goacc_thread *goacc_tls_data;
+static inline struct goacc_thread *
+goacc_thread (void)
+{
+  return goacc_tls_data;
+}
+#else
+extern pthread_key_t goacc_tls_key;
+static inline struct goacc_thread *
+goacc_thread (void)
+{
+  return pthread_getspecific (goacc_tls_key);
+}
+#endif
+
+void goacc_register (struct gomp_device_descr *) __GOACC_NOTHROW;
+
+/* Current dispatcher.  */
+extern struct gomp_device_descr *base_dev;
+
+void goacc_runtime_initialize (void);
+void goacc_save_and_set_bind (acc_device_t);
+void goacc_restore_bind (void);
+void goacc_lazy_initialize (void);
+
+#ifdef HAVE_ATTRIBUTE_VISIBILITY
+# pragma GCC visibility pop
+#endif
+
+#endif
diff --git a/libgomp/oacc-mem.c b/libgomp/oacc-mem.c
new file mode 100644
index 00000000000..0096d514297
--- /dev/null
+++ b/libgomp/oacc-mem.c
@@ -0,0 +1,585 @@
+/* OpenACC Runtime initialization routines
+
+   Copyright (C) 2013-2015 Free Software Foundation, Inc.
+
+   Contributed by Mentor Embedded.
+
+   This file is part of the GNU Offloading and Multi Processing Library
+   (libgomp).
+
+   Libgomp is free software; you can redistribute it and/or modify it
+   under the terms of the GNU General Public License as published by
+   the Free Software Foundation; either version 3, or (at your option)
+   any later version.
+
+   Libgomp is distributed in the hope that it will be useful, but WITHOUT ANY
+   WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
+   FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+   more details.
+
+   Under Section 7 of GPL version 3, you are granted additional
+   permissions described in the GCC Runtime Library Exception, version
+   3.1, as published by the Free Software Foundation.
+
+   You should have received a copy of the GNU General Public License and
+   a copy of the GCC Runtime Library Exception along with this program;
+   see the files COPYING3 and COPYING.RUNTIME respectively.  If not, see
+   <http://www.gnu.org/licenses/>.  */
+
+#include "openacc.h"
+#include "config.h"
+#include "libgomp.h"
+#include "gomp-constants.h"
+#include "oacc-int.h"
+#include "splay-tree.h"
+#include <stdint.h>
+#include <assert.h>
+
+/* Return block containing [H->S), or NULL if not contained.  */
+
+static splay_tree_key
+lookup_host (struct gomp_memory_mapping *mem_map, void *h, size_t s)
+{
+  struct splay_tree_key_s node;
+  splay_tree_key key;
+
+  node.host_start = (uintptr_t) h;
+  node.host_end = (uintptr_t) h + s;
+
+  gomp_mutex_lock (&mem_map->lock);
+
+  key = splay_tree_lookup (&mem_map->splay_tree, &node);
+
+  gomp_mutex_unlock (&mem_map->lock);
+
+  return key;
+}
+
+/* Return block containing [D->S), or NULL if not contained.
+   The list isn't ordered by device address, so we have to iterate
+   over the whole array.  This is not expected to be a common
+   operation.  */
+
+static splay_tree_key
+lookup_dev (struct target_mem_desc *tgt, void *d, size_t s)
+{
+  int i;
+  struct target_mem_desc *t;
+  struct gomp_memory_mapping *mem_map;
+
+  if (!tgt)
+    return NULL;
+
+  mem_map = tgt->mem_map;
+
+  gomp_mutex_lock (&mem_map->lock);
+
+  for (t = tgt; t != NULL; t = t->prev)
+    {
+      if (t->tgt_start <= (uintptr_t) d && t->tgt_end >= (uintptr_t) d + s)
+        break;
+    }
+
+  gomp_mutex_unlock (&mem_map->lock);
+
+  if (!t)
+    return NULL;
+
+  for (i = 0; i < t->list_count; i++)
+    {
+      void * offset;
+
+      splay_tree_key k = &t->array[i].key;
+      offset = d - t->tgt_start + k->tgt_offset;
+
+      if (k->host_start + offset <= (void *) k->host_end)
+        return k;
+    }
+
+  return NULL;
+}
+
+/* OpenACC is silent on how memory exhaustion is indicated.  We return
+   NULL.  */
+
+void *
+acc_malloc (size_t s)
+{
+  if (!s)
+    return NULL;
+
+  goacc_lazy_initialize ();
+
+  struct goacc_thread *thr = goacc_thread ();
+
+  return base_dev->alloc_func (thr->dev->target_id, s);
+}
+
+/* OpenACC 2.0a (3.2.16) doesn't specify what to do in the event
+   the device address is mapped. We choose to check if it mapped,
+   and if it is, to unmap it. */
+void
+acc_free (void *d)
+{
+  splay_tree_key k;
+  struct goacc_thread *thr = goacc_thread ();
+
+  if (!d)
+    return;
+
+  /* We don't have to call lazy open here, as the ptr value must have
+     been returned by acc_malloc.  It's not permitted to pass NULL in
+     (unless you got that null from acc_malloc).  */
+  if ((k = lookup_dev (thr->dev->openacc.data_environ, d, 1)))
+   {
+     void *offset;
+
+     offset = d - k->tgt->tgt_start + k->tgt_offset;
+
+     acc_unmap_data ((void *)(k->host_start + offset));
+   }
+
+  base_dev->free_func (thr->dev->target_id, d);
+}
+
+void
+acc_memcpy_to_device (void *d, void *h, size_t s)
+{
+  /* No need to call lazy open here, as the device pointer must have
+     been obtained from a routine that did that.  */
+  struct goacc_thread *thr = goacc_thread ();
+
+  base_dev->host2dev_func (thr->dev->target_id, d, h, s);
+}
+
+void
+acc_memcpy_from_device (void *h, void *d, size_t s)
+{
+  /* No need to call lazy open here, as the device pointer must have
+     been obtained from a routine that did that.  */
+  struct goacc_thread *thr = goacc_thread ();
+
+  base_dev->dev2host_func (thr->dev->target_id, h, d, s);
+}
+
+/* Return the device pointer that corresponds to host data H.  Or NULL
+   if no mapping.  */
+
+void *
+acc_deviceptr (void *h)
+{
+  splay_tree_key n;
+  void *d;
+  void *offset;
+
+  goacc_lazy_initialize ();
+
+  struct goacc_thread *thr = goacc_thread ();
+
+  n = lookup_host (&thr->dev->mem_map, h, 1);
+
+  if (!n)
+    return NULL;
+
+  offset = h - n->host_start;
+
+  d = n->tgt->tgt_start + n->tgt_offset + offset;
+
+  return d;
+}
+
+/* Return the host pointer that corresponds to device data D.  Or NULL
+   if no mapping.  */
+
+void *
+acc_hostptr (void *d)
+{
+  splay_tree_key n;
+  void *h;
+  void *offset;
+
+  goacc_lazy_initialize ();
+
+  struct goacc_thread *thr = goacc_thread ();
+
+  n = lookup_dev (thr->dev->openacc.data_environ, d, 1);
+
+  if (!n)
+    return NULL;
+
+  offset = d - n->tgt->tgt_start + n->tgt_offset;
+
+  h = n->host_start + offset;
+
+  return h;
+}
+
+/* Return 1 if host data [H,+S] is present on the device.  */
+
+int
+acc_is_present (void *h, size_t s)
+{
+  splay_tree_key n;
+
+  if (!s || !h)
+    return 0;
+
+  goacc_lazy_initialize ();
+
+  struct goacc_thread *thr = goacc_thread ();
+  struct gomp_device_descr *acc_dev = thr->dev;
+
+  n = lookup_host (&acc_dev->mem_map, h, s);
+
+  if (n && ((uintptr_t)h < n->host_start
+	    || (uintptr_t)h + s > n->host_end
+	    || s > n->host_end - n->host_start))
+    n = NULL;
+
+  return n != NULL;
+}
+
+/* Create a mapping for host [H,+S] -> device [D,+S] */
+
+void
+acc_map_data (void *h, void *d, size_t s)
+{
+  struct target_mem_desc *tgt;
+  size_t mapnum = 1;
+  void *hostaddrs = h;
+  void *devaddrs = d;
+  size_t sizes = s;
+  unsigned short kinds = GOMP_MAP_ALLOC;
+
+  goacc_lazy_initialize ();
+
+  struct goacc_thread *thr = goacc_thread ();
+  struct gomp_device_descr *acc_dev = thr->dev;
+
+  if (acc_dev->capabilities & GOMP_OFFLOAD_CAP_SHARED_MEM)
+    {
+      if (d != h)
+        gomp_fatal ("cannot map data on shared-memory system");
+
+      tgt = gomp_map_vars (NULL, 0, NULL, NULL, NULL, NULL, true, false);
+    }
+  else
+    {
+      struct goacc_thread *thr = goacc_thread ();
+
+      if (!d || !h || !s)
+	gomp_fatal ("[%p,+%d]->[%p,+%d] is a bad map",
+                    (void *)h, (int)s, (void *)d, (int)s);
+
+      if (lookup_host (&acc_dev->mem_map, h, s))
+	gomp_fatal ("host address [%p, +%d] is already mapped", (void *)h,
+		    (int)s);
+
+      if (lookup_dev (thr->dev->openacc.data_environ, d, s))
+	gomp_fatal ("device address [%p, +%d] is already mapped", (void *)d,
+		    (int)s);
+
+      tgt = gomp_map_vars (acc_dev, mapnum, &hostaddrs, &devaddrs, &sizes,
+			   &kinds, true, false);
+    }
+
+  tgt->prev = acc_dev->openacc.data_environ;
+  acc_dev->openacc.data_environ = tgt;
+}
+
+void
+acc_unmap_data (void *h)
+{
+  struct goacc_thread *thr = goacc_thread ();
+  struct gomp_device_descr *acc_dev = thr->dev;
+
+  /* No need to call lazy open, as the address must have been mapped.  */
+
+  size_t host_size;
+  splay_tree_key n = lookup_host (&acc_dev->mem_map, h, 1);
+  struct target_mem_desc *t;
+
+  if (!n)
+    gomp_fatal ("%p is not a mapped block", (void *)h);
+
+  host_size = n->host_end - n->host_start;
+
+  if (n->host_start != (uintptr_t) h)
+    gomp_fatal ("[%p,%d] surrounds1 %p",
+		(void *) n->host_start, (int) host_size, (void *) h);
+
+  t = n->tgt;
+
+  if (t->refcount == 2)
+    {
+      struct target_mem_desc *tp;
+
+      /* This is the last reference, so pull the descriptor off the
+         chain. This avoids gomp_unmap_vars via gomp_unmap_tgt from
+         freeing the device memory. */
+      t->tgt_end = 0;
+      t->to_free = 0;
+
+      gomp_mutex_lock (&acc_dev->mem_map.lock);
+
+      for (tp = NULL, t = acc_dev->openacc.data_environ; t != NULL;
+	   tp = t, t = t->prev)
+	if (n->tgt == t)
+	  {
+	    if (tp)
+	      tp->prev = t->prev;
+	    else
+	      acc_dev->openacc.data_environ = t->prev;
+
+	    break;
+	  }
+
+      gomp_mutex_unlock (&acc_dev->mem_map.lock);
+    }
+
+  gomp_unmap_vars (t, true);
+}
+
+#define FLAG_PRESENT (1 << 0)
+#define FLAG_CREATE (1 << 1)
+#define FLAG_COPY (1 << 2)
+
+static void *
+present_create_copy (unsigned f, void *h, size_t s)
+{
+  void *d;
+  splay_tree_key n;
+
+  if (!h || !s)
+    gomp_fatal ("[%p,+%d] is a bad range", (void *)h, (int)s);
+
+  goacc_lazy_initialize ();
+
+  struct goacc_thread *thr = goacc_thread ();
+  struct gomp_device_descr *acc_dev = thr->dev;
+
+  n = lookup_host (&acc_dev->mem_map, h, s);
+  if (n)
+    {
+      /* Present. */
+      d = (void *) (n->tgt->tgt_start + n->tgt_offset);
+
+      if (!(f & FLAG_PRESENT))
+        gomp_fatal ("[%p,+%d] already mapped to [%p,+%d]",
+            (void *)h, (int)s, (void *)d, (int)s);
+      if ((h + s) > (void *)n->host_end)
+        gomp_fatal ("[%p,+%d] not mapped", (void *)h, (int)s);
+    }
+  else if (!(f & FLAG_CREATE))
+    {
+      gomp_fatal ("[%p,+%d] not mapped", (void *)h, (int)s);
+    }
+  else
+    {
+      struct target_mem_desc *tgt;
+      size_t mapnum = 1;
+      unsigned short kinds;
+      void *hostaddrs = h;
+
+      if (f & FLAG_COPY)
+	kinds = GOMP_MAP_TO;
+      else
+	kinds = GOMP_MAP_ALLOC;
+
+      tgt = gomp_map_vars (acc_dev, mapnum, &hostaddrs, NULL, &s, &kinds, true,
+			   false);
+
+      gomp_mutex_lock (&acc_dev->mem_map.lock);
+
+      d = tgt->to_free;
+      tgt->prev = acc_dev->openacc.data_environ;
+      acc_dev->openacc.data_environ = tgt;
+
+      gomp_mutex_unlock (&acc_dev->mem_map.lock);
+    }
+
+  return d;
+}
+
+void *
+acc_create (void *h, size_t s)
+{
+  return present_create_copy (FLAG_CREATE, h, s);
+}
+
+void *
+acc_copyin (void *h, size_t s)
+{
+  return present_create_copy (FLAG_CREATE | FLAG_COPY, h, s);
+}
+
+void *
+acc_present_or_create (void *h, size_t s)
+{
+  return present_create_copy (FLAG_PRESENT | FLAG_CREATE, h, s);
+}
+
+void *
+acc_present_or_copyin (void *h, size_t s)
+{
+  return present_create_copy (FLAG_PRESENT | FLAG_CREATE | FLAG_COPY, h, s);
+}
+
+#define FLAG_COPYOUT (1 << 0)
+
+static void
+delete_copyout (unsigned f, void *h, size_t s)
+{
+  size_t host_size;
+  splay_tree_key n;
+  void *d;
+  struct goacc_thread *thr = goacc_thread ();
+  struct gomp_device_descr *acc_dev = thr->dev;
+
+  n = lookup_host (&acc_dev->mem_map, h, s);
+
+  /* No need to call lazy open, as the data must already have been
+     mapped.  */
+
+  if (!n)
+    gomp_fatal ("[%p,%d] is not mapped", (void *)h, (int)s);
+
+  d = (void *) (n->tgt->tgt_start + n->tgt_offset);
+
+  host_size = n->host_end - n->host_start;
+
+  if (n->host_start != (uintptr_t) h || host_size != s)
+    gomp_fatal ("[%p,%d] surrounds2 [%p,+%d]",
+		(void *) n->host_start, (int) host_size, (void *) h, (int) s);
+
+  if (f & FLAG_COPYOUT)
+    acc_dev->dev2host_func (acc_dev->target_id, h, d, s);
+
+  acc_unmap_data (h);
+
+  acc_dev->free_func (acc_dev->target_id, d);
+}
+
+void
+acc_delete (void *h , size_t s)
+{
+  delete_copyout (0, h, s);
+}
+
+void acc_copyout (void *h, size_t s)
+{
+  delete_copyout (FLAG_COPYOUT, h, s);
+}
+
+static void
+update_dev_host (int is_dev, void *h, size_t s)
+{
+  splay_tree_key n;
+  void *d;
+  struct goacc_thread *thr = goacc_thread ();
+  struct gomp_device_descr *acc_dev = thr->dev;
+
+  n = lookup_host (&acc_dev->mem_map, h, s);
+
+  /* No need to call lazy open, as the data must already have been
+     mapped.  */
+
+  if (!n)
+    gomp_fatal ("[%p,%d] is not mapped", h, (int)s);
+
+  d = (void *) (n->tgt->tgt_start + n->tgt_offset);
+
+  if (is_dev)
+    acc_dev->host2dev_func (acc_dev->target_id, d, h, s);
+  else
+    acc_dev->dev2host_func (acc_dev->target_id, h, d, s);
+}
+
+void
+acc_update_device (void *h, size_t s)
+{
+  update_dev_host (1, h, s);
+}
+
+void
+acc_update_self (void *h, size_t s)
+{
+  update_dev_host (0, h, s);
+}
+
+void
+gomp_acc_insert_pointer (size_t mapnum, void **hostaddrs, size_t *sizes,
+			 void *kinds)
+{
+  struct target_mem_desc *tgt;
+  struct goacc_thread *thr = goacc_thread ();
+  struct gomp_device_descr *acc_dev = thr->dev;
+
+  gomp_debug (0, "  %s: prepare mappings\n", __FUNCTION__);
+  tgt = gomp_map_vars (acc_dev, mapnum, hostaddrs,
+		       NULL, sizes, kinds, true, false);
+  gomp_debug (0, "  %s: mappings prepared\n", __FUNCTION__);
+  tgt->prev = acc_dev->openacc.data_environ;
+  acc_dev->openacc.data_environ = tgt;
+}
+
+void
+gomp_acc_remove_pointer (void *h, bool force_copyfrom, int async, int mapnum)
+{
+  struct goacc_thread *thr = goacc_thread ();
+  struct gomp_device_descr *acc_dev = thr->dev;
+  splay_tree_key n;
+  struct target_mem_desc *t;
+  int minrefs = (mapnum == 1) ? 2 : 3;
+
+  n = lookup_host (&acc_dev->mem_map, h, 1);
+
+  if (!n)
+    gomp_fatal ("%p is not a mapped block", (void *)h);
+
+  gomp_debug (0, "  %s: restore mappings\n", __FUNCTION__);
+
+  t = n->tgt;
+
+  struct target_mem_desc *tp;
+
+  gomp_mutex_lock (&acc_dev->mem_map.lock);
+
+  if (t->refcount == minrefs)
+    {
+      /* This is the last reference, so pull the descriptor off the
+	 chain. This avoids gomp_unmap_vars via gomp_unmap_tgt from
+	 freeing the device memory. */
+      t->tgt_end = 0;
+      t->to_free = 0;
+
+      for (tp = NULL, t = acc_dev->openacc.data_environ; t != NULL;
+	   tp = t, t = t->prev)
+	{
+	  if (n->tgt == t)
+	    {
+	      if (tp)
+		tp->prev = t->prev;
+	      else
+		acc_dev->openacc.data_environ = t->prev;
+	      break;
+	    }
+	}
+    }
+
+  if (force_copyfrom)
+    t->list[0]->copy_from = 1;
+
+  gomp_mutex_unlock (&acc_dev->mem_map.lock);
+
+  /* If running synchronously, unmap immediately.  */
+  if (async < acc_async_noval)
+    gomp_unmap_vars (t, true);
+  else
+    {
+      gomp_copy_from_async (t);
+      acc_dev->openacc.register_async_cleanup_func (t);
+    }
+
+  gomp_debug (0, "  %s: mappings restored\n", __FUNCTION__);
+}
diff --git a/libgomp/oacc-parallel.c b/libgomp/oacc-parallel.c
new file mode 100644
index 00000000000..6d5386b1f94
--- /dev/null
+++ b/libgomp/oacc-parallel.c
@@ -0,0 +1,490 @@
+/* Copyright (C) 2013-2015 Free Software Foundation, Inc.
+
+   Contributed by Mentor Embedded.
+
+   This file is part of the GNU Offloading and Multi Processing Library
+   (libgomp).
+
+   Libgomp is free software; you can redistribute it and/or modify it
+   under the terms of the GNU General Public License as published by
+   the Free Software Foundation; either version 3, or (at your option)
+   any later version.
+
+   Libgomp is distributed in the hope that it will be useful, but WITHOUT ANY
+   WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
+   FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+   more details.
+
+   Under Section 7 of GPL version 3, you are granted additional
+   permissions described in the GCC Runtime Library Exception, version
+   3.1, as published by the Free Software Foundation.
+
+   You should have received a copy of the GNU General Public License and
+   a copy of the GCC Runtime Library Exception along with this program;
+   see the files COPYING3 and COPYING.RUNTIME respectively.  If not, see
+   <http://www.gnu.org/licenses/>.  */
+
+/* This file handles OpenACC constructs.  */
+
+#include "openacc.h"
+#include "libgomp.h"
+#include "libgomp_g.h"
+#include "gomp-constants.h"
+#include "oacc-int.h"
+#include <string.h>
+#include <stdarg.h>
+#include <assert.h>
+#include <alloca.h>
+
+static int
+find_pset (int pos, size_t mapnum, unsigned short *kinds)
+{
+  if (pos + 1 >= mapnum)
+    return 0;
+
+  unsigned char kind = kinds[pos+1] & 0xff;
+
+  return kind == GOMP_MAP_TO_PSET;
+}
+
+
+/* Ensure that the target device for DEVICE_TYPE is initialised (and that
+   plugins have been loaded if appropriate).  The ACC_dev variable for the
+   current thread will be set appropriately for the given device type on
+   return.  */
+
+attribute_hidden void
+select_acc_device (int device_type)
+{
+  goacc_lazy_initialize ();
+
+  if (device_type == GOMP_DEVICE_HOST_FALLBACK)
+    return;
+
+  if (device_type == acc_device_none)
+    device_type = acc_device_host;
+
+  if (device_type >= 0)
+    {
+      /* NOTE: this will go badly if the surrounding data environment is set up
+         to use a different device type.  We'll just have to trust that users
+	 know what they're doing...  */
+      acc_set_device_type (device_type);
+    }
+}
+
+static void goacc_wait (int async, int num_waits, va_list ap);
+
+void
+GOACC_parallel (int device, void (*fn) (void *), const void *offload_table,
+		size_t mapnum, void **hostaddrs, size_t *sizes,
+		unsigned short *kinds,
+		int num_gangs, int num_workers, int vector_length,
+		int async, int num_waits, ...)
+{
+  bool host_fallback = device == GOMP_DEVICE_HOST_FALLBACK;
+  va_list ap;
+  struct goacc_thread *thr;
+  struct gomp_device_descr *acc_dev;
+  struct target_mem_desc *tgt;
+  void **devaddrs;
+  unsigned int i;
+  struct splay_tree_key_s k;
+  splay_tree_key tgt_fn_key;
+  void (*tgt_fn);
+
+  if (num_gangs != 1)
+    gomp_fatal ("num_gangs (%d) different from one is not yet supported",
+		num_gangs);
+  if (num_workers != 1)
+    gomp_fatal ("num_workers (%d) different from one is not yet supported",
+		num_workers);
+
+  gomp_debug (0, "%s: mapnum=%zd, hostaddrs=%p, sizes=%p, kinds=%p, async=%d\n",
+	      __FUNCTION__, mapnum, hostaddrs, sizes, kinds, async);
+
+  select_acc_device (device);
+
+  thr = goacc_thread ();
+  acc_dev = thr->dev;
+
+  /* Host fallback if "if" clause is false or if the current device is set to
+     the host.  */
+  if (host_fallback)
+    {
+      goacc_save_and_set_bind (acc_device_host);
+      fn (hostaddrs);
+      goacc_restore_bind ();
+      return;
+    }
+  else if (acc_device_type (acc_dev->type) == acc_device_host)
+    {
+      fn (hostaddrs);
+      return;
+    }
+
+  va_start (ap, num_waits);
+  
+  if (num_waits > 0)
+    goacc_wait (async, num_waits, ap);
+
+  va_end (ap);
+
+  acc_dev->openacc.async_set_async_func (async);
+
+  if (!(acc_dev->capabilities & GOMP_OFFLOAD_CAP_NATIVE_EXEC))
+    {
+      k.host_start = (uintptr_t) fn;
+      k.host_end = k.host_start + 1;
+      gomp_mutex_lock (&acc_dev->mem_map.lock);
+      tgt_fn_key = splay_tree_lookup (&acc_dev->mem_map.splay_tree, &k);
+      gomp_mutex_unlock (&acc_dev->mem_map.lock);
+
+      if (tgt_fn_key == NULL)
+	gomp_fatal ("target function wasn't mapped");
+
+      tgt_fn = (void (*)) tgt_fn_key->tgt->tgt_start;
+    }
+  else
+    tgt_fn = (void (*)) fn;
+
+  tgt = gomp_map_vars (acc_dev, mapnum, hostaddrs, NULL, sizes, kinds, true,
+		       false);
+
+  devaddrs = alloca (sizeof (void *) * mapnum);
+  for (i = 0; i < mapnum; i++)
+    devaddrs[i] = (void *) (tgt->list[i]->tgt->tgt_start
+			    + tgt->list[i]->tgt_offset);
+
+  acc_dev->openacc.exec_func (tgt_fn, mapnum, hostaddrs, devaddrs, sizes, kinds,
+			      num_gangs, num_workers, vector_length, async,
+			      tgt);
+
+  /* If running synchronously, unmap immediately.  */
+  if (async < acc_async_noval)
+    gomp_unmap_vars (tgt, true);
+  else
+    {
+      gomp_copy_from_async (tgt);
+      acc_dev->openacc.register_async_cleanup_func (tgt);
+    }
+
+  acc_dev->openacc.async_set_async_func (acc_async_sync);
+}
+
+void
+GOACC_data_start (int device, const void *offload_table, size_t mapnum,
+		  void **hostaddrs, size_t *sizes, unsigned short *kinds)
+{
+  bool host_fallback = device == GOMP_DEVICE_HOST_FALLBACK;
+  struct target_mem_desc *tgt;
+
+  gomp_debug (0, "%s: mapnum=%zd, hostaddrs=%p, sizes=%p, kinds=%p\n",
+	      __FUNCTION__, mapnum, hostaddrs, sizes, kinds);
+
+  select_acc_device (device);
+
+  struct goacc_thread *thr = goacc_thread ();
+  struct gomp_device_descr *acc_dev = thr->dev;
+
+  /* Host fallback or 'do nothing'.  */
+  if ((acc_dev->capabilities & GOMP_OFFLOAD_CAP_SHARED_MEM)
+      || host_fallback)
+    {
+      tgt = gomp_map_vars (NULL, 0, NULL, NULL, NULL, NULL, true, false);
+      tgt->prev = thr->mapped_data;
+      thr->mapped_data = tgt;
+
+      return;
+    }
+
+  gomp_debug (0, "  %s: prepare mappings\n", __FUNCTION__);
+  tgt = gomp_map_vars (acc_dev, mapnum, hostaddrs, NULL, sizes, kinds, true,
+		       false);
+  gomp_debug (0, "  %s: mappings prepared\n", __FUNCTION__);
+  tgt->prev = thr->mapped_data;
+  thr->mapped_data = tgt;
+}
+
+void
+GOACC_data_end (void)
+{
+  struct goacc_thread *thr = goacc_thread ();
+  struct target_mem_desc *tgt = thr->mapped_data;
+
+  gomp_debug (0, "  %s: restore mappings\n", __FUNCTION__);
+  thr->mapped_data = tgt->prev;
+  gomp_unmap_vars (tgt, true);
+  gomp_debug (0, "  %s: mappings restored\n", __FUNCTION__);
+}
+
+void
+GOACC_enter_exit_data (int device, const void *offload_table, size_t mapnum,
+		       void **hostaddrs, size_t *sizes, unsigned short *kinds,
+		       int async, int num_waits, ...)
+{
+  struct goacc_thread *thr;
+  struct gomp_device_descr *acc_dev;
+  bool host_fallback = device == GOMP_DEVICE_HOST_FALLBACK;
+  bool data_enter = false;
+  size_t i;
+
+  select_acc_device (device);
+
+  thr = goacc_thread ();
+  acc_dev = thr->dev;
+
+  if ((acc_dev->capabilities & GOMP_OFFLOAD_CAP_SHARED_MEM)
+      || host_fallback)
+    return;
+
+  if (num_waits > 0)
+    {
+      va_list ap;
+
+      va_start (ap, num_waits);
+
+      goacc_wait (async, num_waits, ap);
+
+      va_end (ap);
+    }
+
+  acc_dev->openacc.async_set_async_func (async);
+
+  /* Determine if this is an "acc enter data".  */
+  for (i = 0; i < mapnum; ++i)
+    {
+      unsigned char kind = kinds[i] & 0xff;
+
+      if (kind == GOMP_MAP_POINTER || kind == GOMP_MAP_TO_PSET)
+	continue;
+
+      if (kind == GOMP_MAP_FORCE_ALLOC
+	  || kind == GOMP_MAP_FORCE_PRESENT
+	  || kind == GOMP_MAP_FORCE_TO)
+	{
+	  data_enter = true;
+	  break;
+	}
+
+      if (kind == GOMP_MAP_FORCE_DEALLOC
+	  || kind == GOMP_MAP_FORCE_FROM)
+	break;
+
+      gomp_fatal (">>>> GOACC_enter_exit_data UNHANDLED kind 0x%.2x",
+		      kind);
+    }
+
+  if (data_enter)
+    {
+      for (i = 0; i < mapnum; i++)
+	{
+	  unsigned char kind = kinds[i] & 0xff;
+
+	  /* Scan for PSETs.  */
+	  int psets = find_pset (i, mapnum, kinds);
+
+	  if (!psets)
+	    {
+	      switch (kind)
+		{
+		case GOMP_MAP_POINTER:
+		  gomp_acc_insert_pointer (1, &hostaddrs[i], &sizes[i],
+					&kinds[i]);
+		  break;
+		case GOMP_MAP_FORCE_ALLOC:
+		  acc_create (hostaddrs[i], sizes[i]);
+		  break;
+		case GOMP_MAP_FORCE_PRESENT:
+		  acc_present_or_copyin (hostaddrs[i], sizes[i]);
+		  break;
+		case GOMP_MAP_FORCE_TO:
+		  acc_present_or_copyin (hostaddrs[i], sizes[i]);
+		  break;
+		default:
+		  gomp_fatal (">>>> GOACC_enter_exit_data UNHANDLED kind 0x%.2x",
+			      kind);
+		  break;
+		}
+	    }
+	  else
+	    {
+	      gomp_acc_insert_pointer (3, &hostaddrs[i], &sizes[i], &kinds[i]);
+	      /* Increment 'i' by two because OpenACC requires fortran
+		 arrays to be contiguous, so each PSET is associated with
+		 one of MAP_FORCE_ALLOC/MAP_FORCE_PRESET/MAP_FORCE_TO, and
+		 one MAP_POINTER.  */
+	      i += 2;
+	    }
+	}
+    }
+  else
+    for (i = 0; i < mapnum; ++i)
+      {
+	unsigned char kind = kinds[i] & 0xff;
+
+	int psets = find_pset (i, mapnum, kinds);
+
+	if (!psets)
+	  {
+	    switch (kind)
+	      {
+	      case GOMP_MAP_POINTER:
+		gomp_acc_remove_pointer (hostaddrs[i], (kinds[i] & 0xff)
+					 == GOMP_MAP_FORCE_FROM,
+					 async, 1);
+		break;
+	      case GOMP_MAP_FORCE_DEALLOC:
+		acc_delete (hostaddrs[i], sizes[i]);
+		break;
+	      case GOMP_MAP_FORCE_FROM:
+		acc_copyout (hostaddrs[i], sizes[i]);
+		break;
+	      default:
+		gomp_fatal (">>>> GOACC_enter_exit_data UNHANDLED kind 0x%.2x",
+			    kind);
+		break;
+	      }
+	  }
+	else
+	  {
+	    gomp_acc_remove_pointer (hostaddrs[i], (kinds[i] & 0xff)
+				     == GOMP_MAP_FORCE_FROM, async, 3);
+	    /* See the above comment.  */
+	    i += 2;
+	  }
+      }
+
+  acc_dev->openacc.async_set_async_func (acc_async_sync);
+}
+
+static void
+goacc_wait (int async, int num_waits, va_list ap)
+{
+  struct goacc_thread *thr = goacc_thread ();
+  struct gomp_device_descr *acc_dev = thr->dev;
+  int i;
+
+  assert (num_waits >= 0);
+
+  if (async == acc_async_sync && num_waits == 0)
+    {
+      acc_wait_all ();
+      return;
+    }
+
+  if (async == acc_async_sync && num_waits)
+    {
+      for (i = 0; i < num_waits; i++)
+        {
+          int qid = va_arg (ap, int);
+
+          if (acc_async_test (qid))
+            continue;
+
+          acc_wait (qid);
+        }
+      return;
+    }
+
+  if (async == acc_async_noval && num_waits == 0)
+    {
+      acc_dev->openacc.async_wait_all_async_func (acc_async_noval);
+      return;
+    }
+
+  for (i = 0; i < num_waits; i++)
+    {
+      int qid = va_arg (ap, int);
+
+      if (acc_async_test (qid))
+	continue;
+
+      /* If we're waiting on the same asynchronous queue as we're launching on,
+         the queue itself will order work as required, so there's no need to
+	 wait explicitly.  */
+      if (qid != async)
+	acc_dev->openacc.async_wait_async_func (qid, async);
+    }
+}
+
+void
+GOACC_update (int device, const void *offload_table, size_t mapnum,
+	      void **hostaddrs, size_t *sizes, unsigned short *kinds,
+	      int async, int num_waits, ...)
+{
+  bool host_fallback = device == GOMP_DEVICE_HOST_FALLBACK;
+  size_t i;
+
+  select_acc_device (device);
+
+  struct goacc_thread *thr = goacc_thread ();
+  struct gomp_device_descr *acc_dev = thr->dev;
+
+  if ((acc_dev->capabilities & GOMP_OFFLOAD_CAP_SHARED_MEM)
+      || host_fallback)
+    return;
+
+  if (num_waits > 0)
+    {
+      va_list ap;
+
+      va_start (ap, num_waits);
+
+      goacc_wait (async, num_waits, ap);
+
+      va_end (ap);
+    }
+
+  acc_dev->openacc.async_set_async_func (async);
+
+  for (i = 0; i < mapnum; ++i)
+    {
+      unsigned char kind = kinds[i] & 0xff;
+
+      switch (kind)
+	{
+	case GOMP_MAP_POINTER:
+	case GOMP_MAP_TO_PSET:
+	  break;
+
+	case GOMP_MAP_FORCE_TO:
+	  acc_update_device (hostaddrs[i], sizes[i]);
+	  break;
+
+	case GOMP_MAP_FORCE_FROM:
+	  acc_update_self (hostaddrs[i], sizes[i]);
+	  break;
+
+	default:
+	  gomp_fatal (">>>> GOACC_update UNHANDLED kind 0x%.2x", kind);
+	  break;
+	}
+    }
+
+  acc_dev->openacc.async_set_async_func (acc_async_sync);
+}
+
+void
+GOACC_wait (int async, int num_waits, ...)
+{
+  va_list ap;
+
+  va_start (ap, num_waits);
+
+  goacc_wait (async, num_waits, ap);
+
+  va_end (ap);
+}
+
+int
+GOACC_get_num_threads (void)
+{
+  return 1;
+}
+
+int
+GOACC_get_thread_num (void)
+{
+  return 0;
+}
diff --git a/libgomp/oacc-plugin.c b/libgomp/oacc-plugin.c
new file mode 100644
index 00000000000..44cd6d68e33
--- /dev/null
+++ b/libgomp/oacc-plugin.c
@@ -0,0 +1,48 @@
+/* Copyright (C) 2014-2015 Free Software Foundation, Inc.
+
+   Contributed by Mentor Embedded.
+
+   This file is part of the GNU Offloading and Multi Processing Library
+   (libgomp).
+
+   Libgomp is free software; you can redistribute it and/or modify it
+   under the terms of the GNU General Public License as published by
+   the Free Software Foundation; either version 3, or (at your option)
+   any later version.
+
+   Libgomp is distributed in the hope that it will be useful, but WITHOUT ANY
+   WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
+   FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+   more details.
+
+   Under Section 7 of GPL version 3, you are granted additional
+   permissions described in the GCC Runtime Library Exception, version
+   3.1, as published by the Free Software Foundation.
+
+   You should have received a copy of the GNU General Public License and
+   a copy of the GCC Runtime Library Exception along with this program;
+   see the files COPYING3 and COPYING.RUNTIME respectively.  If not, see
+   <http://www.gnu.org/licenses/>.  */
+
+/* Initialize and register OpenACC dispatch table from libgomp plugin.  */
+
+#include "libgomp.h"
+#include "oacc-plugin.h"
+#include "oacc-int.h"
+
+void
+GOMP_PLUGIN_async_unmap_vars (void *ptr)
+{
+  struct target_mem_desc *tgt = ptr;
+
+  gomp_unmap_vars (tgt, false);
+}
+
+/* Return the target-specific part of the TLS data for the current thread.  */
+
+void *
+GOMP_PLUGIN_acc_thread (void)
+{
+  struct goacc_thread *thr = goacc_thread ();
+  return thr ? thr->target_tls : NULL;
+}
diff --git a/libgomp/libgomp_target.h b/libgomp/oacc-plugin.h
index 2e18a64b164..c60eb9c5ddd 100644
--- a/libgomp/libgomp_target.h
+++ b/libgomp/oacc-plugin.h
@@ -1,5 +1,7 @@
 /* Copyright (C) 2014-2015 Free Software Foundation, Inc.
 
+   Contributed by Mentor Embedded.
+
    This file is part of the GNU Offloading and Multi Processing Library
    (libgomp).
 
@@ -22,24 +24,10 @@
    see the files COPYING3 and COPYING.RUNTIME respectively.  If not, see
    <http://www.gnu.org/licenses/>.  */
 
-#ifndef LIBGOMP_TARGET_H
-#define LIBGOMP_TARGET_H 1
-
-/* Type of offload target device.  */
-enum offload_target_type
-{
-  OFFLOAD_TARGET_TYPE_HOST,
-  OFFLOAD_TARGET_TYPE_INTEL_MIC
-};
-
-/* Auxiliary struct, used for transferring a host-target address range mapping
-   from plugin to libgomp.  */
-struct mapping_table
-{
-  uintptr_t host_start;
-  uintptr_t host_end;
-  uintptr_t tgt_start;
-  uintptr_t tgt_end;
-};
-
-#endif /* LIBGOMP_TARGET_H */
+#ifndef OACC_PLUGIN_H
+#define OACC_PLUGIN_H 1
+
+extern void GOMP_PLUGIN_async_unmap_vars (void *);
+extern void *GOMP_PLUGIN_acc_thread (void);
+
+#endif
diff --git a/libgomp/oacc-ptx.h b/libgomp/oacc-ptx.h
new file mode 100644
index 00000000000..13ff86fb4af
--- /dev/null
+++ b/libgomp/oacc-ptx.h
@@ -0,0 +1,202 @@
+/* Copyright (C) 2014-2015 Free Software Foundation, Inc.
+
+   Contributed by Mentor Embedded.
+
+   This file is part of the GNU Offloading and Multi Processing Library
+   (libgomp).
+
+   Libgomp is free software; you can redistribute it and/or modify it
+   under the terms of the GNU General Public License as published by
+   the Free Software Foundation; either version 3, or (at your option)
+   any later version.
+
+   Libgomp is distributed in the hope that it will be useful, but WITHOUT ANY
+   WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
+   FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+   more details.
+
+   Under Section 7 of GPL version 3, you are granted additional
+   permissions described in the GCC Runtime Library Exception, version
+   3.1, as published by the Free Software Foundation.
+
+   You should have received a copy of the GNU General Public License and
+   a copy of the GCC Runtime Library Exception along with this program;
+   see the files COPYING3 and COPYING.RUNTIME respectively.  If not, see
+   <http://www.gnu.org/licenses/>.  */
+
+#define ABORT_PTX				\
+  ".version 3.1\n"				\
+  ".target sm_30\n"				\
+  ".address_size 64\n"				\
+  ".visible .func abort;\n"			\
+  ".visible .func abort\n"			\
+  "{\n"						\
+  "trap;\n"					\
+  "ret;\n"					\
+  "}\n"						\
+  ".visible .func _gfortran_abort;\n"		\
+  ".visible .func _gfortran_abort\n"		\
+  "{\n"						\
+  "trap;\n"					\
+  "ret;\n"					\
+  "}\n" \
+
+/* Generated with:
+
+   $ echo 'int acc_on_device(int d) { return __builtin_acc_on_device(d); } int acc_on_device_h_(int *d) { return acc_on_device(*d); }' | accel-gcc/xgcc -Baccel-gcc -x c - -o - -S -m64 -O3 -fno-builtin-acc_on_device -fno-inline
+*/
+#define ACC_ON_DEVICE_PTX						\
+  "        .version        3.1\n"					\
+  "        .target sm_30\n"						\
+  "        .address_size 64\n"						\
+  ".visible .func (.param.u32 %out_retval)acc_on_device(.param.u32 %in_ar1);\n" \
+  ".visible .func (.param.u32 %out_retval)acc_on_device(.param.u32 %in_ar1)\n" \
+  "{\n"									\
+  "        .reg.u32 %ar1;\n"						\
+  ".reg.u32 %retval;\n"							\
+  "        .reg.u64 %hr10;\n"						\
+  "        .reg.u32 %r24;\n"						\
+  "        .reg.u32 %r25;\n"						\
+  "        .reg.pred %r27;\n"						\
+  "        .reg.u32 %r30;\n"						\
+  "        ld.param.u32 %ar1, [%in_ar1];\n"				\
+  "                mov.u32 %r24, %ar1;\n"				\
+  "                setp.ne.u32 %r27,%r24,4;\n"				\
+  "                set.u32.eq.u32 %r30,%r24,5;\n"			\
+  "                neg.s32 %r25, %r30;\n"				\
+  "        @%r27   bra     $L3;\n"					\
+  "                mov.u32 %r25, 1;\n"					\
+  "$L3:\n"								\
+  "                mov.u32 %retval, %r25;\n"				\
+  "        st.param.u32    [%out_retval], %retval;\n"			\
+  "        ret;\n"							\
+  "        }\n"								\
+  ".visible .func (.param.u32 %out_retval)acc_on_device_h_(.param.u64 %in_ar1);\n" \
+  ".visible .func (.param.u32 %out_retval)acc_on_device_h_(.param.u64 %in_ar1)\n" \
+  "{\n"									\
+  "        .reg.u64 %ar1;\n"						\
+  ".reg.u32 %retval;\n"							\
+  "        .reg.u64 %hr10;\n"						\
+  "        .reg.u64 %r25;\n"						\
+  "        .reg.u32 %r26;\n"						\
+  "        .reg.u32 %r27;\n"						\
+  "        ld.param.u64 %ar1, [%in_ar1];\n"				\
+  "                mov.u64 %r25, %ar1;\n"				\
+  "                ld.u32  %r26, [%r25];\n"				\
+  "        {\n"								\
+  "                .param.u32 %retval_in;\n"				\
+  "        {\n"								\
+  "                .param.u32 %out_arg0;\n"				\
+  "                st.param.u32 [%out_arg0], %r26;\n"			\
+  "                call (%retval_in), acc_on_device, (%out_arg0);\n"	\
+  "        }\n"								\
+  "                ld.param.u32    %r27, [%retval_in];\n"		\
+  "}\n"									\
+  "                mov.u32 %retval, %r27;\n"				\
+  "        st.param.u32    [%out_retval], %retval;\n"			\
+  "        ret;\n"							\
+  "        }"
+
+ #define GOACC_INTERNAL_PTX						\
+  ".version 3.1\n" \
+  ".target sm_30\n" \
+  ".address_size 64\n" \
+  ".visible .func (.param .u32 %out_retval) GOACC_get_num_threads;\n" \
+  ".visible .func (.param .u32 %out_retval) GOACC_get_thread_num;\n" \
+  ".extern .func abort;\n" \
+  ".visible .func (.param .u32 %out_retval) GOACC_get_num_threads\n"	\
+  "{\n"									\
+  ".reg .u32 %retval;\n"						\
+  ".reg .u64 %hr10;\n"							\
+  ".reg .u32 %r22;\n"							\
+  ".reg .u32 %r23;\n"							\
+  ".reg .u32 %r24;\n"							\
+  ".reg .u32 %r25;\n"							\
+  ".reg .u32 %r26;\n"							\
+  ".reg .u32 %r27;\n"							\
+  ".reg .u32 %r28;\n"							\
+  ".reg .u32 %r29;\n"							\
+  "mov.u32 %r26,0;\n"							\
+  "{\n"									\
+  ".param .u32 %retval_in;\n"						\
+  "{\n"									\
+  ".param .u32 %out_arg0;\n"						\
+  "st.param.u32 [%out_arg0],%r26;\n"					\
+  "call (%retval_in),GOACC_ntid,(%out_arg0);\n"				\
+  "}\n"									\
+  "ld.param.u32 %r27,[%retval_in];\n"					\
+  "}\n"									\
+  "mov.u32 %r22,%r27;\n"						\
+  "mov.u32 %r28,0;\n"							\
+  "{\n"									\
+  ".param .u32 %retval_in;\n"						\
+  "{\n"									\
+  ".param .u32 %out_arg0;\n"						\
+  "st.param.u32 [%out_arg0],%r28;\n"					\
+  "call (%retval_in),GOACC_nctaid,(%out_arg0);\n"			\
+  "}\n"									\
+  "ld.param.u32 %r29,[%retval_in];\n"					\
+  "}\n"									\
+  "mov.u32 %r23,%r29;\n"						\
+  "mul.lo.u32 %r24,%r22,%r23;\n"					\
+  "mov.u32 %r25,%r24;\n"						\
+  "mov.u32 %retval,%r25;\n"						\
+  "st.param.u32 [%out_retval],%retval;\n"				\
+  "ret;\n"								\
+  "}\n"									\
+  ".visible .func (.param .u32 %out_retval) GOACC_get_thread_num\n"	\
+  "{\n"									\
+  ".reg .u32 %retval;\n"						\
+  ".reg .u64 %hr10;\n"							\
+  ".reg .u32 %r22;\n"							\
+  ".reg .u32 %r23;\n"							\
+  ".reg .u32 %r24;\n"							\
+  ".reg .u32 %r25;\n"							\
+  ".reg .u32 %r26;\n"							\
+  ".reg .u32 %r27;\n"							\
+  ".reg .u32 %r28;\n"							\
+  ".reg .u32 %r29;\n"							\
+  ".reg .u32 %r30;\n"							\
+  ".reg .u32 %r31;\n"							\
+  ".reg .u32 %r32;\n"							\
+  ".reg .u32 %r33;\n"							\
+  "mov.u32 %r28,0;\n"							\
+  "{\n"									\
+  ".param .u32 %retval_in;\n"						\
+  "{\n"									\
+  ".param .u32 %out_arg0;\n"						\
+  "st.param.u32 [%out_arg0],%r28;\n"					\
+  "call (%retval_in),GOACC_ntid,(%out_arg0);\n"				\
+  "}\n"									\
+  "ld.param.u32 %r29,[%retval_in];\n"					\
+  "}\n"									\
+  "mov.u32 %r22,%r29;\n"						\
+  "mov.u32 %r30,0;\n"							\
+  "{\n"									\
+  ".param .u32 %retval_in;\n"						\
+  "{\n"									\
+  ".param .u32 %out_arg0;\n"						\
+  "st.param.u32 [%out_arg0],%r30;\n"					\
+  "call (%retval_in),GOACC_ctaid,(%out_arg0);\n"			\
+  "}\n"									\
+  "ld.param.u32 %r31,[%retval_in];\n"					\
+  "}\n"									\
+  "mov.u32 %r23,%r31;\n"						\
+  "mul.lo.u32 %r24,%r22,%r23;\n"					\
+  "mov.u32 %r32,0;\n"							\
+  "{\n"									\
+  ".param .u32 %retval_in;\n"						\
+  "{\n"									\
+  ".param .u32 %out_arg0;\n"						\
+  "st.param.u32 [%out_arg0],%r32;\n"					\
+  "call (%retval_in),GOACC_tid,(%out_arg0);\n"				\
+  "}\n"									\
+  "ld.param.u32 %r33,[%retval_in];\n"					\
+  "}\n"									\
+  "mov.u32 %r25,%r33;\n"						\
+  "add.u32 %r26,%r24,%r25;\n"						\
+  "mov.u32 %r27,%r26;\n"						\
+  "mov.u32 %retval,%r27;\n"						\
+  "st.param.u32 [%out_retval],%retval;\n"				\
+  "ret;\n"								\
+  "}\n"
diff --git a/libgomp/openacc.f90 b/libgomp/openacc.f90
new file mode 100644
index 00000000000..04d80886a98
--- /dev/null
+++ b/libgomp/openacc.f90
@@ -0,0 +1,956 @@
+!  OpenACC Runtime Library Definitions.
+
+!  Copyright (C) 2014-2015 Free Software Foundation, Inc.
+
+!  Contributed by Tobias Burnus <burnus@net-b.de>
+!              and Mentor Embedded.
+
+!  This file is part of the GNU Offloading and Multi Processing Library
+!  (libgomp).
+
+!  Libgomp is free software; you can redistribute it and/or modify it
+!  under the terms of the GNU General Public License as published by
+!  the Free Software Foundation; either version 3, or (at your option)
+!  any later version.
+
+!  Libgomp is distributed in the hope that it will be useful, but WITHOUT ANY
+!  WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
+!  FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+!  more details.
+
+!  Under Section 7 of GPL version 3, you are granted additional
+!  permissions described in the GCC Runtime Library Exception, version
+!  3.1, as published by the Free Software Foundation.
+
+!  You should have received a copy of the GNU General Public License and
+!  a copy of the GCC Runtime Library Exception along with this program;
+!  see the files COPYING3 and COPYING.RUNTIME respectively.  If not, see
+!  <http://www.gnu.org/licenses/>.
+
+module openacc_kinds
+  use iso_fortran_env, only: int32
+  implicit none
+
+  private :: int32
+  public :: acc_device_kind
+
+  integer, parameter :: acc_device_kind = int32
+
+  public :: acc_device_none, acc_device_default, acc_device_host
+  public :: acc_device_not_host, acc_device_nvidia
+
+  ! Keep in sync with include/gomp-constants.h.
+  integer (acc_device_kind), parameter :: acc_device_none = 0
+  integer (acc_device_kind), parameter :: acc_device_default = 1
+  integer (acc_device_kind), parameter :: acc_device_host = 2
+  integer (acc_device_kind), parameter :: acc_device_host_nonshm = 3
+  integer (acc_device_kind), parameter :: acc_device_not_host = 4
+  integer (acc_device_kind), parameter :: acc_device_nvidia = 5
+
+  public :: acc_handle_kind
+
+  integer, parameter :: acc_handle_kind = int32
+
+  public :: acc_async_noval, acc_async_sync
+
+  ! Keep in sync with include/gomp-constants.h.
+  integer (acc_handle_kind), parameter :: acc_async_noval = -1
+  integer (acc_handle_kind), parameter :: acc_async_sync = -2
+
+end module
+
+module openacc_internal
+  use openacc_kinds
+  implicit none
+
+  interface
+    function acc_get_num_devices_h (d)
+      import
+      integer acc_get_num_devices_h
+      integer (acc_device_kind) d
+    end function
+
+    subroutine acc_set_device_type_h (d)
+      import
+      integer (acc_device_kind) d
+    end subroutine
+
+    function acc_get_device_type_h ()
+      import
+      integer (acc_device_kind) acc_get_device_type_h
+    end function
+
+    subroutine acc_set_device_num_h (n, d)
+      import
+      integer n
+      integer (acc_device_kind) d
+    end subroutine
+
+    function acc_get_device_num_h (d)
+      import
+      integer acc_get_device_num_h
+      integer (acc_device_kind) d
+    end function
+
+    function acc_async_test_h (a)
+      logical acc_async_test_h
+      integer a
+    end function
+
+    function acc_async_test_all_h ()
+      logical acc_async_test_all_h
+    end function
+
+    subroutine acc_wait_h (a)
+      integer a
+    end subroutine
+
+    subroutine acc_wait_async_h (a1, a2)
+      integer a1, a2
+    end subroutine
+
+    subroutine acc_wait_all_h ()
+    end subroutine
+
+    subroutine acc_wait_all_async_h (a)
+      integer a
+    end subroutine
+
+    subroutine acc_init_h (d)
+      import
+      integer (acc_device_kind) d
+    end subroutine
+
+    subroutine acc_shutdown_h (d)
+      import
+      integer (acc_device_kind) d
+    end subroutine
+
+    function acc_on_device_h (d)
+      import
+      integer (acc_device_kind) d
+      logical acc_on_device_h
+    end function
+
+    subroutine acc_copyin_32_h (a, len)
+      use iso_c_binding, only: c_int32_t
+      !GCC$ ATTRIBUTES NO_ARG_CHECK :: a
+      type (*), dimension (*) :: a
+      integer (c_int32_t) len
+    end subroutine
+
+    subroutine acc_copyin_64_h (a, len)
+      use iso_c_binding, only: c_int64_t
+      !GCC$ ATTRIBUTES NO_ARG_CHECK :: a
+      type (*), dimension (*) :: a
+      integer (c_int64_t) len
+    end subroutine
+
+    subroutine acc_copyin_array_h (a)
+      type (*), dimension (..), contiguous :: a
+    end subroutine
+
+    subroutine acc_present_or_copyin_32_h (a, len)
+      use iso_c_binding, only: c_int32_t
+      !GCC$ ATTRIBUTES NO_ARG_CHECK :: a
+      type (*), dimension (*) :: a
+      integer (c_int32_t) len
+    end subroutine
+
+    subroutine acc_present_or_copyin_64_h (a, len)
+      use iso_c_binding, only: c_int64_t
+      !GCC$ ATTRIBUTES NO_ARG_CHECK :: a
+      type (*), dimension (*) :: a
+      integer (c_int64_t) len
+    end subroutine
+
+    subroutine acc_present_or_copyin_array_h (a)
+      type (*), dimension (..), contiguous :: a
+    end subroutine
+
+    subroutine acc_create_32_h (a, len)
+      use iso_c_binding, only: c_int32_t
+      !GCC$ ATTRIBUTES NO_ARG_CHECK :: a
+      type (*), dimension (*) :: a
+      integer (c_int32_t) len
+    end subroutine
+
+    subroutine acc_create_64_h (a, len)
+      use iso_c_binding, only: c_int64_t
+      !GCC$ ATTRIBUTES NO_ARG_CHECK :: a
+      type (*), dimension (*) :: a
+      integer (c_int64_t) len
+    end subroutine
+
+    subroutine acc_create_array_h (a)
+      type (*), dimension (..), contiguous :: a
+    end subroutine
+
+    subroutine acc_present_or_create_32_h (a, len)
+      use iso_c_binding, only: c_int32_t
+      !GCC$ ATTRIBUTES NO_ARG_CHECK :: a
+      type (*), dimension (*) :: a
+      integer (c_int32_t) len
+    end subroutine
+
+    subroutine acc_present_or_create_64_h (a, len)
+      use iso_c_binding, only: c_int64_t
+      !GCC$ ATTRIBUTES NO_ARG_CHECK :: a
+      type (*), dimension (*) :: a
+      integer (c_int64_t) len
+    end subroutine
+
+    subroutine acc_present_or_create_array_h (a)
+      type (*), dimension (..), contiguous :: a
+    end subroutine
+
+    subroutine acc_copyout_32_h (a, len)
+      use iso_c_binding, only: c_int32_t
+      !GCC$ ATTRIBUTES NO_ARG_CHECK :: a
+      type (*), dimension (*) :: a
+      integer (c_int32_t) len
+    end subroutine
+
+    subroutine acc_copyout_64_h (a, len)
+      use iso_c_binding, only: c_int64_t
+      !GCC$ ATTRIBUTES NO_ARG_CHECK :: a
+      type (*), dimension (*) :: a
+      integer (c_int64_t) len
+    end subroutine
+
+    subroutine acc_copyout_array_h (a)
+      type (*), dimension (..), contiguous :: a
+    end subroutine
+
+    subroutine acc_delete_32_h (a, len)
+      use iso_c_binding, only: c_int32_t
+      !GCC$ ATTRIBUTES NO_ARG_CHECK :: a
+      type (*), dimension (*) :: a
+      integer (c_int32_t) len
+    end subroutine
+
+    subroutine acc_delete_64_h (a, len)
+      use iso_c_binding, only: c_int64_t
+      !GCC$ ATTRIBUTES NO_ARG_CHECK :: a
+      type (*), dimension (*) :: a
+      integer (c_int64_t) len
+    end subroutine
+
+    subroutine acc_delete_array_h (a)
+      type (*), dimension (..), contiguous :: a
+    end subroutine
+
+    subroutine acc_update_device_32_h (a, len)
+      use iso_c_binding, only: c_int32_t
+      !GCC$ ATTRIBUTES NO_ARG_CHECK :: a
+      type (*), dimension (*) :: a
+      integer (c_int32_t) len
+    end subroutine
+
+    subroutine acc_update_device_64_h (a, len)
+      use iso_c_binding, only: c_int64_t
+      !GCC$ ATTRIBUTES NO_ARG_CHECK :: a
+      type (*), dimension (*) :: a
+      integer (c_int64_t) len
+    end subroutine
+
+    subroutine acc_update_device_array_h (a)
+      type (*), dimension (..), contiguous :: a
+    end subroutine
+
+    subroutine acc_update_self_32_h (a, len)
+      use iso_c_binding, only: c_int32_t
+      !GCC$ ATTRIBUTES NO_ARG_CHECK :: a
+      type (*), dimension (*) :: a
+      integer (c_int32_t) len
+    end subroutine
+
+    subroutine acc_update_self_64_h (a, len)
+      use iso_c_binding, only: c_int64_t
+      !GCC$ ATTRIBUTES NO_ARG_CHECK :: a
+      type (*), dimension (*) :: a
+      integer (c_int64_t) len
+    end subroutine
+
+    subroutine acc_update_self_array_h (a)
+      type (*), dimension (..), contiguous :: a
+    end subroutine
+
+    function acc_is_present_32_h (a, len)
+      use iso_c_binding, only: c_int32_t
+      logical acc_is_present_32_h
+      !GCC$ ATTRIBUTES NO_ARG_CHECK :: a
+      type (*), dimension (*) :: a
+      integer (c_int32_t) len
+    end function
+
+    function acc_is_present_64_h (a, len)
+      use iso_c_binding, only: c_int64_t
+      logical acc_is_present_64_h
+      !GCC$ ATTRIBUTES NO_ARG_CHECK :: a
+      type (*), dimension (*) :: a
+      integer (c_int64_t) len
+    end function
+
+    function acc_is_present_array_h (a)
+      logical acc_is_present_array_h
+      type (*), dimension (..), contiguous :: a
+    end function
+  end interface
+
+  interface
+    function acc_get_num_devices_l (d) &
+        bind (C, name = "acc_get_num_devices")
+      use iso_c_binding, only: c_int
+      integer (c_int) :: acc_get_num_devices_l
+      integer (c_int), value :: d
+    end function
+
+    subroutine acc_set_device_type_l (d) &
+        bind (C, name = "acc_set_device_type")
+      use iso_c_binding, only: c_int
+      integer (c_int), value :: d
+    end subroutine
+
+    function acc_get_device_type_l () &
+        bind (C, name = "acc_get_device_type")
+      use iso_c_binding, only: c_int
+      integer (c_int) :: acc_get_device_type_l
+    end function
+
+    subroutine acc_set_device_num_l (n, d) &
+        bind (C, name = "acc_set_device_num")
+      use iso_c_binding, only: c_int
+      integer (c_int), value :: n, d
+    end subroutine
+
+    function acc_get_device_num_l (d) &
+        bind (C, name = "acc_get_device_num")
+      use iso_c_binding, only: c_int
+      integer (c_int) :: acc_get_device_num_l
+      integer (c_int), value :: d
+    end function
+
+    function acc_async_test_l (a) &
+        bind (C, name = "acc_async_test")
+      use iso_c_binding, only: c_int
+      integer (c_int) :: acc_async_test_l
+      integer (c_int), value :: a
+    end function
+
+    function acc_async_test_all_l () &
+        bind (C, name = "acc_async_test_all")
+      use iso_c_binding, only: c_int
+      integer (c_int) :: acc_async_test_all_l
+    end function
+
+    subroutine acc_wait_l (a) &
+        bind (C, name = "acc_wait")
+      use iso_c_binding, only: c_int
+      integer (c_int), value :: a
+    end subroutine
+
+    subroutine acc_wait_async_l (a1, a2) &
+        bind (C, name = "acc_wait_async")
+      use iso_c_binding, only: c_int
+      integer (c_int), value :: a1, a2
+    end subroutine
+
+    subroutine acc_wait_all_l () &
+        bind (C, name = "acc_wait_all")
+      use iso_c_binding, only: c_int
+    end subroutine
+
+    subroutine acc_wait_all_async_l (a) &
+        bind (C, name = "acc_wait_all_async")
+      use iso_c_binding, only: c_int
+      integer (c_int), value :: a
+    end subroutine
+
+    subroutine acc_init_l (d) &
+        bind (C, name = "acc_init")
+      use iso_c_binding, only: c_int
+      integer (c_int), value :: d
+    end subroutine
+
+    subroutine acc_shutdown_l (d) &
+        bind (C, name = "acc_shutdown")
+      use iso_c_binding, only: c_int
+      integer (c_int), value :: d
+    end subroutine
+
+    function acc_on_device_l (d) &
+        bind (C, name = "acc_on_device")
+      use iso_c_binding, only: c_int
+      integer (c_int) :: acc_on_device_l
+      integer (c_int), value :: d
+    end function
+
+    subroutine acc_copyin_l (a, len) &
+        bind (C, name = "acc_copyin")
+      use iso_c_binding, only: c_size_t
+      !GCC$ ATTRIBUTES NO_ARG_CHECK :: a
+      type (*), dimension (*) :: a
+      integer (c_size_t), value :: len
+    end subroutine
+
+    subroutine acc_present_or_copyin_l (a, len) &
+        bind (C, name = "acc_present_or_copyin")
+      use iso_c_binding, only: c_size_t
+      !GCC$ ATTRIBUTES NO_ARG_CHECK :: a
+      type (*), dimension (*) :: a
+      integer (c_size_t), value :: len
+    end subroutine
+
+    subroutine acc_create_l (a, len) &
+        bind (C, name = "acc_create")
+      use iso_c_binding, only: c_size_t
+      !GCC$ ATTRIBUTES NO_ARG_CHECK :: a
+      type (*), dimension (*) :: a
+      integer (c_size_t), value :: len
+    end subroutine
+
+    subroutine acc_present_or_create_l (a, len) &
+        bind (C, name = "acc_present_or_create")
+      use iso_c_binding, only: c_size_t
+      !GCC$ ATTRIBUTES NO_ARG_CHECK :: a
+      type (*), dimension (*) :: a
+      integer (c_size_t), value :: len
+    end subroutine
+
+    subroutine acc_copyout_l (a, len) &
+        bind (C, name = "acc_copyout")
+      use iso_c_binding, only: c_size_t
+      !GCC$ ATTRIBUTES NO_ARG_CHECK :: a
+      type (*), dimension (*) :: a
+      integer (c_size_t), value :: len
+    end subroutine
+
+    subroutine acc_delete_l (a, len) &
+        bind (C, name = "acc_delete")
+      use iso_c_binding, only: c_size_t
+      !GCC$ ATTRIBUTES NO_ARG_CHECK :: a
+      type (*), dimension (*) :: a
+      integer (c_size_t), value :: len
+    end subroutine
+
+    subroutine acc_update_device_l (a, len) &
+        bind (C, name = "acc_update_device")
+      use iso_c_binding, only: c_size_t
+      !GCC$ ATTRIBUTES NO_ARG_CHECK :: a
+      type (*), dimension (*) :: a
+      integer (c_size_t), value :: len
+    end subroutine
+
+    subroutine acc_update_self_l (a, len) &
+        bind (C, name = "acc_update_self")
+      use iso_c_binding, only: c_size_t
+      !GCC$ ATTRIBUTES NO_ARG_CHECK :: a
+      type (*), dimension (*) :: a
+      integer (c_size_t), value :: len
+    end subroutine
+
+    function acc_is_present_l (a, len) &
+        bind (C, name = "acc_is_present")
+      use iso_c_binding, only: c_int32_t, c_size_t
+      !GCC$ ATTRIBUTES NO_ARG_CHECK :: a
+      integer (c_int32_t) :: acc_is_present_l
+      type (*), dimension (*) :: a
+      integer (c_size_t), value :: len
+    end function
+  end interface
+end module
+
+module openacc
+  use openacc_kinds
+  use openacc_internal
+  implicit none
+
+  public :: openacc_version
+
+  public :: acc_get_num_devices, acc_set_device_type, acc_get_device_type
+  public :: acc_set_device_num, acc_get_device_num, acc_async_test
+  public :: acc_async_test_all, acc_wait, acc_wait_async, acc_wait_all
+  public :: acc_wait_all_async, acc_init, acc_shutdown, acc_on_device
+  public :: acc_copyin, acc_present_or_copyin, acc_pcopyin, acc_create
+  public :: acc_present_or_create, acc_pcreate, acc_copyout, acc_delete
+  public :: acc_update_device, acc_update_self, acc_is_present
+
+  integer, parameter :: openacc_version = 201306
+
+  interface acc_get_num_devices
+    procedure :: acc_get_num_devices_h
+  end interface
+
+  interface acc_set_device_type
+    procedure :: acc_set_device_type_h
+  end interface
+
+  interface acc_get_device_type
+    procedure :: acc_get_device_type_h
+  end interface
+
+  interface acc_set_device_num
+    procedure :: acc_set_device_num_h
+  end interface
+
+  interface acc_get_device_num
+    procedure :: acc_get_device_num_h
+  end interface
+
+  interface acc_async_test
+    procedure :: acc_async_test_h
+  end interface
+
+  interface acc_async_test_all
+    procedure :: acc_async_test_all_h
+  end interface
+
+  interface acc_wait
+    procedure :: acc_wait_h
+  end interface
+
+  interface acc_wait_async
+    procedure :: acc_wait_async_h
+  end interface
+
+  interface acc_wait_all
+    procedure :: acc_wait_all_h
+  end interface
+
+  interface acc_wait_all_async
+    procedure :: acc_wait_all_async_h
+  end interface
+
+  interface acc_init
+    procedure :: acc_init_h
+  end interface
+
+  interface acc_shutdown
+    procedure :: acc_shutdown_h
+  end interface
+
+  interface acc_on_device
+    procedure :: acc_on_device_h
+  end interface
+
+  ! acc_malloc: Only available in C/C++
+  ! acc_free: Only available in C/C++
+
+  ! As vendor extension, the following code supports both 32bit and 64bit
+  ! arguments for "size"; the OpenACC standard only permits default-kind
+  ! integers, which are of kind 4 (i.e. 32 bits).
+  ! Additionally, the two-argument version also takes arrays as argument.
+  ! and the one argument version also scalars. Note that the code assumes
+  ! that the arrays are contiguous.
+
+  interface acc_copyin
+    procedure :: acc_copyin_32_h
+    procedure :: acc_copyin_64_h
+    procedure :: acc_copyin_array_h
+  end interface
+
+  interface acc_present_or_copyin
+    procedure :: acc_present_or_copyin_32_h
+    procedure :: acc_present_or_copyin_64_h
+    procedure :: acc_present_or_copyin_array_h
+  end interface
+
+  interface acc_pcopyin
+    procedure :: acc_present_or_copyin_32_h
+    procedure :: acc_present_or_copyin_64_h
+    procedure :: acc_present_or_copyin_array_h
+  end interface
+
+  interface acc_create
+    procedure :: acc_create_32_h
+    procedure :: acc_create_64_h
+    procedure :: acc_create_array_h
+  end interface
+
+  interface acc_present_or_create
+    procedure :: acc_present_or_create_32_h
+    procedure :: acc_present_or_create_64_h
+    procedure :: acc_present_or_create_array_h
+  end interface
+
+  interface acc_pcreate
+    procedure :: acc_present_or_create_32_h
+    procedure :: acc_present_or_create_64_h
+    procedure :: acc_present_or_create_array_h
+  end interface
+
+  interface acc_copyout
+    procedure :: acc_copyout_32_h
+    procedure :: acc_copyout_64_h
+    procedure :: acc_copyout_array_h
+  end interface
+
+  interface acc_delete
+    procedure :: acc_delete_32_h
+    procedure :: acc_delete_64_h
+    procedure :: acc_delete_array_h
+  end interface
+
+  interface acc_update_device
+    procedure :: acc_update_device_32_h
+    procedure :: acc_update_device_64_h
+    procedure :: acc_update_device_array_h
+  end interface
+
+  interface acc_update_self
+    procedure :: acc_update_self_32_h
+    procedure :: acc_update_self_64_h
+    procedure :: acc_update_self_array_h
+  end interface
+
+  ! acc_map_data: Only available in C/C++
+  ! acc_unmap_data: Only available in C/C++
+  ! acc_deviceptr: Only available in C/C++
+  ! acc_hostptr: Only available in C/C++
+
+  interface acc_is_present
+    procedure :: acc_is_present_32_h
+    procedure :: acc_is_present_64_h
+    procedure :: acc_is_present_array_h
+  end interface
+
+  ! acc_memcpy_to_device: Only available in C/C++
+  ! acc_memcpy_from_device: Only available in C/C++
+
+end module
+
+function acc_get_num_devices_h (d)
+  use openacc_internal, only: acc_get_num_devices_l
+  use openacc_kinds
+  integer acc_get_num_devices_h
+  integer (acc_device_kind) d
+  acc_get_num_devices_h = acc_get_num_devices_l (d)
+end function
+
+subroutine acc_set_device_type_h (d)
+  use openacc_internal, only: acc_set_device_type_l
+  use openacc_kinds
+  integer (acc_device_kind) d
+  call acc_set_device_type_l (d)
+end subroutine
+
+function acc_get_device_type_h ()
+  use openacc_internal, only: acc_get_device_type_l
+  use openacc_kinds
+  integer (acc_device_kind) acc_get_device_type_h
+  acc_get_device_type_h = acc_get_device_type_l ()
+end function
+
+subroutine acc_set_device_num_h (n, d)
+  use openacc_internal, only: acc_set_device_num_l
+  use openacc_kinds
+  integer n
+  integer (acc_device_kind) d
+  call acc_set_device_num_l (n, d)
+end subroutine
+
+function acc_get_device_num_h (d)
+  use openacc_internal, only: acc_get_device_num_l
+  use openacc_kinds
+  integer acc_get_device_num_h
+  integer (acc_device_kind) d
+  acc_get_device_num_h = acc_get_device_num_l (d)
+end function
+
+function acc_async_test_h (a)
+  use openacc_internal, only: acc_async_test_l
+  logical acc_async_test_h
+  integer a
+  if (acc_async_test_l (a) .eq. 1) then
+    acc_async_test_h = .TRUE.
+  else
+    acc_async_test_h = .FALSE.
+  end if
+end function
+
+function acc_async_test_all_h ()
+  use openacc_internal, only: acc_async_test_all_l
+  logical acc_async_test_all_h
+  if (acc_async_test_all_l () .eq. 1) then
+    acc_async_test_all_h = .TRUE.
+  else
+    acc_async_test_all_h = .FALSE.
+  end if
+end function
+
+subroutine acc_wait_h (a)
+  use openacc_internal, only: acc_wait_l
+  integer a
+  call acc_wait_l (a)
+end subroutine
+
+subroutine acc_wait_async_h (a1, a2)
+  use openacc_internal, only: acc_wait_async_l
+  integer a1, a2
+  call acc_wait_async_l (a1, a2)
+end subroutine
+
+subroutine acc_wait_all_h ()
+  use openacc_internal, only: acc_wait_all_l
+  call acc_wait_all_l ()
+end subroutine
+
+subroutine acc_wait_all_async_h (a)
+  use openacc_internal, only: acc_wait_all_async_l
+  integer a
+  call acc_wait_all_async_l (a)
+end subroutine
+
+subroutine acc_init_h (d)
+  use openacc_internal, only: acc_init_l
+  use openacc_kinds
+  integer (acc_device_kind) d
+  call acc_init_l (d)
+end subroutine
+
+subroutine acc_shutdown_h (d)
+  use openacc_internal, only: acc_shutdown_l
+  use openacc_kinds
+  integer (acc_device_kind) d
+  call acc_shutdown_l (d)
+end subroutine
+
+function acc_on_device_h (d)
+  use openacc_internal, only: acc_on_device_l
+  use openacc_kinds
+  integer (acc_device_kind) d
+  logical acc_on_device_h
+  if (acc_on_device_l (d) .eq. 1) then
+    acc_on_device_h = .TRUE.
+  else
+    acc_on_device_h = .FALSE.
+  end if
+end function
+
+subroutine acc_copyin_32_h (a, len)
+  use iso_c_binding, only: c_int32_t, c_size_t
+  use openacc_internal, only: acc_copyin_l
+  !GCC$ ATTRIBUTES NO_ARG_CHECK :: a
+  type (*), dimension (*) :: a
+  integer (c_int32_t) len
+  call acc_copyin_l (a, int (len, kind = c_size_t))
+end subroutine
+
+subroutine acc_copyin_64_h (a, len)
+  use iso_c_binding, only: c_int64_t, c_size_t
+  use openacc_internal, only: acc_copyin_l
+  !GCC$ ATTRIBUTES NO_ARG_CHECK :: a
+  type (*), dimension (*) :: a
+  integer (c_int64_t) len
+  call acc_copyin_l (a, int (len, kind = c_size_t))
+end subroutine
+
+subroutine acc_copyin_array_h (a)
+  use openacc_internal, only: acc_copyin_l
+  type (*), dimension (..), contiguous :: a
+  call acc_copyin_l (a, sizeof (a))
+end subroutine
+
+subroutine acc_present_or_copyin_32_h (a, len)
+  use iso_c_binding, only: c_int32_t, c_size_t
+  use openacc_internal, only: acc_present_or_copyin_l
+  !GCC$ ATTRIBUTES NO_ARG_CHECK :: a
+  type (*), dimension (*) :: a
+  integer (c_int32_t) len
+  call acc_present_or_copyin_l (a, int (len, kind = c_size_t))
+end subroutine
+
+subroutine acc_present_or_copyin_64_h (a, len)
+  use iso_c_binding, only: c_int64_t, c_size_t
+  use openacc_internal, only: acc_present_or_copyin_l
+  !GCC$ ATTRIBUTES NO_ARG_CHECK :: a
+  type (*), dimension (*) :: a
+  integer (c_int64_t) len
+  call acc_present_or_copyin_l (a, int (len, kind = c_size_t))
+end subroutine
+
+subroutine acc_present_or_copyin_array_h (a)
+  use openacc_internal, only: acc_present_or_copyin_l
+  type (*), dimension (..), contiguous :: a
+  call acc_present_or_copyin_l (a, sizeof (a))
+end subroutine
+
+subroutine acc_create_32_h (a, len)
+  use iso_c_binding, only: c_int32_t, c_size_t
+  use openacc_internal, only: acc_create_l
+  !GCC$ ATTRIBUTES NO_ARG_CHECK :: a
+  type (*), dimension (*) :: a
+  integer (c_int32_t) len
+  call acc_create_l (a, int (len, kind = c_size_t))
+end subroutine
+
+subroutine acc_create_64_h (a, len)
+  use iso_c_binding, only: c_int64_t, c_size_t
+  use openacc_internal, only: acc_create_l
+  !GCC$ ATTRIBUTES NO_ARG_CHECK :: a
+  type (*), dimension (*) :: a
+  integer (c_int64_t) len
+  call acc_create_l (a, int (len, kind = c_size_t))
+end subroutine
+
+subroutine acc_create_array_h (a)
+  use openacc_internal, only: acc_create_l
+  type (*), dimension (..), contiguous :: a
+  call acc_create_l (a, sizeof (a))
+end subroutine
+
+subroutine acc_present_or_create_32_h (a, len)
+  use iso_c_binding, only: c_int32_t, c_size_t
+  use openacc_internal, only: acc_present_or_create_l
+  !GCC$ ATTRIBUTES NO_ARG_CHECK :: a
+  type (*), dimension (*) :: a
+  integer (c_int32_t) len
+  call acc_present_or_create_l (a, int (len, kind = c_size_t))
+end subroutine
+
+subroutine acc_present_or_create_64_h (a, len)
+  use iso_c_binding, only: c_int64_t, c_size_t
+  use openacc_internal, only: acc_present_or_create_l
+  !GCC$ ATTRIBUTES NO_ARG_CHECK :: a
+  type (*), dimension (*) :: a
+  integer (c_int64_t) len
+  call acc_present_or_create_l (a, int (len, kind = c_size_t))
+end subroutine
+
+subroutine acc_present_or_create_array_h (a)
+  use openacc_internal, only: acc_present_or_create_l
+  type (*), dimension (..), contiguous :: a
+  call acc_present_or_create_l (a, sizeof (a))
+end subroutine
+
+subroutine acc_copyout_32_h (a, len)
+  use iso_c_binding, only: c_int32_t, c_size_t
+  use openacc_internal, only: acc_copyout_l
+  !GCC$ ATTRIBUTES NO_ARG_CHECK :: a
+  type (*), dimension (*) :: a
+  integer (c_int32_t) len
+  call acc_copyout_l (a, int (len, kind = c_size_t))
+end subroutine
+
+subroutine acc_copyout_64_h (a, len)
+  use iso_c_binding, only: c_int64_t, c_size_t
+  use openacc_internal, only: acc_copyout_l
+  !GCC$ ATTRIBUTES NO_ARG_CHECK :: a
+  type (*), dimension (*) :: a
+  integer (c_int64_t) len
+  call acc_copyout_l (a, int (len, kind = c_size_t))
+end subroutine
+
+subroutine acc_copyout_array_h (a)
+  use openacc_internal, only: acc_copyout_l
+  type (*), dimension (..), contiguous :: a
+  call acc_copyout_l (a, sizeof (a))
+end subroutine
+
+subroutine acc_delete_32_h (a, len)
+  use iso_c_binding, only: c_int32_t, c_size_t
+  use openacc_internal, only: acc_delete_l
+  !GCC$ ATTRIBUTES NO_ARG_CHECK :: a
+  type (*), dimension (*) :: a
+  integer (c_int32_t) len
+  call acc_delete_l (a, int (len, kind = c_size_t))
+end subroutine
+
+subroutine acc_delete_64_h (a, len)
+  use iso_c_binding, only: c_int64_t, c_size_t
+  use openacc_internal, only: acc_delete_l
+  !GCC$ ATTRIBUTES NO_ARG_CHECK :: a
+  type (*), dimension (*) :: a
+  integer (c_int64_t) len
+  call acc_delete_l (a, int (len, kind = c_size_t))
+end subroutine
+
+subroutine acc_delete_array_h (a)
+  use openacc_internal, only: acc_delete_l
+  type (*), dimension (..), contiguous :: a
+  call acc_delete_l (a, sizeof (a))
+end subroutine
+
+subroutine acc_update_device_32_h (a, len)
+  use iso_c_binding, only: c_int32_t, c_size_t
+  use openacc_internal, only: acc_update_device_l
+  !GCC$ ATTRIBUTES NO_ARG_CHECK :: a
+  type (*), dimension (*) :: a
+  integer (c_int32_t) len
+  call acc_update_device_l (a, int (len, kind = c_size_t))
+end subroutine
+
+subroutine acc_update_device_64_h (a, len)
+  use iso_c_binding, only: c_int64_t, c_size_t
+  use openacc_internal, only: acc_update_device_l
+  !GCC$ ATTRIBUTES NO_ARG_CHECK :: a
+  type (*), dimension (*) :: a
+  integer (c_int64_t) len
+  call acc_update_device_l (a, int (len, kind = c_size_t))
+end subroutine
+
+subroutine acc_update_device_array_h (a)
+  use openacc_internal, only: acc_update_device_l
+  type (*), dimension (..), contiguous :: a
+  call acc_update_device_l (a, sizeof (a))
+end subroutine
+
+subroutine acc_update_self_32_h (a, len)
+  use iso_c_binding, only: c_int32_t, c_size_t
+  use openacc_internal, only: acc_update_self_l
+  !GCC$ ATTRIBUTES NO_ARG_CHECK :: a
+  type (*), dimension (*) :: a
+  integer (c_int32_t) len
+  call acc_update_self_l (a, int (len, kind = c_size_t))
+end subroutine
+
+subroutine acc_update_self_64_h (a, len)
+  use iso_c_binding, only: c_int64_t, c_size_t
+  use openacc_internal, only: acc_update_self_l
+  !GCC$ ATTRIBUTES NO_ARG_CHECK :: a
+  type (*), dimension (*) :: a
+  integer (c_int64_t) len
+  call acc_update_self_l (a, int (len, kind = c_size_t))
+end subroutine
+
+subroutine acc_update_self_array_h (a)
+  use openacc_internal, only: acc_update_self_l
+  type (*), dimension (..), contiguous :: a
+  call acc_update_self_l (a, sizeof (a))
+end subroutine
+
+function acc_is_present_32_h (a, len)
+  use iso_c_binding, only: c_int32_t, c_size_t
+  use openacc_internal, only: acc_is_present_l
+  logical acc_is_present_32_h
+  !GCC$ ATTRIBUTES NO_ARG_CHECK :: a
+  type (*), dimension (*) :: a
+  integer (c_int32_t) len
+  if (acc_is_present_l (a, int (len, kind = c_size_t)) .eq. 1) then
+    acc_is_present_32_h = .TRUE.
+  else
+    acc_is_present_32_h = .FALSE.
+  end if
+end function
+
+function acc_is_present_64_h (a, len)
+  use iso_c_binding, only: c_int64_t, c_size_t
+  use openacc_internal, only: acc_is_present_l
+  logical acc_is_present_64_h
+  !GCC$ ATTRIBUTES NO_ARG_CHECK :: a
+  type (*), dimension (*) :: a
+  integer (c_int64_t) len
+  if (acc_is_present_l (a, int (len, kind = c_size_t)) .eq. 1) then
+    acc_is_present_64_h = .TRUE.
+  else
+    acc_is_present_64_h = .FALSE.
+  end if
+end function
+
+function acc_is_present_array_h (a)
+  use openacc_internal, only: acc_is_present_l
+  logical acc_is_present_array_h
+  type (*), dimension (..), contiguous :: a
+  acc_is_present_array_h = acc_is_present_l (a, sizeof (a)) == 1
+end function
diff --git a/libgomp/openacc.h b/libgomp/openacc.h
new file mode 100644
index 00000000000..334324114f1
--- /dev/null
+++ b/libgomp/openacc.h
@@ -0,0 +1,118 @@
+/* OpenACC Runtime Library User-facing Declarations
+
+   Copyright (C) 2013-2015 Free Software Foundation, Inc.
+
+   Contributed by Mentor Embedded.
+
+   This file is part of the GNU Offloading and Multi Processing Library
+   (libgomp).
+
+   Libgomp is free software; you can redistribute it and/or modify it
+   under the terms of the GNU General Public License as published by
+   the Free Software Foundation; either version 3, or (at your option)
+   any later version.
+
+   Libgomp is distributed in the hope that it will be useful, but WITHOUT ANY
+   WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
+   FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+   more details.
+
+   Under Section 7 of GPL version 3, you are granted additional
+   permissions described in the GCC Runtime Library Exception, version
+   3.1, as published by the Free Software Foundation.
+
+   You should have received a copy of the GNU General Public License and
+   a copy of the GCC Runtime Library Exception along with this program;
+   see the files COPYING3 and COPYING.RUNTIME respectively.  If not, see
+   <http://www.gnu.org/licenses/>.  */
+
+#ifndef _OPENACC_H
+#define _OPENACC_H 1
+
+/* The OpenACC standard is silent on whether or not including <openacc.h>
+   might or must not include other header files.  We chose to include
+   some.  */
+#include <stddef.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#if __cplusplus >= 201103
+# define __GOACC_NOTHROW noexcept ()
+#elif __cplusplus
+# define __GOACC_NOTHROW throw ()
+#else /* Not C++ */
+# define __GOACC_NOTHROW __attribute__ ((__nothrow__))
+#endif
+
+/* Types */
+typedef enum acc_device_t
+  {
+    /* Keep in sync with include/gomp-constants.h.  */
+    acc_device_none = 0,
+    acc_device_default = 1,
+    acc_device_host = 2,
+    acc_device_host_nonshm = 3,
+    acc_device_not_host = 4,
+    acc_device_nvidia = 5,
+    _ACC_device_hwm
+  } acc_device_t;
+
+typedef enum acc_async_t
+  {
+    /* Keep in sync with include/gomp-constants.h.  */
+    acc_async_noval = -1,
+    acc_async_sync  = -2
+  } acc_async_t;
+
+int acc_get_num_devices (acc_device_t) __GOACC_NOTHROW;
+void acc_set_device_type (acc_device_t) __GOACC_NOTHROW;
+acc_device_t acc_get_device_type (void) __GOACC_NOTHROW;
+void acc_set_device_num (int, acc_device_t) __GOACC_NOTHROW;
+int acc_get_device_num (acc_device_t) __GOACC_NOTHROW;
+int acc_async_test (int) __GOACC_NOTHROW;
+int acc_async_test_all (void) __GOACC_NOTHROW;
+void acc_wait (int) __GOACC_NOTHROW;
+void acc_wait_async (int, int) __GOACC_NOTHROW;
+void acc_wait_all (void) __GOACC_NOTHROW;
+void acc_wait_all_async (int) __GOACC_NOTHROW;
+void acc_init (acc_device_t) __GOACC_NOTHROW;
+void acc_shutdown (acc_device_t) __GOACC_NOTHROW;
+int acc_on_device (acc_device_t) __GOACC_NOTHROW;
+void *acc_malloc (size_t) __GOACC_NOTHROW;
+void acc_free (void *) __GOACC_NOTHROW;
+/* Some of these would be more correct with const qualifiers, but
+   the standard specifies otherwise.  */
+void *acc_copyin (void *, size_t) __GOACC_NOTHROW;
+void *acc_present_or_copyin (void *, size_t) __GOACC_NOTHROW;
+void *acc_create (void *, size_t) __GOACC_NOTHROW;
+void *acc_present_or_create (void *, size_t) __GOACC_NOTHROW;
+void acc_copyout (void *, size_t) __GOACC_NOTHROW;
+void acc_delete (void *, size_t) __GOACC_NOTHROW;
+void acc_update_device (void *, size_t) __GOACC_NOTHROW;
+void acc_update_self (void *, size_t) __GOACC_NOTHROW;
+void acc_map_data (void *, void *, size_t) __GOACC_NOTHROW;
+void acc_unmap_data (void *) __GOACC_NOTHROW;
+void *acc_deviceptr (void *) __GOACC_NOTHROW;
+void *acc_hostptr (void *) __GOACC_NOTHROW;
+int acc_is_present (void *, size_t) __GOACC_NOTHROW;
+void acc_memcpy_to_device (void *, void *, size_t) __GOACC_NOTHROW;
+void acc_memcpy_from_device (void *, void *, size_t) __GOACC_NOTHROW;
+
+/* Old names.  OpenACC does not specify whether these can or must
+   not be macros, inlines or aliases for the new names.  */
+#define acc_pcreate acc_present_or_create
+#define acc_pcopyin acc_present_or_copyin
+
+/* CUDA-specific routines.  */
+void *acc_get_current_cuda_device (void) __GOACC_NOTHROW;
+void *acc_get_current_cuda_context (void) __GOACC_NOTHROW;
+void *acc_get_cuda_stream (int) __GOACC_NOTHROW;
+int acc_set_cuda_stream (int, void *) __GOACC_NOTHROW;
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _OPENACC_H */
diff --git a/libgomp/openacc_lib.h b/libgomp/openacc_lib.h
new file mode 100644
index 00000000000..28659a1e0b0
--- /dev/null
+++ b/libgomp/openacc_lib.h
@@ -0,0 +1,381 @@
+!  OpenACC Runtime Library Definitions.			-*- mode: fortran -*-
+
+!  Copyright (C) 2014-2015 Free Software Foundation, Inc.
+
+!  Contributed by Tobias Burnus <burnus@net-b.de>
+!              and Mentor Embedded.
+
+!  This file is part of the GNU Offloading and Multi Processing Library
+!  (libgomp).
+
+!  Libgomp is free software; you can redistribute it and/or modify it
+!  under the terms of the GNU General Public License as published by
+!  the Free Software Foundation; either version 3, or (at your option)
+!  any later version.
+
+!  Libgomp is distributed in the hope that it will be useful, but WITHOUT ANY
+!  WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
+!  FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+!  more details.
+
+!  Under Section 7 of GPL version 3, you are granted additional
+!  permissions described in the GCC Runtime Library Exception, version
+!  3.1, as published by the Free Software Foundation.
+
+!  You should have received a copy of the GNU General Public License and
+!  a copy of the GCC Runtime Library Exception along with this program;
+!  see the files COPYING3 and COPYING.RUNTIME respectively.  If not, see
+!  <http://www.gnu.org/licenses/>.
+
+! NOTE: Due to the use of dimension (..), the code only works when compiled
+! with -std=f2008ts/gnu/legacy but not with other standard settings.
+! Alternatively, the user can use the module version, which permits
+! compilation with -std=f95.
+
+      integer, parameter :: acc_device_kind = 4
+
+!     Keep in sync with include/gomp-constants.h.
+      integer (acc_device_kind), parameter :: acc_device_none = 0
+      integer (acc_device_kind), parameter :: acc_device_default = 1
+      integer (acc_device_kind), parameter :: acc_device_host = 2
+      integer (acc_device_kind), parameter :: acc_device_host_nonshm = 3
+      integer (acc_device_kind), parameter :: acc_device_not_host = 4
+      integer (acc_device_kind), parameter :: acc_device_nvidia = 5
+
+      integer, parameter :: acc_handle_kind = 4
+
+!     Keep in sync with include/gomp-constants.h.
+      integer (acc_handle_kind), parameter :: acc_async_noval = -1
+      integer (acc_handle_kind), parameter :: acc_async_sync = -2
+
+      integer, parameter :: openacc_version = 201306
+
+      interface acc_get_num_devices
+        function acc_get_num_devices_h (d)
+          import acc_device_kind
+          integer acc_get_num_devices_h
+          integer (acc_device_kind) d
+        end function
+      end interface
+
+      interface acc_set_device_type
+        subroutine acc_set_device_type_h (d)
+          import acc_device_kind
+          integer (acc_device_kind) d
+        end subroutine
+      end interface
+
+      interface acc_get_device_type
+        function acc_get_device_type_h ()
+          import acc_device_kind
+          integer (acc_device_kind) acc_get_device_type_h
+        end function
+      end interface
+
+      interface acc_set_device_num
+        subroutine acc_set_device_num_h (n, d)
+          import acc_device_kind
+          integer n
+          integer (acc_device_kind) d
+        end subroutine
+      end interface
+
+      interface acc_get_device_num
+        function acc_get_device_num_h (d)
+          import acc_device_kind
+          integer acc_get_device_num_h
+          integer (acc_device_kind) d
+        end function
+      end interface
+
+      interface acc_async_test
+        function acc_async_test_h (a)
+          logical acc_async_test_h
+          integer a
+        end function
+      end interface
+
+      interface acc_async_test_all
+        function acc_async_test_all_h ()
+          logical acc_async_test_all_h
+        end function
+      end interface
+
+      interface acc_wait
+        subroutine acc_wait_h (a)
+          integer a
+        end subroutine
+      end interface
+
+      interface acc_wait_async
+        subroutine acc_wait_async_h (a1, a2)
+          integer a1, a2
+        end subroutine
+      end interface
+
+      interface acc_wait_all
+        subroutine acc_wait_all_h ()
+        end subroutine
+      end interface
+
+      interface acc_wait_all_async
+        subroutine acc_wait_all_async_h (a)
+          integer a
+        end subroutine
+      end interface
+
+      interface acc_init
+        subroutine acc_init_h (devicetype)
+          import acc_device_kind
+          integer (acc_device_kind) devicetype
+        end subroutine
+      end interface
+
+      interface acc_shutdown
+        subroutine acc_shutdown_h (devicetype)
+          import acc_device_kind
+          integer (acc_device_kind) devicetype
+        end subroutine
+      end interface
+
+      interface acc_on_device
+        function acc_on_device_h (devicetype)
+          import acc_device_kind
+          logical acc_on_device_h
+          integer (acc_device_kind) devicetype
+        end function
+      end interface
+
+      ! acc_malloc: Only available in C/C++
+      ! acc_free: Only available in C/C++
+
+      interface acc_copyin
+        subroutine acc_copyin_32_h (a, len)
+          use iso_c_binding, only: c_int32_t
+          !GCC$ ATTRIBUTES NO_ARG_CHECK :: a
+          type (*), dimension (*) :: a
+          integer (c_int32_t) len
+        end subroutine
+
+        subroutine acc_copyin_64_h (a, len)
+          use iso_c_binding, only: c_int64_t
+          !GCC$ ATTRIBUTES NO_ARG_CHECK :: a
+          type (*), dimension (*) :: a
+          integer (c_int64_t) len
+        end subroutine
+
+        subroutine acc_copyin_array_h (a)
+          type (*), dimension (..), contiguous :: a
+          end subroutine
+      end interface
+
+      interface acc_present_or_copyin
+        subroutine acc_present_or_copyin_32_h (a, len)
+          use iso_c_binding, only: c_int32_t
+          !GCC$ ATTRIBUTES NO_ARG_CHECK :: a
+          type (*), dimension (*) :: a
+          integer (c_int32_t) len
+        end subroutine
+
+        subroutine acc_present_or_copyin_64_h (a, len)
+          use iso_c_binding, only: c_int64_t
+          !GCC$ ATTRIBUTES NO_ARG_CHECK :: a
+          type (*), dimension (*) :: a
+          integer (c_int64_t) len
+        end subroutine
+
+        subroutine acc_present_or_copyin_array_h (a)
+          type (*), dimension (..), contiguous :: a
+          end subroutine
+      end interface
+
+      interface acc_pcopyin
+        subroutine acc_pcopyin_32_h (a, len)
+          use iso_c_binding, only: c_int32_t
+          !GCC$ ATTRIBUTES NO_ARG_CHECK :: a
+          type (*), dimension (*) :: a
+          integer (c_int32_t) len
+        end subroutine
+
+        subroutine acc_pcopyin_64_h (a, len)
+          use iso_c_binding, only: c_int64_t
+          !GCC$ ATTRIBUTES NO_ARG_CHECK :: a
+          type (*), dimension (*) :: a
+          integer (c_int64_t) len
+        end subroutine
+
+        subroutine acc_pcopyin_array_h (a)
+          type (*), dimension (..), contiguous :: a
+          end subroutine
+      end interface
+
+      interface acc_create
+        subroutine acc_create_32_h (a, len)
+          use iso_c_binding, only: c_int32_t
+          !GCC$ ATTRIBUTES NO_ARG_CHECK :: a
+          type (*), dimension (*) :: a
+          integer (c_int32_t) len
+        end subroutine
+
+        subroutine acc_create_64_h (a, len)
+          use iso_c_binding, only: c_int64_t
+          !GCC$ ATTRIBUTES NO_ARG_CHECK :: a
+          type (*), dimension (*) :: a
+          integer (c_int64_t) len
+        end subroutine
+
+        subroutine acc_create_array_h (a)
+          type (*), dimension (..), contiguous :: a
+          end subroutine
+      end interface
+
+      interface acc_present_or_create
+        subroutine acc_present_or_create_32_h (a, len)
+          use iso_c_binding, only: c_int32_t
+          !GCC$ ATTRIBUTES NO_ARG_CHECK :: a
+          type (*), dimension (*) :: a
+          integer (c_int32_t) len
+        end subroutine
+
+        subroutine acc_present_or_create_64_h (a, len)
+          use iso_c_binding, only: c_int64_t
+          !GCC$ ATTRIBUTES NO_ARG_CHECK :: a
+          type (*), dimension (*) :: a
+          integer (c_int64_t) len
+        end subroutine
+
+        subroutine acc_present_or_create_array_h (a)
+          type (*), dimension (..), contiguous :: a
+          end subroutine
+      end interface
+
+      interface acc_pcreate
+        subroutine acc_pcreate_32_h (a, len)
+          use iso_c_binding, only: c_int32_t
+          !GCC$ ATTRIBUTES NO_ARG_CHECK :: a
+          type (*), dimension (*) :: a
+          integer (c_int32_t) len
+        end subroutine
+
+        subroutine acc_pcreate_64_h (a, len)
+          use iso_c_binding, only: c_int64_t
+          !GCC$ ATTRIBUTES NO_ARG_CHECK :: a
+          type (*), dimension (*) :: a
+          integer (c_int64_t) len
+        end subroutine
+
+        subroutine acc_pcreate_array_h (a)
+          type (*), dimension (..), contiguous :: a
+          end subroutine
+      end interface
+
+      interface acc_copyout
+        subroutine acc_copyout_32_h (a, len)
+          use iso_c_binding, only: c_int32_t
+          !GCC$ ATTRIBUTES NO_ARG_CHECK :: a
+          type (*), dimension (*) :: a
+          integer (c_int32_t) len
+        end subroutine
+
+        subroutine acc_copyout_64_h (a, len)
+          use iso_c_binding, only: c_int64_t
+          !GCC$ ATTRIBUTES NO_ARG_CHECK :: a
+          type (*), dimension (*) :: a
+          integer (c_int64_t) len
+        end subroutine
+
+        subroutine acc_copyout_array_h (a)
+          type (*), dimension (..), contiguous :: a
+        end subroutine
+      end interface
+
+      interface acc_delete
+        subroutine acc_delete_32_h (a, len)
+          use iso_c_binding, only: c_int32_t
+          !GCC$ ATTRIBUTES NO_ARG_CHECK :: a
+          type (*), dimension (*) :: a
+          integer (c_int32_t) len
+        end subroutine
+
+        subroutine acc_delete_64_h (a, len)
+          use iso_c_binding, only: c_int64_t
+          !GCC$ ATTRIBUTES NO_ARG_CHECK :: a
+          type (*), dimension (*) :: a
+          integer (c_int64_t) len
+        end subroutine
+
+        subroutine acc_delete_array_h (a)
+          type (*), dimension (..), contiguous :: a
+        end subroutine
+      end interface
+
+      interface acc_update_device
+        subroutine acc_update_device_32_h (a, len)
+          use iso_c_binding, only: c_int32_t
+          !GCC$ ATTRIBUTES NO_ARG_CHECK :: a
+          type (*), dimension (*) :: a
+          integer (c_int32_t) len
+        end subroutine
+
+        subroutine acc_update_device_64_h (a, len)
+          use iso_c_binding, only: c_int64_t
+          !GCC$ ATTRIBUTES NO_ARG_CHECK :: a
+          type (*), dimension (*) :: a
+          integer (c_int64_t) len
+        end subroutine
+
+        subroutine acc_update_device_array_h (a)
+          type (*), dimension (..), contiguous :: a
+        end subroutine
+      end interface
+
+      interface acc_update_self
+        subroutine acc_update_self_32_h (a, len)
+          use iso_c_binding, only: c_int32_t
+          !GCC$ ATTRIBUTES NO_ARG_CHECK :: a
+          type (*), dimension (*) :: a
+          integer (c_int32_t) len
+        end subroutine
+
+        subroutine acc_update_self_64_h (a, len)
+          use iso_c_binding, only: c_int64_t
+          !GCC$ ATTRIBUTES NO_ARG_CHECK :: a
+          type (*), dimension (*) :: a
+          integer (c_int64_t) len
+        end subroutine
+
+        subroutine acc_update_self_array_h (a)
+          type (*), dimension (..), contiguous :: a
+        end subroutine
+      end interface
+
+      ! acc_map_data: Only available in C/C++
+      ! acc_unmap_data: Only available in C/C++
+      ! acc_deviceptr: Only available in C/C++
+      ! acc_ostptr: Only available in C/C++
+
+      interface acc_is_present
+        function acc_is_present_32_h (a, len)
+          use iso_c_binding, only: c_int32_t
+          logical acc_is_present_32_h
+          !GCC$ ATTRIBUTES NO_ARG_CHECK :: a
+          type (*), dimension (*) :: a
+          integer (c_int32_t) len
+        end function
+
+        function acc_is_present_64_h (a, len)
+          use iso_c_binding, only: c_int64_t
+          logical acc_is_present_64_h
+          !GCC$ ATTRIBUTES NO_ARG_CHECK :: a
+          type (*), dimension (*) :: a
+          integer (c_int64_t) len
+        end function
+
+        function acc_is_present_array_h (a)
+          logical acc_is_present_array_h
+          type (*), dimension (..), contiguous :: a
+        end function
+      end interface
+
+      ! acc_memcpy_to_device: Only available in C/C++
+      ! acc_memcpy_from_device: Only available in C/C++
diff --git a/libgomp/plugin/Makefrag.am b/libgomp/plugin/Makefrag.am
new file mode 100644
index 00000000000..167485f52c1
--- /dev/null
+++ b/libgomp/plugin/Makefrag.am
@@ -0,0 +1,49 @@
+# Plugins for offload execution, Makefile.am fragment.
+#
+# Copyright (C) 2014-2015 Free Software Foundation, Inc.
+#
+# Contributed by Mentor Embedded.
+#
+# This file is part of the GNU Offloading and Multi Processing Library
+# (libgomp).
+#
+# Libgomp is free software; you can redistribute it and/or modify it
+# under the terms of the GNU General Public License as published by
+# the Free Software Foundation; either version 3, or (at your option)
+# any later version.
+#
+# Libgomp is distributed in the hope that it will be useful, but WITHOUT ANY
+# WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
+# FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+# more details.
+#
+# Under Section 7 of GPL version 3, you are granted additional
+# permissions described in the GCC Runtime Library Exception, version
+# 3.1, as published by the Free Software Foundation.
+#
+# You should have received a copy of the GNU General Public License and
+# a copy of the GCC Runtime Library Exception along with this program;
+# see the files COPYING3 and COPYING.RUNTIME respectively.  If not, see
+# <http://www.gnu.org/licenses/>.
+
+if PLUGIN_NVPTX
+# Nvidia PTX OpenACC plugin.
+libgomp_plugin_nvptx_version_info = -version-info $(libtool_VERSION)
+toolexeclib_LTLIBRARIES += libgomp-plugin-nvptx.la
+libgomp_plugin_nvptx_la_SOURCES = plugin/plugin-nvptx.c
+libgomp_plugin_nvptx_la_CPPFLAGS = $(AM_CPPFLAGS) $(PLUGIN_NVPTX_CPPFLAGS)
+libgomp_plugin_nvptx_la_LDFLAGS = $(libgomp_plugin_nvptx_version_info) \
+	$(lt_host_flags)
+libgomp_plugin_nvptx_la_LDFLAGS += $(PLUGIN_NVPTX_LDFLAGS)
+libgomp_plugin_nvptx_la_LIBADD = libgomp.la $(PLUGIN_NVPTX_LIBS)
+libgomp_plugin_nvptx_la_LIBTOOLFLAGS = --tag=disable-static
+endif
+
+libgomp_plugin_host_nonshm_version_info = -version-info $(libtool_VERSION)
+toolexeclib_LTLIBRARIES += libgomp-plugin-host_nonshm.la
+libgomp_plugin_host_nonshm_la_SOURCES = plugin/plugin-host.c
+libgomp_plugin_host_nonshm_la_CPPFLAGS = $(AM_CPPFLAGS) -DHOST_NONSHM_PLUGIN
+libgomp_plugin_host_nonshm_la_LDFLAGS = \
+	$(libgomp_plugin_host_nonshm_version_info) $(lt_host_flags)
+libgomp_plugin_host_nonshm_la_LIBADD = libgomp.la
+libgomp_plugin_host_nonshm_la_LIBTOOLFLAGS = --tag=disable-static
diff --git a/libgomp/plugin/configfrag.ac b/libgomp/plugin/configfrag.ac
new file mode 100644
index 00000000000..254c68853cb
--- /dev/null
+++ b/libgomp/plugin/configfrag.ac
@@ -0,0 +1,148 @@
+# Plugins for offload execution, configure.ac fragment.  -*- mode: autoconf -*-
+#
+# Copyright (C) 2014-2015 Free Software Foundation, Inc.
+#
+# Contributed by Mentor Embedded.
+#
+# This file is part of the GNU Offloading and Multi Processing Library
+# (libgomp).
+#
+# Libgomp is free software; you can redistribute it and/or modify it
+# under the terms of the GNU General Public License as published by
+# the Free Software Foundation; either version 3, or (at your option)
+# any later version.
+#
+# Libgomp is distributed in the hope that it will be useful, but WITHOUT ANY
+# WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
+# FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+# more details.
+#
+# Under Section 7 of GPL version 3, you are granted additional
+# permissions described in the GCC Runtime Library Exception, version
+# 3.1, as published by the Free Software Foundation.
+#
+# You should have received a copy of the GNU General Public License and
+# a copy of the GCC Runtime Library Exception along with this program;
+# see the files COPYING3 and COPYING.RUNTIME respectively.  If not, see
+# <http://www.gnu.org/licenses/>.
+
+offload_targets=
+AC_SUBST(offload_targets)
+plugin_support=yes
+AC_CHECK_LIB(dl, dlsym, , [plugin_support=no])
+if test x"$plugin_support" = xyes; then
+  AC_DEFINE(PLUGIN_SUPPORT, 1,
+    [Define if all infrastructure, needed for plugins, is supported.])
+  offload_targets=host_nonshm
+elif test "x${enable_offload_targets-no}" != xno; then
+  AC_MSG_ERROR([Can't support offloading without support for plugins])
+fi
+
+# Look for the CUDA driver package.
+CUDA_DRIVER_INCLUDE=
+CUDA_DRIVER_LIB=
+AC_SUBST(CUDA_DRIVER_INCLUDE)
+AC_SUBST(CUDA_DRIVER_LIB)
+CUDA_DRIVER_CPPFLAGS=
+CUDA_DRIVER_LDFLAGS=
+AC_ARG_WITH(cuda-driver,
+	[AS_HELP_STRING([--with-cuda-driver=PATH],
+		[specify prefix directory for installed CUDA driver package.
+		 Equivalent to --with-cuda-driver-include=PATH/include
+		 plus --with-cuda-driver-lib=PATH/lib])])
+AC_ARG_WITH(cuda-driver-include,
+	[AS_HELP_STRING([--with-cuda-driver-include=PATH],
+		[specify directory for installed CUDA driver include files])])
+AC_ARG_WITH(cuda-driver-lib,
+	[AS_HELP_STRING([--with-cuda-driver-lib=PATH],
+		[specify directory for the installed CUDA driver library])])
+if test "x$with_cuda_driver" != x; then
+  CUDA_DRIVER_INCLUDE=$with_cuda_driver/include
+  CUDA_DRIVER_LIB=$with_cuda_driver/lib
+fi
+if test "x$with_cuda_driver_include" != x; then
+  CUDA_DRIVER_INCLUDE=$with_cuda_driver_include
+fi
+if test "x$with_cuda_driver_lib" != x; then
+  CUDA_DRIVER_LIB=$with_cuda_driver_lib
+fi
+if test "x$CUDA_DRIVER_INCLUDE" != x; then
+  CUDA_DRIVER_CPPFLAGS=-I$CUDA_DRIVER_INCLUDE
+fi
+if test "x$CUDA_DRIVER_LIB" != x; then
+  CUDA_DRIVER_LDFLAGS=-L$CUDA_DRIVER_LIB
+fi
+
+PLUGIN_NVPTX=0
+PLUGIN_NVPTX_CPPFLAGS=
+PLUGIN_NVPTX_LDFLAGS=
+PLUGIN_NVPTX_LIBS=
+AC_SUBST(PLUGIN_NVPTX)
+AC_SUBST(PLUGIN_NVPTX_CPPFLAGS)
+AC_SUBST(PLUGIN_NVPTX_LDFLAGS)
+AC_SUBST(PLUGIN_NVPTX_LIBS)
+
+# Get offload targets and path to install tree of offloading compiler.
+offload_additional_options=
+offload_additional_lib_paths=
+AC_SUBST(offload_additional_options)
+AC_SUBST(offload_additional_lib_paths)
+if test x"$enable_offload_targets" != x; then
+  for tgt in `echo $enable_offload_targets | sed -e 's#,# #g'`; do
+    tgt_dir=`echo $tgt | grep '=' | sed 's/.*=//'`
+    tgt=`echo $tgt | sed 's/=.*//'`
+    case $tgt in
+      *-intelmic-* | *-intelmicemul-*)
+	tgt_name=intelmic
+	;;
+      nvptx*)
+        tgt_name=nvptx
+	PLUGIN_NVPTX=$tgt
+	PLUGIN_NVPTX_CPPFLAGS=$CUDA_DRIVER_CPPFLAGS
+	PLUGIN_NVPTX_LDFLAGS=$CUDA_DRIVER_LDFLAGS
+	PLUGIN_NVPTX_LIBS='-lcuda'
+
+	PLUGIN_NVPTX_save_CPPFLAGS=$CPPFLAGS
+	CPPFLAGS="$PLUGIN_NVPTX_CPPFLAGS $CPPFLAGS"
+	PLUGIN_NVPTX_save_LDFLAGS=$LDFLAGS
+	LDFLAGS="$PLUGIN_NVPTX_LDFLAGS $LDFLAGS"
+	PLUGIN_NVPTX_save_LIBS=$LIBS
+	LIBS="$PLUGIN_NVPTX_LIBS $LIBS"
+	AC_LINK_IFELSE(
+	  [AC_LANG_PROGRAM(
+	    [#include "cuda.h"],
+	      [CUresult r = cuCtxPushCurrent (NULL);])],
+	  [PLUGIN_NVPTX=1])
+	CPPFLAGS=$PLUGIN_NVPTX_save_CPPFLAGS
+	LDFLAGS=$PLUGIN_NVPTX_save_LDFLAGS
+	LIBS=$PLUGIN_NVPTX_save_LIBS
+	case $PLUGIN_NVPTX in
+	  nvptx*)
+	    PLUGIN_NVPTX=0
+	    AC_MSG_ERROR([CUDA driver package required for nvptx support])
+	    ;;
+	esac
+	;;
+      *)
+	AC_MSG_ERROR([unknown offload target specified])
+	;;
+    esac
+    if test x"$offload_targets" = x; then
+      offload_targets=$tgt_name
+    else
+      offload_targets=$offload_targets,$tgt_name
+    fi
+    if test x"$tgt_dir" != x; then
+      offload_additional_options="$offload_additional_options -B$tgt_dir/libexec/gcc/\$(target_alias)/\$(gcc_version) -B$tgt_dir/bin"
+      offload_additional_lib_paths="$offload_additional_lib_paths:$tgt_dir/lib64:$tgt_dir/lib:$tgt_dir/lib32"
+    else
+      offload_additional_options="$offload_additional_options -B\$(libexecdir)/gcc/\$(target_alias)/\$(gcc_version) -B\$(bindir)"
+      offload_additional_lib_paths="$offload_additional_lib_paths:$toolexeclibdir"
+    fi
+  done
+fi
+AC_DEFINE_UNQUOTED(OFFLOAD_TARGETS, "$offload_targets",
+  [Define to hold the list of target names suitable for offloading.])
+AM_CONDITIONAL([PLUGIN_NVPTX], [test $PLUGIN_NVPTX = 1])
+AC_DEFINE_UNQUOTED([PLUGIN_NVPTX], [$PLUGIN_NVPTX],
+  [Define to 1 if the NVIDIA plugin is built, 0 if not.])
diff --git a/libgomp/plugin/plugin-host.c b/libgomp/plugin/plugin-host.c
new file mode 100644
index 00000000000..ebf7f11caf7
--- /dev/null
+++ b/libgomp/plugin/plugin-host.c
@@ -0,0 +1,266 @@
+/* OpenACC Runtime Library: acc_device_host, acc_device_host_nonshm.
+
+   Copyright (C) 2013-2015 Free Software Foundation, Inc.
+
+   Contributed by Mentor Embedded.
+
+   This file is part of the GNU Offloading and Multi Processing Library
+   (libgomp).
+
+   Libgomp is free software; you can redistribute it and/or modify it
+   under the terms of the GNU General Public License as published by
+   the Free Software Foundation; either version 3, or (at your option)
+   any later version.
+
+   Libgomp is distributed in the hope that it will be useful, but WITHOUT ANY
+   WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
+   FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+   more details.
+
+   Under Section 7 of GPL version 3, you are granted additional
+   permissions described in the GCC Runtime Library Exception, version
+   3.1, as published by the Free Software Foundation.
+
+   You should have received a copy of the GNU General Public License and
+   a copy of the GCC Runtime Library Exception along with this program;
+   see the files COPYING3 and COPYING.RUNTIME respectively.  If not, see
+   <http://www.gnu.org/licenses/>.  */
+
+/* Simple implementation of support routines for a shared-memory
+   acc_device_host, and a non-shared memory acc_device_host_nonshm, with the
+   latter built as a plugin.  */
+
+#include "openacc.h"
+#include "config.h"
+#ifdef HOST_NONSHM_PLUGIN
+#include "libgomp-plugin.h"
+#include "oacc-plugin.h"
+#else
+#include "libgomp.h"
+#include "oacc-int.h"
+#endif
+
+#include <stdint.h>
+#include <stdlib.h>
+#include <string.h>
+#include <stdio.h>
+
+#ifdef HOST_NONSHM_PLUGIN
+#define STATIC
+#define GOMP(X) GOMP_PLUGIN_##X
+#define SELF "host_nonshm plugin: "
+#else
+#define STATIC static
+#define GOMP(X) gomp_##X
+#define SELF "host: "
+#endif
+
+STATIC const char *
+GOMP_OFFLOAD_get_name (void)
+{
+#ifdef HOST_NONSHM_PLUGIN
+  return "host_nonshm";
+#else
+  return "host";
+#endif
+}
+
+STATIC unsigned int
+GOMP_OFFLOAD_get_caps (void)
+{
+  unsigned int caps = (GOMP_OFFLOAD_CAP_OPENACC_200
+		       | GOMP_OFFLOAD_CAP_NATIVE_EXEC);
+
+#ifndef HOST_NONSHM_PLUGIN
+  caps |= GOMP_OFFLOAD_CAP_SHARED_MEM;
+#endif
+
+  return caps;
+}
+
+STATIC int
+GOMP_OFFLOAD_get_type (void)
+{
+#ifdef HOST_NONSHM_PLUGIN
+  return OFFLOAD_TARGET_TYPE_HOST_NONSHM;
+#else
+  return OFFLOAD_TARGET_TYPE_HOST;
+#endif
+}
+
+STATIC int
+GOMP_OFFLOAD_get_num_devices (void)
+{
+  return 1;
+}
+
+STATIC void
+GOMP_OFFLOAD_register_image (void *host_table __attribute__ ((unused)),
+			     void *target_data __attribute__ ((unused)))
+{
+}
+
+STATIC void
+GOMP_OFFLOAD_init_device (int n __attribute__ ((unused)))
+{
+}
+
+STATIC void
+GOMP_OFFLOAD_fini_device (int n __attribute__ ((unused)))
+{
+}
+
+STATIC int
+GOMP_OFFLOAD_get_table (int n __attribute__ ((unused)),
+			struct mapping_table **table __attribute__ ((unused)))
+{
+  return 0;
+}
+
+STATIC void *
+GOMP_OFFLOAD_openacc_open_device (int n)
+{
+  return (void *) (intptr_t) n;
+}
+
+STATIC int
+GOMP_OFFLOAD_openacc_close_device (void *hnd)
+{
+  return 0;
+}
+
+STATIC int
+GOMP_OFFLOAD_openacc_get_device_num (void)
+{
+  return 0;
+}
+
+STATIC void
+GOMP_OFFLOAD_openacc_set_device_num (int n)
+{
+  if (n > 0)
+    GOMP (fatal) ("device number %u out of range for host execution", n);
+}
+
+STATIC void *
+GOMP_OFFLOAD_alloc (int n __attribute__ ((unused)), size_t s)
+{
+  return GOMP (malloc) (s);
+}
+
+STATIC void
+GOMP_OFFLOAD_free (int n __attribute__ ((unused)), void *p)
+{
+  free (p);
+}
+
+STATIC void *
+GOMP_OFFLOAD_host2dev (int n __attribute__ ((unused)), void *d, const void *h,
+		       size_t s)
+{
+#ifdef HOST_NONSHM_PLUGIN
+  memcpy (d, h, s);
+#endif
+
+  return 0;
+}
+
+STATIC void *
+GOMP_OFFLOAD_dev2host (int n __attribute__ ((unused)), void *h, const void *d,
+		       size_t s)
+{
+#ifdef HOST_NONSHM_PLUGIN
+  memcpy (h, d, s);
+#endif
+
+  return 0;
+}
+
+STATIC void
+GOMP_OFFLOAD_run (int n __attribute__ ((unused)), void *fn_ptr, void *vars)
+{
+  void (*fn)(void *) = (void (*)(void *)) fn_ptr;
+
+  fn (vars);
+}
+
+STATIC void
+GOMP_OFFLOAD_openacc_parallel (void (*fn) (void *),
+			       size_t mapnum __attribute__ ((unused)),
+			       void **hostaddrs __attribute__ ((unused)),
+			       void **devaddrs __attribute__ ((unused)),
+			       size_t *sizes __attribute__ ((unused)),
+			       unsigned short *kinds __attribute__ ((unused)),
+			       int num_gangs __attribute__ ((unused)),
+			       int num_workers __attribute__ ((unused)),
+			       int vector_length __attribute__ ((unused)),
+			       int async __attribute__ ((unused)),
+			       void *targ_mem_desc __attribute__ ((unused)))
+{
+#ifdef HOST_NONSHM_PLUGIN
+  fn (devaddrs);
+#else
+  fn (hostaddrs);
+#endif
+}
+
+STATIC void
+GOMP_OFFLOAD_openacc_register_async_cleanup (void *targ_mem_desc)
+{
+#ifdef HOST_NONSHM_PLUGIN
+  /* "Asynchronous" launches are executed synchronously on the (non-SHM) host,
+     so there's no point in delaying host-side cleanup -- just do it now.  */
+  GOMP_PLUGIN_async_unmap_vars (targ_mem_desc);
+#endif
+}
+
+STATIC void
+GOMP_OFFLOAD_openacc_async_set_async (int async __attribute__ ((unused)))
+{
+}
+
+STATIC int
+GOMP_OFFLOAD_openacc_async_test (int async __attribute__ ((unused)))
+{
+  return 1;
+}
+
+STATIC int
+GOMP_OFFLOAD_openacc_async_test_all (void)
+{
+  return 1;
+}
+
+STATIC void
+GOMP_OFFLOAD_openacc_async_wait (int async __attribute__ ((unused)))
+{
+}
+
+STATIC void
+GOMP_OFFLOAD_openacc_async_wait_all (void)
+{
+}
+
+STATIC void
+GOMP_OFFLOAD_openacc_async_wait_async (int async1 __attribute__ ((unused)),
+				       int async2 __attribute__ ((unused)))
+{
+}
+
+STATIC void
+GOMP_OFFLOAD_openacc_async_wait_all_async (int async __attribute__ ((unused)))
+{
+}
+
+STATIC void *
+GOMP_OFFLOAD_openacc_create_thread_data (void *targ_data
+					 __attribute__ ((unused)))
+{
+  return NULL;
+}
+
+STATIC void
+GOMP_OFFLOAD_openacc_destroy_thread_data (void *tls_data
+					  __attribute__ ((unused)))
+{
+}
diff --git a/libgomp/plugin/plugin-nvptx.c b/libgomp/plugin/plugin-nvptx.c
new file mode 100644
index 00000000000..483cb7559e8
--- /dev/null
+++ b/libgomp/plugin/plugin-nvptx.c
@@ -0,0 +1,1791 @@
+/* Plugin for NVPTX execution.
+
+   Copyright (C) 2013-2015 Free Software Foundation, Inc.
+
+   Contributed by Mentor Embedded.
+
+   This file is part of the GNU Offloading and Multi Processing Library
+   (libgomp).
+
+   Libgomp is free software; you can redistribute it and/or modify it
+   under the terms of the GNU General Public License as published by
+   the Free Software Foundation; either version 3, or (at your option)
+   any later version.
+
+   Libgomp is distributed in the hope that it will be useful, but WITHOUT ANY
+   WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
+   FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+   more details.
+
+   Under Section 7 of GPL version 3, you are granted additional
+   permissions described in the GCC Runtime Library Exception, version
+   3.1, as published by the Free Software Foundation.
+
+   You should have received a copy of the GNU General Public License and
+   a copy of the GCC Runtime Library Exception along with this program;
+   see the files COPYING3 and COPYING.RUNTIME respectively.  If not, see
+   <http://www.gnu.org/licenses/>.  */
+
+/* Nvidia PTX-specific parts of OpenACC support.  The cuda driver
+   library appears to hold some implicit state, but the documentation
+   is not clear as to what that state might be.  Or how one might
+   propagate it from one thread to another.  */
+
+#include "openacc.h"
+#include "config.h"
+#include "libgomp-plugin.h"
+#include "oacc-ptx.h"
+#include "oacc-plugin.h"
+
+#include <pthread.h>
+#include <cuda.h>
+#include <stdbool.h>
+#include <stdint.h>
+#include <string.h>
+#include <stdio.h>
+#include <dlfcn.h>
+#include <unistd.h>
+#include <assert.h>
+
+#define	ARRAYSIZE(X) (sizeof (X) / sizeof ((X)[0]))
+
+static struct
+{
+  CUresult r;
+  char *m;
+} cuda_errlist[]=
+{
+  { CUDA_ERROR_INVALID_VALUE, "invalid value" },
+  { CUDA_ERROR_OUT_OF_MEMORY, "out of memory" },
+  { CUDA_ERROR_NOT_INITIALIZED, "not initialized" },
+  { CUDA_ERROR_DEINITIALIZED, "deinitialized" },
+  { CUDA_ERROR_PROFILER_DISABLED, "profiler disabled" },
+  { CUDA_ERROR_PROFILER_NOT_INITIALIZED, "profiler not initialized" },
+  { CUDA_ERROR_PROFILER_ALREADY_STARTED, "already started" },
+  { CUDA_ERROR_PROFILER_ALREADY_STOPPED, "already stopped" },
+  { CUDA_ERROR_NO_DEVICE, "no device" },
+  { CUDA_ERROR_INVALID_DEVICE, "invalid device" },
+  { CUDA_ERROR_INVALID_IMAGE, "invalid image" },
+  { CUDA_ERROR_INVALID_CONTEXT, "invalid context" },
+  { CUDA_ERROR_CONTEXT_ALREADY_CURRENT, "context already current" },
+  { CUDA_ERROR_MAP_FAILED, "map error" },
+  { CUDA_ERROR_UNMAP_FAILED, "unmap error" },
+  { CUDA_ERROR_ARRAY_IS_MAPPED, "array is mapped" },
+  { CUDA_ERROR_ALREADY_MAPPED, "already mapped" },
+  { CUDA_ERROR_NO_BINARY_FOR_GPU, "no binary for gpu" },
+  { CUDA_ERROR_ALREADY_ACQUIRED, "already acquired" },
+  { CUDA_ERROR_NOT_MAPPED, "not mapped" },
+  { CUDA_ERROR_NOT_MAPPED_AS_ARRAY, "not mapped as array" },
+  { CUDA_ERROR_NOT_MAPPED_AS_POINTER, "not mapped as pointer" },
+  { CUDA_ERROR_ECC_UNCORRECTABLE, "ecc uncorrectable" },
+  { CUDA_ERROR_UNSUPPORTED_LIMIT, "unsupported limit" },
+  { CUDA_ERROR_CONTEXT_ALREADY_IN_USE, "context already in use" },
+  { CUDA_ERROR_PEER_ACCESS_UNSUPPORTED, "peer access unsupported" },
+  { CUDA_ERROR_INVALID_SOURCE, "invalid source" },
+  { CUDA_ERROR_FILE_NOT_FOUND, "file not found" },
+  { CUDA_ERROR_SHARED_OBJECT_SYMBOL_NOT_FOUND,
+                                           "shared object symbol not found" },
+  { CUDA_ERROR_SHARED_OBJECT_INIT_FAILED, "shared object init error" },
+  { CUDA_ERROR_OPERATING_SYSTEM, "operating system" },
+  { CUDA_ERROR_INVALID_HANDLE, "invalid handle" },
+  { CUDA_ERROR_NOT_FOUND, "not found" },
+  { CUDA_ERROR_NOT_READY, "not ready" },
+  { CUDA_ERROR_LAUNCH_FAILED, "launch error" },
+  { CUDA_ERROR_LAUNCH_OUT_OF_RESOURCES, "launch out of resources" },
+  { CUDA_ERROR_LAUNCH_TIMEOUT, "launch timeout" },
+  { CUDA_ERROR_LAUNCH_INCOMPATIBLE_TEXTURING,
+                                             "launch incompatibe texturing" },
+  { CUDA_ERROR_PEER_ACCESS_ALREADY_ENABLED, "peer access already enabled" },
+  { CUDA_ERROR_PEER_ACCESS_NOT_ENABLED, "peer access not enabled " },
+  { CUDA_ERROR_PRIMARY_CONTEXT_ACTIVE, "primary cotext active" },
+  { CUDA_ERROR_CONTEXT_IS_DESTROYED, "context is destroyed" },
+  { CUDA_ERROR_ASSERT, "assert" },
+  { CUDA_ERROR_TOO_MANY_PEERS, "too many peers" },
+  { CUDA_ERROR_HOST_MEMORY_ALREADY_REGISTERED,
+                                           "host memory already registered" },
+  { CUDA_ERROR_HOST_MEMORY_NOT_REGISTERED, "host memory not registered" },
+  { CUDA_ERROR_NOT_PERMITTED, "not permitted" },
+  { CUDA_ERROR_NOT_SUPPORTED, "not supported" },
+  { CUDA_ERROR_UNKNOWN, "unknown" }
+};
+
+static char errmsg[128];
+
+static char *
+cuda_error (CUresult r)
+{
+  int i;
+
+  for (i = 0; i < ARRAYSIZE (cuda_errlist); i++)
+    {
+      if (cuda_errlist[i].r == r)
+	return &cuda_errlist[i].m[0];
+    }
+
+  sprintf (&errmsg[0], "unknown result code: %5d", r);
+
+  return &errmsg[0];
+}
+
+struct targ_fn_descriptor
+{
+  CUfunction fn;
+  const char *name;
+};
+
+static bool ptx_inited = false;
+
+struct ptx_stream
+{
+  CUstream stream;
+  pthread_t host_thread;
+  bool multithreaded;
+
+  CUdeviceptr d;
+  void *h;
+  void *h_begin;
+  void *h_end;
+  void *h_next;
+  void *h_prev;
+  void *h_tail;
+
+  struct ptx_stream *next;
+};
+
+/* Thread-specific data for PTX.  */
+
+struct nvptx_thread
+{
+  struct ptx_stream *current_stream;
+  struct ptx_device *ptx_dev;
+};
+
+struct map
+{
+  int     async;
+  size_t  size;
+  char    mappings[0];
+};
+
+static void
+map_init (struct ptx_stream *s)
+{
+  CUresult r;
+
+  int size = getpagesize ();
+
+  assert (s);
+  assert (!s->d);
+  assert (!s->h);
+
+  r = cuMemAllocHost (&s->h, size);
+  if (r != CUDA_SUCCESS)
+    GOMP_PLUGIN_fatal ("cuMemAllocHost error: %s", cuda_error (r));
+
+  r = cuMemHostGetDevicePointer (&s->d, s->h, 0);
+  if (r != CUDA_SUCCESS)
+    GOMP_PLUGIN_fatal ("cuMemHostGetDevicePointer error: %s", cuda_error (r));
+
+  assert (s->h);
+
+  s->h_begin = s->h;
+  s->h_end = s->h_begin + size;
+  s->h_next = s->h_prev = s->h_tail = s->h_begin;
+
+  assert (s->h_next);
+  assert (s->h_end);
+}
+
+static void
+map_fini (struct ptx_stream *s)
+{
+  CUresult r;
+
+  r = cuMemFreeHost (s->h);
+  if (r != CUDA_SUCCESS)
+    GOMP_PLUGIN_fatal ("cuMemFreeHost error: %s", cuda_error (r));
+}
+
+static void
+map_pop (struct ptx_stream *s)
+{
+  struct map *m;
+
+  assert (s != NULL);
+  assert (s->h_next);
+  assert (s->h_prev);
+  assert (s->h_tail);
+
+  m = s->h_tail;
+
+  s->h_tail += m->size;
+
+  if (s->h_tail >= s->h_end)
+    s->h_tail = s->h_begin + (int) (s->h_tail - s->h_end);
+
+  if (s->h_next == s->h_tail)
+    s->h_prev = s->h_next;
+
+  assert (s->h_next >= s->h_begin);
+  assert (s->h_tail >= s->h_begin);
+  assert (s->h_prev >= s->h_begin);
+
+  assert (s->h_next <= s->h_end);
+  assert (s->h_tail <= s->h_end);
+  assert (s->h_prev <= s->h_end);
+}
+
+static void
+map_push (struct ptx_stream *s, int async, size_t size, void **h, void **d)
+{
+  int left;
+  int offset;
+  struct map *m;
+
+  assert (s != NULL);
+
+  left = s->h_end - s->h_next;
+  size += sizeof (struct map);
+
+  assert (s->h_prev);
+  assert (s->h_next);
+
+  if (size >= left)
+    {
+      m = s->h_prev;
+      m->size += left;
+      s->h_next = s->h_begin;
+
+      if (s->h_next + size > s->h_end)
+	GOMP_PLUGIN_fatal ("unable to push map");
+    }
+
+  assert (s->h_next);
+
+  m = s->h_next;
+  m->async = async;
+  m->size = size;
+
+  offset = (void *)&m->mappings[0] - s->h;
+
+  *d = (void *)(s->d + offset);
+  *h = (void *)(s->h + offset);
+
+  s->h_prev = s->h_next;
+  s->h_next += size;
+
+  assert (s->h_prev);
+  assert (s->h_next);
+
+  assert (s->h_next >= s->h_begin);
+  assert (s->h_tail >= s->h_begin);
+  assert (s->h_prev >= s->h_begin);
+  assert (s->h_next <= s->h_end);
+  assert (s->h_tail <= s->h_end);
+  assert (s->h_prev <= s->h_end);
+
+  return;
+}
+
+struct ptx_device
+{
+  CUcontext ctx;
+  bool ctx_shared;
+  CUdevice dev;
+  struct ptx_stream *null_stream;
+  /* All non-null streams associated with this device (actually context),
+     either created implicitly or passed in from the user (via
+     acc_set_cuda_stream).  */
+  struct ptx_stream *active_streams;
+  struct {
+    struct ptx_stream **arr;
+    int size;
+  } async_streams;
+  /* A lock for use when manipulating the above stream list and array.  */
+  pthread_mutex_t stream_lock;
+  int ord;
+  bool overlap;
+  bool map;
+  bool concur;
+  int  mode;
+  bool mkern;
+
+  struct ptx_device *next;
+};
+
+enum ptx_event_type
+{
+  PTX_EVT_MEM,
+  PTX_EVT_KNL,
+  PTX_EVT_SYNC,
+  PTX_EVT_ASYNC_CLEANUP
+};
+
+struct ptx_event
+{
+  CUevent *evt;
+  int type;
+  void *addr;
+  int ord;
+
+  struct ptx_event *next;
+};
+
+static pthread_mutex_t ptx_event_lock;
+static struct ptx_event *ptx_events;
+
+#define _XSTR(s) _STR(s)
+#define _STR(s) #s
+
+static struct _synames
+{
+  char *n;
+} cuda_symnames[] =
+{
+  { _XSTR (cuCtxCreate) },
+  { _XSTR (cuCtxDestroy) },
+  { _XSTR (cuCtxGetCurrent) },
+  { _XSTR (cuCtxPushCurrent) },
+  { _XSTR (cuCtxSynchronize) },
+  { _XSTR (cuDeviceGet) },
+  { _XSTR (cuDeviceGetAttribute) },
+  { _XSTR (cuDeviceGetCount) },
+  { _XSTR (cuEventCreate) },
+  { _XSTR (cuEventDestroy) },
+  { _XSTR (cuEventQuery) },
+  { _XSTR (cuEventRecord) },
+  { _XSTR (cuInit) },
+  { _XSTR (cuLaunchKernel) },
+  { _XSTR (cuLinkAddData) },
+  { _XSTR (cuLinkComplete) },
+  { _XSTR (cuLinkCreate) },
+  { _XSTR (cuMemAlloc) },
+  { _XSTR (cuMemAllocHost) },
+  { _XSTR (cuMemcpy) },
+  { _XSTR (cuMemcpyDtoH) },
+  { _XSTR (cuMemcpyDtoHAsync) },
+  { _XSTR (cuMemcpyHtoD) },
+  { _XSTR (cuMemcpyHtoDAsync) },
+  { _XSTR (cuMemFree) },
+  { _XSTR (cuMemFreeHost) },
+  { _XSTR (cuMemGetAddressRange) },
+  { _XSTR (cuMemHostGetDevicePointer) },
+  { _XSTR (cuMemHostRegister) },
+  { _XSTR (cuMemHostUnregister) },
+  { _XSTR (cuModuleGetFunction) },
+  { _XSTR (cuModuleLoadData) },
+  { _XSTR (cuStreamDestroy) },
+  { _XSTR (cuStreamQuery) },
+  { _XSTR (cuStreamSynchronize) },
+  { _XSTR (cuStreamWaitEvent) }
+};
+
+static int
+verify_device_library (void)
+{
+  int i;
+  void *dh, *ds;
+
+  dh = dlopen ("libcuda.so", RTLD_LAZY);
+  if (!dh)
+    return -1;
+
+  for (i = 0; i < ARRAYSIZE (cuda_symnames); i++)
+    {
+      ds = dlsym (dh, cuda_symnames[i].n);
+      if (!ds)
+        return -1;
+    }
+
+  dlclose (dh);
+
+  return 0;
+}
+
+static inline struct nvptx_thread *
+nvptx_thread (void)
+{
+  return (struct nvptx_thread *) GOMP_PLUGIN_acc_thread ();
+}
+
+static void
+init_streams_for_device (struct ptx_device *ptx_dev, int concurrency)
+{
+  int i;
+  struct ptx_stream *null_stream
+    = GOMP_PLUGIN_malloc (sizeof (struct ptx_stream));
+
+  null_stream->stream = NULL;
+  null_stream->host_thread = pthread_self ();
+  null_stream->multithreaded = true;
+  null_stream->d = (CUdeviceptr) NULL;
+  null_stream->h = NULL;
+  map_init (null_stream);
+  ptx_dev->null_stream = null_stream;
+
+  ptx_dev->active_streams = NULL;
+  pthread_mutex_init (&ptx_dev->stream_lock, NULL);
+
+  if (concurrency < 1)
+    concurrency = 1;
+
+  /* This is just a guess -- make space for as many async streams as the
+     current device is capable of concurrently executing.  This can grow
+     later as necessary.  No streams are created yet.  */
+  ptx_dev->async_streams.arr
+    = GOMP_PLUGIN_malloc (concurrency * sizeof (struct ptx_stream *));
+  ptx_dev->async_streams.size = concurrency;
+
+  for (i = 0; i < concurrency; i++)
+    ptx_dev->async_streams.arr[i] = NULL;
+}
+
+static void
+fini_streams_for_device (struct ptx_device *ptx_dev)
+{
+  free (ptx_dev->async_streams.arr);
+
+  while (ptx_dev->active_streams != NULL)
+    {
+      struct ptx_stream *s = ptx_dev->active_streams;
+      ptx_dev->active_streams = ptx_dev->active_streams->next;
+
+      cuStreamDestroy (s->stream);
+      map_fini (s);
+      free (s);
+    }
+
+  map_fini (ptx_dev->null_stream);
+  free (ptx_dev->null_stream);
+}
+
+/* Select a stream for (OpenACC-semantics) ASYNC argument for the current
+   thread THREAD (and also current device/context).  If CREATE is true, create
+   the stream if it does not exist (or use EXISTING if it is non-NULL), and
+   associate the stream with the same thread argument.  Returns stream to use
+   as result.  */
+
+static struct ptx_stream *
+select_stream_for_async (int async, pthread_t thread, bool create,
+			 CUstream existing)
+{
+  struct nvptx_thread *nvthd = nvptx_thread ();
+  /* Local copy of TLS variable.  */
+  struct ptx_device *ptx_dev = nvthd->ptx_dev;
+  struct ptx_stream *stream = NULL;
+  int orig_async = async;
+
+  /* The special value acc_async_noval (-1) maps (for now) to an
+     implicitly-created stream, which is then handled the same as any other
+     numbered async stream.  Other options are available, e.g. using the null
+     stream for anonymous async operations, or choosing an idle stream from an
+     active set.  But, stick with this for now.  */
+  if (async > acc_async_sync)
+    async++;
+
+  if (create)
+    pthread_mutex_lock (&ptx_dev->stream_lock);
+
+  /* NOTE: AFAICT there's no particular need for acc_async_sync to map to the
+     null stream, and in fact better performance may be obtainable if it doesn't
+     (because the null stream enforces overly-strict synchronisation with
+     respect to other streams for legacy reasons, and that's probably not
+     needed with OpenACC).  Maybe investigate later.  */
+  if (async == acc_async_sync)
+    stream = ptx_dev->null_stream;
+  else if (async >= 0 && async < ptx_dev->async_streams.size
+	   && ptx_dev->async_streams.arr[async] && !(create && existing))
+    stream = ptx_dev->async_streams.arr[async];
+  else if (async >= 0 && create)
+    {
+      if (async >= ptx_dev->async_streams.size)
+	{
+	  int i, newsize = ptx_dev->async_streams.size * 2;
+
+	  if (async >= newsize)
+	    newsize = async + 1;
+
+	  ptx_dev->async_streams.arr
+	    = GOMP_PLUGIN_realloc (ptx_dev->async_streams.arr,
+				   newsize * sizeof (struct ptx_stream *));
+
+	  for (i = ptx_dev->async_streams.size; i < newsize; i++)
+	    ptx_dev->async_streams.arr[i] = NULL;
+
+	  ptx_dev->async_streams.size = newsize;
+	}
+
+      /* Create a new stream on-demand if there isn't one already, or if we're
+	 setting a particular async value to an existing (externally-provided)
+	 stream.  */
+      if (!ptx_dev->async_streams.arr[async] || existing)
+        {
+	  CUresult r;
+	  struct ptx_stream *s
+	    = GOMP_PLUGIN_malloc (sizeof (struct ptx_stream));
+
+	  if (existing)
+	    s->stream = existing;
+	  else
+	    {
+	      r = cuStreamCreate (&s->stream, CU_STREAM_DEFAULT);
+	      if (r != CUDA_SUCCESS)
+		GOMP_PLUGIN_fatal ("cuStreamCreate error: %s", cuda_error (r));
+	    }
+
+	  /* If CREATE is true, we're going to be queueing some work on this
+	     stream.  Associate it with the current host thread.  */
+	  s->host_thread = thread;
+	  s->multithreaded = false;
+
+	  s->d = (CUdeviceptr) NULL;
+	  s->h = NULL;
+	  map_init (s);
+
+	  s->next = ptx_dev->active_streams;
+	  ptx_dev->active_streams = s;
+	  ptx_dev->async_streams.arr[async] = s;
+	}
+
+      stream = ptx_dev->async_streams.arr[async];
+    }
+  else if (async < 0)
+    GOMP_PLUGIN_fatal ("bad async %d", async);
+
+  if (create)
+    {
+      assert (stream != NULL);
+
+      /* If we're trying to use the same stream from different threads
+	 simultaneously, set stream->multithreaded to true.  This affects the
+	 behaviour of acc_async_test_all and acc_wait_all, which are supposed to
+	 only wait for asynchronous launches from the same host thread they are
+	 invoked on.  If multiple threads use the same async value, we make note
+	 of that here and fall back to testing/waiting for all threads in those
+	 functions.  */
+      if (thread != stream->host_thread)
+        stream->multithreaded = true;
+
+      pthread_mutex_unlock (&ptx_dev->stream_lock);
+    }
+  else if (stream && !stream->multithreaded
+	   && !pthread_equal (stream->host_thread, thread))
+    GOMP_PLUGIN_fatal ("async %d used on wrong thread", orig_async);
+
+  return stream;
+}
+
+static int nvptx_get_num_devices (void);
+
+/* Initialize the device.  */
+static int
+nvptx_init (void)
+{
+  CUresult r;
+  int rc;
+
+  if (ptx_inited)
+    return nvptx_get_num_devices ();
+
+  rc = verify_device_library ();
+  if (rc < 0)
+    return -1;
+
+  r = cuInit (0);
+  if (r != CUDA_SUCCESS)
+    GOMP_PLUGIN_fatal ("cuInit error: %s", cuda_error (r));
+
+  ptx_events = NULL;
+
+  pthread_mutex_init (&ptx_event_lock, NULL);
+
+  ptx_inited = true;
+
+  return nvptx_get_num_devices ();
+}
+
+static void
+nvptx_fini (void)
+{
+  ptx_inited = false;
+}
+
+static void *
+nvptx_open_device (int n)
+{
+  struct ptx_device *ptx_dev;
+  CUdevice dev;
+  CUresult r;
+  int async_engines, pi;
+
+  r = cuDeviceGet (&dev, n);
+  if (r != CUDA_SUCCESS)
+    GOMP_PLUGIN_fatal ("cuDeviceGet error: %s", cuda_error (r));
+
+  ptx_dev = GOMP_PLUGIN_malloc (sizeof (struct ptx_device));
+
+  ptx_dev->ord = n;
+  ptx_dev->dev = dev;
+  ptx_dev->ctx_shared = false;
+
+  r = cuCtxGetCurrent (&ptx_dev->ctx);
+  if (r != CUDA_SUCCESS)
+    GOMP_PLUGIN_fatal ("cuCtxGetCurrent error: %s", cuda_error (r));
+
+  if (!ptx_dev->ctx)
+    {
+      r = cuCtxCreate (&ptx_dev->ctx, CU_CTX_SCHED_AUTO, dev);
+      if (r != CUDA_SUCCESS)
+	GOMP_PLUGIN_fatal ("cuCtxCreate error: %s", cuda_error (r));
+    }
+  else
+    ptx_dev->ctx_shared = true;
+
+  r = cuDeviceGetAttribute (&pi, CU_DEVICE_ATTRIBUTE_GPU_OVERLAP, dev);
+  if (r != CUDA_SUCCESS)
+    GOMP_PLUGIN_fatal ("cuDeviceGetAttribute error: %s", cuda_error (r));
+
+  ptx_dev->overlap = pi;
+
+  r = cuDeviceGetAttribute (&pi, CU_DEVICE_ATTRIBUTE_CAN_MAP_HOST_MEMORY, dev);
+  if (r != CUDA_SUCCESS)
+    GOMP_PLUGIN_fatal ("cuDeviceGetAttribute error: %s", cuda_error (r));
+
+  ptx_dev->map = pi;
+
+  r = cuDeviceGetAttribute (&pi, CU_DEVICE_ATTRIBUTE_CONCURRENT_KERNELS, dev);
+  if (r != CUDA_SUCCESS)
+    GOMP_PLUGIN_fatal ("cuDeviceGetAttribute error: %s", cuda_error (r));
+
+  ptx_dev->concur = pi;
+
+  r = cuDeviceGetAttribute (&pi, CU_DEVICE_ATTRIBUTE_COMPUTE_MODE, dev);
+  if (r != CUDA_SUCCESS)
+    GOMP_PLUGIN_fatal ("cuDeviceGetAttribute error: %s", cuda_error (r));
+
+  ptx_dev->mode = pi;
+
+  r = cuDeviceGetAttribute (&pi, CU_DEVICE_ATTRIBUTE_INTEGRATED, dev);
+  if (r != CUDA_SUCCESS)
+    GOMP_PLUGIN_fatal ("cuDeviceGetAttribute error: %s", cuda_error (r));
+
+  ptx_dev->mkern = pi;
+
+  r = cuDeviceGetAttribute (&async_engines,
+			    CU_DEVICE_ATTRIBUTE_ASYNC_ENGINE_COUNT, dev);
+  if (r != CUDA_SUCCESS)
+    async_engines = 1;
+
+  init_streams_for_device (ptx_dev, async_engines);
+
+  return (void *) ptx_dev;
+}
+
+static int
+nvptx_close_device (void *targ_data)
+{
+  CUresult r;
+  struct ptx_device *ptx_dev = targ_data;
+
+  if (!ptx_dev)
+    return 0;
+
+  fini_streams_for_device (ptx_dev);
+
+  if (!ptx_dev->ctx_shared)
+    {
+      r = cuCtxDestroy (ptx_dev->ctx);
+      if (r != CUDA_SUCCESS)
+	GOMP_PLUGIN_fatal ("cuCtxDestroy error: %s", cuda_error (r));
+    }
+
+  free (ptx_dev);
+
+  return 0;
+}
+
+static int
+nvptx_get_num_devices (void)
+{
+  int n;
+  CUresult r;
+
+  /* This function will be called before the plugin has been initialized in
+     order to enumerate available devices, but CUDA API routines can't be used
+     until cuInit has been called.  Just call it now (but don't yet do any
+     further initialization).  */
+  if (!ptx_inited)
+    cuInit (0);
+
+  r = cuDeviceGetCount (&n);
+  if (r!= CUDA_SUCCESS)
+    GOMP_PLUGIN_fatal ("cuDeviceGetCount error: %s", cuda_error (r));
+
+  return n;
+}
+
+
+static void
+link_ptx (CUmodule *module, char *ptx_code)
+{
+  CUjit_option opts[7];
+  void *optvals[7];
+  float elapsed = 0.0;
+#define LOGSIZE 8192
+  char elog[LOGSIZE];
+  char ilog[LOGSIZE];
+  unsigned long logsize = LOGSIZE;
+  CUlinkState linkstate;
+  CUresult r;
+  void *linkout;
+  size_t linkoutsize __attribute__ ((unused));
+
+  GOMP_PLUGIN_debug (0, "attempting to load:\n---\n%s\n---\n", ptx_code);
+
+  opts[0] = CU_JIT_WALL_TIME;
+  optvals[0] = &elapsed;
+
+  opts[1] = CU_JIT_INFO_LOG_BUFFER;
+  optvals[1] = &ilog[0];
+
+  opts[2] = CU_JIT_INFO_LOG_BUFFER_SIZE_BYTES;
+  optvals[2] = (void *) logsize;
+
+  opts[3] = CU_JIT_ERROR_LOG_BUFFER;
+  optvals[3] = &elog[0];
+
+  opts[4] = CU_JIT_ERROR_LOG_BUFFER_SIZE_BYTES;
+  optvals[4] = (void *) logsize;
+
+  opts[5] = CU_JIT_LOG_VERBOSE;
+  optvals[5] = (void *) 1;
+
+  opts[6] = CU_JIT_TARGET;
+  optvals[6] = (void *) CU_TARGET_COMPUTE_30;
+
+  r = cuLinkCreate (7, opts, optvals, &linkstate);
+  if (r != CUDA_SUCCESS)
+    GOMP_PLUGIN_fatal ("cuLinkCreate error: %s", cuda_error (r));
+
+  char *abort_ptx = ABORT_PTX;
+  r = cuLinkAddData (linkstate, CU_JIT_INPUT_PTX, abort_ptx,
+		     strlen (abort_ptx) + 1, 0, 0, 0, 0);
+  if (r != CUDA_SUCCESS)
+    {
+      GOMP_PLUGIN_error ("Link error log %s\n", &elog[0]);
+      GOMP_PLUGIN_fatal ("cuLinkAddData (abort) error: %s", cuda_error (r));
+    }
+
+  char *acc_on_device_ptx = ACC_ON_DEVICE_PTX;
+  r = cuLinkAddData (linkstate, CU_JIT_INPUT_PTX, acc_on_device_ptx,
+		     strlen (acc_on_device_ptx) + 1, 0, 0, 0, 0);
+  if (r != CUDA_SUCCESS)
+    {
+      GOMP_PLUGIN_error ("Link error log %s\n", &elog[0]);
+      GOMP_PLUGIN_fatal ("cuLinkAddData (acc_on_device) error: %s",
+			 cuda_error (r));
+    }
+
+  char *goacc_internal_ptx = GOACC_INTERNAL_PTX;
+  r = cuLinkAddData (linkstate, CU_JIT_INPUT_PTX, goacc_internal_ptx,
+		     strlen (goacc_internal_ptx) + 1, 0, 0, 0, 0);
+  if (r != CUDA_SUCCESS)
+    {
+      GOMP_PLUGIN_error ("Link error log %s\n", &elog[0]);
+      GOMP_PLUGIN_fatal ("cuLinkAddData (goacc_internal_ptx) error: %s",
+			 cuda_error (r));
+    }
+
+  r = cuLinkAddData (linkstate, CU_JIT_INPUT_PTX, ptx_code,
+              strlen (ptx_code) + 1, 0, 0, 0, 0);
+  if (r != CUDA_SUCCESS)
+    {
+      GOMP_PLUGIN_error ("Link error log %s\n", &elog[0]);
+      GOMP_PLUGIN_fatal ("cuLinkAddData (ptx_code) error: %s", cuda_error (r));
+    }
+
+  r = cuLinkComplete (linkstate, &linkout, &linkoutsize);
+  if (r != CUDA_SUCCESS)
+    GOMP_PLUGIN_fatal ("cuLinkComplete error: %s", cuda_error (r));
+
+  GOMP_PLUGIN_debug (0, "Link complete: %fms\n", elapsed);
+  GOMP_PLUGIN_debug (0, "Link log %s\n", &ilog[0]);
+
+  r = cuModuleLoadData (module, linkout);
+  if (r != CUDA_SUCCESS)
+    GOMP_PLUGIN_fatal ("cuModuleLoadData error: %s", cuda_error (r));
+}
+
+static void
+event_gc (bool memmap_lockable)
+{
+  struct ptx_event *ptx_event = ptx_events;
+  struct nvptx_thread *nvthd = nvptx_thread ();
+
+  pthread_mutex_lock (&ptx_event_lock);
+
+  while (ptx_event != NULL)
+    {
+      CUresult r;
+      struct ptx_event *e = ptx_event;
+
+      ptx_event = ptx_event->next;
+
+      if (e->ord != nvthd->ptx_dev->ord)
+	continue;
+
+      r = cuEventQuery (*e->evt);
+      if (r == CUDA_SUCCESS)
+	{
+	  CUevent *te;
+
+	  te = e->evt;
+
+	  switch (e->type)
+	    {
+	    case PTX_EVT_MEM:
+	    case PTX_EVT_SYNC:
+	      break;
+
+	    case PTX_EVT_KNL:
+	      map_pop (e->addr);
+	      break;
+
+	    case PTX_EVT_ASYNC_CLEANUP:
+	      {
+		/* The function gomp_plugin_async_unmap_vars needs to claim the
+		   memory-map splay tree lock for the current device, so we
+		   can't call it when one of our callers has already claimed
+		   the lock.  In that case, just delay the GC for this event
+		   until later.  */
+		if (!memmap_lockable)
+		  continue;
+
+		GOMP_PLUGIN_async_unmap_vars (e->addr);
+	      }
+	      break;
+	    }
+
+	  cuEventDestroy (*te);
+	  free ((void *)te);
+
+	  if (ptx_events == e)
+	    ptx_events = ptx_events->next;
+	  else
+	    {
+	      struct ptx_event *e_ = ptx_events;
+	      while (e_->next != e)
+		e_ = e_->next;
+	      e_->next = e_->next->next;
+	    }
+
+	  free (e);
+	}
+    }
+
+  pthread_mutex_unlock (&ptx_event_lock);
+}
+
+static void
+event_add (enum ptx_event_type type, CUevent *e, void *h)
+{
+  struct ptx_event *ptx_event;
+  struct nvptx_thread *nvthd = nvptx_thread ();
+
+  assert (type == PTX_EVT_MEM || type == PTX_EVT_KNL || type == PTX_EVT_SYNC
+	  || type == PTX_EVT_ASYNC_CLEANUP);
+
+  ptx_event = GOMP_PLUGIN_malloc (sizeof (struct ptx_event));
+  ptx_event->type = type;
+  ptx_event->evt = e;
+  ptx_event->addr = h;
+  ptx_event->ord = nvthd->ptx_dev->ord;
+
+  pthread_mutex_lock (&ptx_event_lock);
+
+  ptx_event->next = ptx_events;
+  ptx_events = ptx_event;
+
+  pthread_mutex_unlock (&ptx_event_lock);
+}
+
+void
+nvptx_exec (void (*fn), size_t mapnum, void **hostaddrs, void **devaddrs,
+	  size_t *sizes, unsigned short *kinds, int num_gangs, int num_workers,
+	  int vector_length, int async, void *targ_mem_desc)
+{
+  struct targ_fn_descriptor *targ_fn = (struct targ_fn_descriptor *) fn;
+  CUfunction function;
+  CUresult r;
+  int i;
+  struct ptx_stream *dev_str;
+  void *kargs[1];
+  void *hp, *dp;
+  unsigned int nthreads_in_block;
+  struct nvptx_thread *nvthd = nvptx_thread ();
+  const char *maybe_abort_msg = "(perhaps abort was called)";
+
+  function = targ_fn->fn;
+
+  dev_str = select_stream_for_async (async, pthread_self (), false, NULL);
+  assert (dev_str == nvthd->current_stream);
+
+  /* This reserves a chunk of a pre-allocated page of memory mapped on both
+     the host and the device. HP is a host pointer to the new chunk, and DP is
+     the corresponding device pointer.  */
+  map_push (dev_str, async, mapnum * sizeof (void *), &hp, &dp);
+
+  GOMP_PLUGIN_debug (0, "  %s: prepare mappings\n", __FUNCTION__);
+
+  /* Copy the array of arguments to the mapped page.  */
+  for (i = 0; i < mapnum; i++)
+    ((void **) hp)[i] = devaddrs[i];
+
+  /* Copy the (device) pointers to arguments to the device (dp and hp might in
+     fact have the same value on a unified-memory system).  */
+  r = cuMemcpy ((CUdeviceptr)dp, (CUdeviceptr)hp, mapnum * sizeof (void *));
+  if (r != CUDA_SUCCESS)
+    GOMP_PLUGIN_fatal ("cuMemcpy failed: %s", cuda_error (r));
+
+  GOMP_PLUGIN_debug (0, "  %s: kernel %s: launch\n", __FUNCTION__, targ_fn->name);
+
+  // OpenACC		CUDA
+  //
+  // num_gangs		blocks
+  // num_workers	warps (where a warp is equivalent to 32 threads)
+  // vector length	threads
+  //
+
+  /* The openacc vector_length clause 'determines the vector length to use for
+     vector or SIMD operations'.  The question is how to map this to CUDA.
+
+     In CUDA, the warp size is the vector length of a CUDA device.  However, the
+     CUDA interface abstracts away from that, and only shows us warp size
+     indirectly in maximum number of threads per block, which is a product of
+     warp size and the number of hyperthreads of a multiprocessor.
+
+     We choose to map openacc vector_length directly onto the number of threads
+     in a block, in the x dimension.  This is reflected in gcc code generation
+     that uses ThreadIdx.x to access vector elements.
+
+     Attempting to use an openacc vector_length of more than the maximum number
+     of threads per block will result in a cuda error.  */
+  nthreads_in_block = vector_length;
+
+  kargs[0] = &dp;
+  r = cuLaunchKernel (function,
+		      num_gangs, 1, 1,
+		      nthreads_in_block, 1, 1,
+		      0, dev_str->stream, kargs, 0);
+  if (r != CUDA_SUCCESS)
+    GOMP_PLUGIN_fatal ("cuLaunchKernel error: %s", cuda_error (r));
+
+#ifndef DISABLE_ASYNC
+  if (async < acc_async_noval)
+    {
+      r = cuStreamSynchronize (dev_str->stream);
+      if (r == CUDA_ERROR_LAUNCH_FAILED)
+	GOMP_PLUGIN_fatal ("cuStreamSynchronize error: %s %s\n", cuda_error (r),
+			   maybe_abort_msg);
+      else if (r != CUDA_SUCCESS)
+        GOMP_PLUGIN_fatal ("cuStreamSynchronize error: %s", cuda_error (r));
+    }
+  else
+    {
+      CUevent *e;
+
+      e = (CUevent *)GOMP_PLUGIN_malloc (sizeof (CUevent));
+
+      r = cuEventCreate (e, CU_EVENT_DISABLE_TIMING);
+      if (r == CUDA_ERROR_LAUNCH_FAILED)
+	GOMP_PLUGIN_fatal ("cuEventCreate error: %s %s\n", cuda_error (r),
+			   maybe_abort_msg);
+      else if (r != CUDA_SUCCESS)
+        GOMP_PLUGIN_fatal ("cuEventCreate error: %s", cuda_error (r));
+
+      event_gc (true);
+
+      r = cuEventRecord (*e, dev_str->stream);
+      if (r != CUDA_SUCCESS)
+        GOMP_PLUGIN_fatal ("cuEventRecord error: %s", cuda_error (r));
+
+      event_add (PTX_EVT_KNL, e, (void *)dev_str);
+    }
+#else
+  r = cuCtxSynchronize ();
+  if (r == CUDA_ERROR_LAUNCH_FAILED)
+    GOMP_PLUGIN_fatal ("cuCtxSynchronize error: %s %s\n", cuda_error (r),
+		       maybe_abort_msg);
+  else if (r != CUDA_SUCCESS)
+    GOMP_PLUGIN_fatal ("cuCtxSynchronize error: %s", cuda_error (r));
+#endif
+
+  GOMP_PLUGIN_debug (0, "  %s: kernel %s: finished\n", __FUNCTION__,
+		     targ_fn->name);
+
+#ifndef DISABLE_ASYNC
+  if (async < acc_async_noval)
+#endif
+    map_pop (dev_str);
+}
+
+void * openacc_get_current_cuda_context (void);
+
+static void *
+nvptx_alloc (size_t s)
+{
+  CUdeviceptr d;
+  CUresult r;
+
+  r = cuMemAlloc (&d, s);
+  if (r == CUDA_ERROR_OUT_OF_MEMORY)
+    return 0;
+  if (r != CUDA_SUCCESS)
+    GOMP_PLUGIN_fatal ("cuMemAlloc error: %s", cuda_error (r));
+  return (void *)d;
+}
+
+static void
+nvptx_free (void *p)
+{
+  CUresult r;
+  CUdeviceptr pb;
+  size_t ps;
+
+  r = cuMemGetAddressRange (&pb, &ps, (CUdeviceptr)p);
+  if (r != CUDA_SUCCESS)
+    GOMP_PLUGIN_fatal ("cuMemGetAddressRange error: %s", cuda_error (r));
+
+  if ((CUdeviceptr)p != pb)
+    GOMP_PLUGIN_fatal ("invalid device address");
+
+  r = cuMemFree ((CUdeviceptr)p);
+  if (r != CUDA_SUCCESS)
+    GOMP_PLUGIN_fatal ("cuMemFree error: %s", cuda_error (r));
+}
+
+static void *
+nvptx_host2dev (void *d, const void *h, size_t s)
+{
+  CUresult r;
+  CUdeviceptr pb;
+  size_t ps;
+  struct nvptx_thread *nvthd = nvptx_thread ();
+
+  if (!s)
+    return 0;
+
+  if (!d)
+    GOMP_PLUGIN_fatal ("invalid device address");
+
+  r = cuMemGetAddressRange (&pb, &ps, (CUdeviceptr)d);
+  if (r != CUDA_SUCCESS)
+    GOMP_PLUGIN_fatal ("cuMemGetAddressRange error: %s", cuda_error (r));
+
+  if (!pb)
+    GOMP_PLUGIN_fatal ("invalid device address");
+
+  if (!h)
+    GOMP_PLUGIN_fatal ("invalid host address");
+
+  if (d == h)
+    GOMP_PLUGIN_fatal ("invalid host or device address");
+
+  if ((void *)(d + s) > (void *)(pb + ps))
+    GOMP_PLUGIN_fatal ("invalid size");
+
+#ifndef DISABLE_ASYNC
+  if (nvthd->current_stream != nvthd->ptx_dev->null_stream)
+    {
+      CUevent *e;
+
+      e = (CUevent *)GOMP_PLUGIN_malloc (sizeof (CUevent));
+
+      r = cuEventCreate (e, CU_EVENT_DISABLE_TIMING);
+      if (r != CUDA_SUCCESS)
+        GOMP_PLUGIN_fatal ("cuEventCreate error: %s", cuda_error (r));
+
+      event_gc (false);
+
+      r = cuMemcpyHtoDAsync ((CUdeviceptr)d, h, s,
+			     nvthd->current_stream->stream);
+      if (r != CUDA_SUCCESS)
+        GOMP_PLUGIN_fatal ("cuMemcpyHtoDAsync error: %s", cuda_error (r));
+
+      r = cuEventRecord (*e, nvthd->current_stream->stream);
+      if (r != CUDA_SUCCESS)
+        GOMP_PLUGIN_fatal ("cuEventRecord error: %s", cuda_error (r));
+
+      event_add (PTX_EVT_MEM, e, (void *)h);
+    }
+  else
+#endif
+    {
+      r = cuMemcpyHtoD ((CUdeviceptr)d, h, s);
+      if (r != CUDA_SUCCESS)
+        GOMP_PLUGIN_fatal ("cuMemcpyHtoD error: %s", cuda_error (r));
+    }
+
+  return 0;
+}
+
+static void *
+nvptx_dev2host (void *h, const void *d, size_t s)
+{
+  CUresult r;
+  CUdeviceptr pb;
+  size_t ps;
+  struct nvptx_thread *nvthd = nvptx_thread ();
+
+  if (!s)
+    return 0;
+
+  if (!d)
+    GOMP_PLUGIN_fatal ("invalid device address");
+
+  r = cuMemGetAddressRange (&pb, &ps, (CUdeviceptr)d);
+  if (r != CUDA_SUCCESS)
+    GOMP_PLUGIN_fatal ("cuMemGetAddressRange error: %s", cuda_error (r));
+
+  if (!pb)
+    GOMP_PLUGIN_fatal ("invalid device address");
+
+  if (!h)
+    GOMP_PLUGIN_fatal ("invalid host address");
+
+  if (d == h)
+    GOMP_PLUGIN_fatal ("invalid host or device address");
+
+  if ((void *)(d + s) > (void *)(pb + ps))
+    GOMP_PLUGIN_fatal ("invalid size");
+
+#ifndef DISABLE_ASYNC
+  if (nvthd->current_stream != nvthd->ptx_dev->null_stream)
+    {
+      CUevent *e;
+
+      e = (CUevent *)GOMP_PLUGIN_malloc (sizeof (CUevent));
+
+      r = cuEventCreate (e, CU_EVENT_DISABLE_TIMING);
+      if (r != CUDA_SUCCESS)
+        GOMP_PLUGIN_fatal ("cuEventCreate error: %s\n", cuda_error (r));
+
+      event_gc (false);
+
+      r = cuMemcpyDtoHAsync (h, (CUdeviceptr)d, s,
+			     nvthd->current_stream->stream);
+      if (r != CUDA_SUCCESS)
+        GOMP_PLUGIN_fatal ("cuMemcpyDtoHAsync error: %s", cuda_error (r));
+
+      r = cuEventRecord (*e, nvthd->current_stream->stream);
+      if (r != CUDA_SUCCESS)
+        GOMP_PLUGIN_fatal ("cuEventRecord error: %s", cuda_error (r));
+
+      event_add (PTX_EVT_MEM, e, (void *)h);
+    }
+  else
+#endif
+    {
+      r = cuMemcpyDtoH (h, (CUdeviceptr)d, s);
+      if (r != CUDA_SUCCESS)
+	GOMP_PLUGIN_fatal ("cuMemcpyDtoH error: %s", cuda_error (r));
+    }
+
+  return 0;
+}
+
+static void
+nvptx_set_async (int async)
+{
+  struct nvptx_thread *nvthd = nvptx_thread ();
+  nvthd->current_stream
+    = select_stream_for_async (async, pthread_self (), true, NULL);
+}
+
+static int
+nvptx_async_test (int async)
+{
+  CUresult r;
+  struct ptx_stream *s;
+
+  s = select_stream_for_async (async, pthread_self (), false, NULL);
+
+  if (!s)
+    GOMP_PLUGIN_fatal ("unknown async %d", async);
+
+  r = cuStreamQuery (s->stream);
+  if (r == CUDA_SUCCESS)
+    {
+      /* The oacc-parallel.c:goacc_wait function calls this hook to determine
+	 whether all work has completed on this stream, and if so omits the call
+	 to the wait hook.  If that happens, event_gc might not get called
+	 (which prevents variables from getting unmapped and their associated
+	 device storage freed), so call it here.  */
+      event_gc (true);
+      return 1;
+    }
+  else if (r == CUDA_ERROR_NOT_READY)
+    return 0;
+
+  GOMP_PLUGIN_fatal ("cuStreamQuery error: %s", cuda_error (r));
+
+  return 0;
+}
+
+static int
+nvptx_async_test_all (void)
+{
+  struct ptx_stream *s;
+  pthread_t self = pthread_self ();
+  struct nvptx_thread *nvthd = nvptx_thread ();
+
+  pthread_mutex_lock (&nvthd->ptx_dev->stream_lock);
+
+  for (s = nvthd->ptx_dev->active_streams; s != NULL; s = s->next)
+    {
+      if ((s->multithreaded || pthread_equal (s->host_thread, self))
+	  && cuStreamQuery (s->stream) == CUDA_ERROR_NOT_READY)
+	{
+	  pthread_mutex_unlock (&nvthd->ptx_dev->stream_lock);
+	  return 0;
+	}
+    }
+
+  pthread_mutex_unlock (&nvthd->ptx_dev->stream_lock);
+
+  event_gc (true);
+
+  return 1;
+}
+
+static void
+nvptx_wait (int async)
+{
+  CUresult r;
+  struct ptx_stream *s;
+
+  s = select_stream_for_async (async, pthread_self (), false, NULL);
+
+  if (!s)
+    GOMP_PLUGIN_fatal ("unknown async %d", async);
+
+  r = cuStreamSynchronize (s->stream);
+  if (r != CUDA_SUCCESS)
+    GOMP_PLUGIN_fatal ("cuStreamSynchronize error: %s", cuda_error (r));
+
+  event_gc (true);
+}
+
+static void
+nvptx_wait_async (int async1, int async2)
+{
+  CUresult r;
+  CUevent *e;
+  struct ptx_stream *s1, *s2;
+  pthread_t self = pthread_self ();
+
+  /* The stream that is waiting (rather than being waited for) doesn't
+     necessarily have to exist already.  */
+  s2 = select_stream_for_async (async2, self, true, NULL);
+
+  s1 = select_stream_for_async (async1, self, false, NULL);
+  if (!s1)
+    GOMP_PLUGIN_fatal ("invalid async 1\n");
+
+  if (s1 == s2)
+    GOMP_PLUGIN_fatal ("identical parameters");
+
+  e = (CUevent *)GOMP_PLUGIN_malloc (sizeof (CUevent));
+
+  r = cuEventCreate (e, CU_EVENT_DISABLE_TIMING);
+  if (r != CUDA_SUCCESS)
+    GOMP_PLUGIN_fatal ("cuEventCreate error: %s", cuda_error (r));
+
+  event_gc (true);
+
+  r = cuEventRecord (*e, s1->stream);
+  if (r != CUDA_SUCCESS)
+    GOMP_PLUGIN_fatal ("cuEventRecord error: %s", cuda_error (r));
+
+  event_add (PTX_EVT_SYNC, e, NULL);
+
+  r = cuStreamWaitEvent (s2->stream, *e, 0);
+  if (r != CUDA_SUCCESS)
+    GOMP_PLUGIN_fatal ("cuStreamWaitEvent error: %s", cuda_error (r));
+}
+
+static void
+nvptx_wait_all (void)
+{
+  CUresult r;
+  struct ptx_stream *s;
+  pthread_t self = pthread_self ();
+  struct nvptx_thread *nvthd = nvptx_thread ();
+
+  pthread_mutex_lock (&nvthd->ptx_dev->stream_lock);
+
+  /* Wait for active streams initiated by this thread (or by multiple threads)
+     to complete.  */
+  for (s = nvthd->ptx_dev->active_streams; s != NULL; s = s->next)
+    {
+      if (s->multithreaded || pthread_equal (s->host_thread, self))
+	{
+	  r = cuStreamQuery (s->stream);
+	  if (r == CUDA_SUCCESS)
+	    continue;
+	  else if (r != CUDA_ERROR_NOT_READY)
+	    GOMP_PLUGIN_fatal ("cuStreamQuery error: %s", cuda_error (r));
+
+	  r = cuStreamSynchronize (s->stream);
+	  if (r != CUDA_SUCCESS)
+	    GOMP_PLUGIN_fatal ("cuStreamSynchronize error: %s", cuda_error (r));
+	}
+    }
+
+  pthread_mutex_unlock (&nvthd->ptx_dev->stream_lock);
+
+  event_gc (true);
+}
+
+static void
+nvptx_wait_all_async (int async)
+{
+  CUresult r;
+  struct ptx_stream *waiting_stream, *other_stream;
+  CUevent *e;
+  struct nvptx_thread *nvthd = nvptx_thread ();
+  pthread_t self = pthread_self ();
+
+  /* The stream doing the waiting.  This could be the first mention of the
+     stream, so create it if necessary.  */
+  waiting_stream
+    = select_stream_for_async (async, pthread_self (), true, NULL);
+
+  /* Launches on the null stream already block on other streams in the
+     context.  */
+  if (!waiting_stream || waiting_stream == nvthd->ptx_dev->null_stream)
+    return;
+
+  event_gc (true);
+
+  pthread_mutex_lock (&nvthd->ptx_dev->stream_lock);
+
+  for (other_stream = nvthd->ptx_dev->active_streams;
+       other_stream != NULL;
+       other_stream = other_stream->next)
+    {
+      if (!other_stream->multithreaded
+	  && !pthread_equal (other_stream->host_thread, self))
+	continue;
+
+      e = (CUevent *) GOMP_PLUGIN_malloc (sizeof (CUevent));
+
+      r = cuEventCreate (e, CU_EVENT_DISABLE_TIMING);
+      if (r != CUDA_SUCCESS)
+	GOMP_PLUGIN_fatal ("cuEventCreate error: %s", cuda_error (r));
+
+      /* Record an event on the waited-for stream.  */
+      r = cuEventRecord (*e, other_stream->stream);
+      if (r != CUDA_SUCCESS)
+	GOMP_PLUGIN_fatal ("cuEventRecord error: %s", cuda_error (r));
+
+      event_add (PTX_EVT_SYNC, e, NULL);
+
+      r = cuStreamWaitEvent (waiting_stream->stream, *e, 0);
+      if (r != CUDA_SUCCESS)
+	GOMP_PLUGIN_fatal ("cuStreamWaitEvent error: %s", cuda_error (r));
+   }
+
+  pthread_mutex_unlock (&nvthd->ptx_dev->stream_lock);
+}
+
+static void *
+nvptx_get_current_cuda_device (void)
+{
+  struct nvptx_thread *nvthd = nvptx_thread ();
+
+  if (!nvthd || !nvthd->ptx_dev)
+    return NULL;
+
+  return &nvthd->ptx_dev->dev;
+}
+
+static void *
+nvptx_get_current_cuda_context (void)
+{
+  struct nvptx_thread *nvthd = nvptx_thread ();
+
+  if (!nvthd || !nvthd->ptx_dev)
+    return NULL;
+
+  return nvthd->ptx_dev->ctx;
+}
+
+static void *
+nvptx_get_cuda_stream (int async)
+{
+  struct ptx_stream *s;
+  struct nvptx_thread *nvthd = nvptx_thread ();
+
+  if (!nvthd || !nvthd->ptx_dev)
+    return NULL;
+
+  s = select_stream_for_async (async, pthread_self (), false, NULL);
+
+  return s ? s->stream : NULL;
+}
+
+static int
+nvptx_set_cuda_stream (int async, void *stream)
+{
+  struct ptx_stream *oldstream;
+  pthread_t self = pthread_self ();
+  struct nvptx_thread *nvthd = nvptx_thread ();
+
+  pthread_mutex_lock (&nvthd->ptx_dev->stream_lock);
+
+  if (async < 0)
+    GOMP_PLUGIN_fatal ("bad async %d", async);
+
+  /* We have a list of active streams and an array mapping async values to
+     entries of that list.  We need to take "ownership" of the passed-in stream,
+     and add it to our list, removing the previous entry also (if there was one)
+     in order to prevent resource leaks.  Note the potential for surprise
+     here: maybe we should keep track of passed-in streams and leave it up to
+     the user to tidy those up, but that doesn't work for stream handles
+     returned from acc_get_cuda_stream above...  */
+
+  oldstream = select_stream_for_async (async, self, false, NULL);
+
+  if (oldstream)
+    {
+      if (nvthd->ptx_dev->active_streams == oldstream)
+	nvthd->ptx_dev->active_streams = nvthd->ptx_dev->active_streams->next;
+      else
+	{
+	  struct ptx_stream *s = nvthd->ptx_dev->active_streams;
+	  while (s->next != oldstream)
+	    s = s->next;
+	  s->next = s->next->next;
+	}
+
+      cuStreamDestroy (oldstream->stream);
+      map_fini (oldstream);
+      free (oldstream);
+    }
+
+  pthread_mutex_unlock (&nvthd->ptx_dev->stream_lock);
+
+  (void) select_stream_for_async (async, self, true, (CUstream) stream);
+
+  return 1;
+}
+
+/* Plugin entry points.  */
+
+const char *
+GOMP_OFFLOAD_get_name (void)
+{
+  return "nvptx";
+}
+
+unsigned int
+GOMP_OFFLOAD_get_caps (void)
+{
+  return GOMP_OFFLOAD_CAP_OPENACC_200;
+}
+
+int
+GOMP_OFFLOAD_get_type (void)
+{
+  return OFFLOAD_TARGET_TYPE_NVIDIA_PTX;
+}
+
+int
+GOMP_OFFLOAD_get_num_devices (void)
+{
+  return nvptx_get_num_devices ();
+}
+
+static void **kernel_target_data;
+static void **kernel_host_table;
+
+void
+GOMP_OFFLOAD_register_image (void *host_table, void *target_data)
+{
+  kernel_target_data = target_data;
+  kernel_host_table = host_table;
+}
+
+void
+GOMP_OFFLOAD_init_device (int n __attribute__ ((unused)))
+{
+  (void) nvptx_init ();
+}
+
+void
+GOMP_OFFLOAD_fini_device (int n __attribute__ ((unused)))
+{
+  nvptx_fini ();
+}
+
+int
+GOMP_OFFLOAD_get_table (int n __attribute__ ((unused)),
+			struct mapping_table **tablep)
+{
+  CUmodule module;
+  void **fn_table;
+  char **fn_names;
+  int fn_entries, i;
+  CUresult r;
+  struct targ_fn_descriptor *targ_fns;
+
+  if (nvptx_init () <= 0)
+    return 0;
+
+  /* This isn't an error, because an image may legitimately have no offloaded
+     regions and so will not call GOMP_offload_register.  */
+  if (kernel_target_data == NULL)
+    return 0;
+
+  link_ptx (&module, kernel_target_data[0]);
+
+  /* kernel_target_data[0] -> ptx code
+     kernel_target_data[1] -> variable mappings
+     kernel_target_data[2] -> array of kernel names in ascii
+
+     kernel_host_table[0] -> start of function addresses (__offload_func_table)
+     kernel_host_table[1] -> end of function addresses (__offload_funcs_end)
+
+     The array of kernel names and the functions addresses form a
+     one-to-one correspondence.  */
+
+  fn_table = kernel_host_table[0];
+  fn_names = (char **) kernel_target_data[2];
+  fn_entries = (kernel_host_table[1] - kernel_host_table[0]) / sizeof (void *);
+
+  *tablep = GOMP_PLUGIN_malloc (sizeof (struct mapping_table) * fn_entries);
+  targ_fns = GOMP_PLUGIN_malloc (sizeof (struct targ_fn_descriptor)
+				 * fn_entries);
+
+  for (i = 0; i < fn_entries; i++)
+    {
+      CUfunction function;
+
+      r = cuModuleGetFunction (&function, module, fn_names[i]);
+      if (r != CUDA_SUCCESS)
+	GOMP_PLUGIN_fatal ("cuModuleGetFunction error: %s", cuda_error (r));
+
+      targ_fns[i].fn = function;
+      targ_fns[i].name = (const char *) fn_names[i];
+
+      (*tablep)[i].host_start = (uintptr_t) fn_table[i];
+      (*tablep)[i].host_end = (*tablep)[i].host_start + 1;
+      (*tablep)[i].tgt_start = (uintptr_t) &targ_fns[i];
+      (*tablep)[i].tgt_end = (*tablep)[i].tgt_start + 1;
+    }
+
+  return fn_entries;
+}
+
+void *
+GOMP_OFFLOAD_alloc (int n __attribute__ ((unused)), size_t size)
+{
+  return nvptx_alloc (size);
+}
+
+void
+GOMP_OFFLOAD_free (int n __attribute__ ((unused)), void *ptr)
+{
+  nvptx_free (ptr);
+}
+
+void *
+GOMP_OFFLOAD_dev2host (int ord __attribute__ ((unused)), void *dst,
+		       const void *src, size_t n)
+{
+  return nvptx_dev2host (dst, src, n);
+}
+
+void *
+GOMP_OFFLOAD_host2dev (int ord __attribute__ ((unused)), void *dst,
+		       const void *src, size_t n)
+{
+  return nvptx_host2dev (dst, src, n);
+}
+
+void (*device_run) (int n, void *fn_ptr, void *vars) = NULL;
+
+void
+GOMP_OFFLOAD_openacc_parallel (void (*fn) (void *), size_t mapnum,
+			       void **hostaddrs, void **devaddrs, size_t *sizes,
+			       unsigned short *kinds, int num_gangs,
+			       int num_workers, int vector_length, int async,
+			       void *targ_mem_desc)
+{
+  nvptx_exec (fn, mapnum, hostaddrs, devaddrs, sizes, kinds, num_gangs,
+	    num_workers, vector_length, async, targ_mem_desc);
+}
+
+void *
+GOMP_OFFLOAD_openacc_open_device (int n)
+{
+  return nvptx_open_device (n);
+}
+
+int
+GOMP_OFFLOAD_openacc_close_device (void *h)
+{
+  return nvptx_close_device (h);
+}
+
+void
+GOMP_OFFLOAD_openacc_set_device_num (int n)
+{
+  struct nvptx_thread *nvthd = nvptx_thread ();
+
+  assert (n >= 0);
+
+  if (!nvthd->ptx_dev || nvthd->ptx_dev->ord != n)
+    (void) nvptx_open_device (n);
+}
+
+/* This can be called before the device is "opened" for the current thread, in
+   which case we can't tell which device number should be returned.  We don't
+   actually want to open the device here, so just return -1 and let the caller
+   (oacc-init.c:acc_get_device_num) handle it.  */
+
+int
+GOMP_OFFLOAD_openacc_get_device_num (void)
+{
+  struct nvptx_thread *nvthd = nvptx_thread ();
+
+  if (nvthd && nvthd->ptx_dev)
+    return nvthd->ptx_dev->ord;
+  else
+    return -1;
+}
+
+void
+GOMP_OFFLOAD_openacc_register_async_cleanup (void *targ_mem_desc)
+{
+  CUevent *e;
+  CUresult r;
+  struct nvptx_thread *nvthd = nvptx_thread ();
+
+  e = (CUevent *) GOMP_PLUGIN_malloc (sizeof (CUevent));
+
+  r = cuEventCreate (e, CU_EVENT_DISABLE_TIMING);
+  if (r != CUDA_SUCCESS)
+    GOMP_PLUGIN_fatal ("cuEventCreate error: %s", cuda_error (r));
+
+  r = cuEventRecord (*e, nvthd->current_stream->stream);
+  if (r != CUDA_SUCCESS)
+    GOMP_PLUGIN_fatal ("cuEventRecord error: %s", cuda_error (r));
+
+  event_add (PTX_EVT_ASYNC_CLEANUP, e, targ_mem_desc);
+}
+
+int
+GOMP_OFFLOAD_openacc_async_test (int async)
+{
+  return nvptx_async_test (async);
+}
+
+int
+GOMP_OFFLOAD_openacc_async_test_all (void)
+{
+  return nvptx_async_test_all ();
+}
+
+void
+GOMP_OFFLOAD_openacc_async_wait (int async)
+{
+  nvptx_wait (async);
+}
+
+void
+GOMP_OFFLOAD_openacc_async_wait_async (int async1, int async2)
+{
+  nvptx_wait_async (async1, async2);
+}
+
+void
+GOMP_OFFLOAD_openacc_async_wait_all (void)
+{
+  nvptx_wait_all ();
+}
+
+void
+GOMP_OFFLOAD_openacc_async_wait_all_async (int async)
+{
+  nvptx_wait_all_async (async);
+}
+
+void
+GOMP_OFFLOAD_openacc_async_set_async (int async)
+{
+  nvptx_set_async (async);
+}
+
+void *
+GOMP_OFFLOAD_openacc_create_thread_data (void *targ_data)
+{
+  struct ptx_device *ptx_dev = (struct ptx_device *) targ_data;
+  struct nvptx_thread *nvthd
+    = GOMP_PLUGIN_malloc (sizeof (struct nvptx_thread));
+  CUresult r;
+  CUcontext thd_ctx;
+
+  r = cuCtxGetCurrent (&thd_ctx);
+  if (r != CUDA_SUCCESS)
+    GOMP_PLUGIN_fatal ("cuCtxGetCurrent error: %s", cuda_error (r));
+
+  assert (ptx_dev->ctx);
+
+  if (!thd_ctx)
+    {
+      r = cuCtxPushCurrent (ptx_dev->ctx);
+      if (r != CUDA_SUCCESS)
+	GOMP_PLUGIN_fatal ("cuCtxPushCurrent error: %s", cuda_error (r));
+    }
+
+  nvthd->current_stream = ptx_dev->null_stream;
+  nvthd->ptx_dev = ptx_dev;
+
+  return (void *) nvthd;
+}
+
+void
+GOMP_OFFLOAD_openacc_destroy_thread_data (void *data)
+{
+  free (data);
+}
+
+void *
+GOMP_OFFLOAD_openacc_get_current_cuda_device (void)
+{
+  return nvptx_get_current_cuda_device ();
+}
+
+void *
+GOMP_OFFLOAD_openacc_get_current_cuda_context (void)
+{
+  return nvptx_get_current_cuda_context ();
+}
+
+/* NOTE: This returns a CUstream, not a ptx_stream pointer.  */
+
+void *
+GOMP_OFFLOAD_openacc_get_cuda_stream (int async)
+{
+  return nvptx_get_cuda_stream (async);
+}
+
+/* NOTE: This takes a CUstream, not a ptx_stream pointer.  */
+
+int
+GOMP_OFFLOAD_openacc_set_cuda_stream (int async, void *stream)
+{
+  return nvptx_set_cuda_stream (async, stream);
+}
diff --git a/libgomp/splay-tree.c b/libgomp/splay-tree.c
new file mode 100644
index 00000000000..030ca8f6c45
--- /dev/null
+++ b/libgomp/splay-tree.c
@@ -0,0 +1,217 @@
+/* A splay-tree datatype.
+   Copyright (C) 1998-2015 Free Software Foundation, Inc.
+   Contributed by Mark Mitchell (mark@markmitchell.com).
+
+   This file is part of the GNU Offloading and Multi Processing Library
+   (libgomp).
+
+   Libgomp is free software; you can redistribute it and/or modify it
+   under the terms of the GNU General Public License as published by
+   the Free Software Foundation; either version 3, or (at your option)
+   any later version.
+
+   Libgomp is distributed in the hope that it will be useful, but WITHOUT ANY
+   WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
+   FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+   more details.
+
+   Under Section 7 of GPL version 3, you are granted additional
+   permissions described in the GCC Runtime Library Exception, version
+   3.1, as published by the Free Software Foundation.
+
+   You should have received a copy of the GNU General Public License and
+   a copy of the GCC Runtime Library Exception along with this program;
+   see the files COPYING3 and COPYING.RUNTIME respectively.  If not, see
+   <http://www.gnu.org/licenses/>.  */
+
+/* The splay tree code copied from include/splay-tree.h and adjusted,
+   so that all the data lives directly in splay_tree_node_s structure
+   and no extra allocations are needed.  */
+
+/* For an easily readable description of splay-trees, see:
+
+     Lewis, Harry R. and Denenberg, Larry.  Data Structures and Their
+     Algorithms.  Harper-Collins, Inc.  1991.
+
+   The major feature of splay trees is that all basic tree operations
+   are amortized O(log n) time for a tree with n nodes.  */
+
+#include "libgomp.h"
+#include "splay-tree.h"
+
+extern int splay_compare (splay_tree_key, splay_tree_key);
+
+/* Rotate the edge joining the left child N with its parent P.  PP is the
+   grandparents' pointer to P.  */
+
+static inline void
+rotate_left (splay_tree_node *pp, splay_tree_node p, splay_tree_node n)
+{
+  splay_tree_node tmp;
+  tmp = n->right;
+  n->right = p;
+  p->left = tmp;
+  *pp = n;
+}
+
+/* Rotate the edge joining the right child N with its parent P.  PP is the
+   grandparents' pointer to P.  */
+
+static inline void
+rotate_right (splay_tree_node *pp, splay_tree_node p, splay_tree_node n)
+{
+  splay_tree_node tmp;
+  tmp = n->left;
+  n->left = p;
+  p->right = tmp;
+  *pp = n;
+}
+
+/* Bottom up splay of KEY.  */
+
+static void
+splay_tree_splay (splay_tree sp, splay_tree_key key)
+{
+  if (sp->root == NULL)
+    return;
+
+  do {
+    int cmp1, cmp2;
+    splay_tree_node n, c;
+
+    n = sp->root;
+    cmp1 = splay_compare (key, &n->key);
+
+    /* Found.  */
+    if (cmp1 == 0)
+      return;
+
+    /* Left or right?  If no child, then we're done.  */
+    if (cmp1 < 0)
+      c = n->left;
+    else
+      c = n->right;
+    if (!c)
+      return;
+
+    /* Next one left or right?  If found or no child, we're done
+       after one rotation.  */
+    cmp2 = splay_compare (key, &c->key);
+    if (cmp2 == 0
+	|| (cmp2 < 0 && !c->left)
+	|| (cmp2 > 0 && !c->right))
+      {
+	if (cmp1 < 0)
+	  rotate_left (&sp->root, n, c);
+	else
+	  rotate_right (&sp->root, n, c);
+	return;
+      }
+
+    /* Now we have the four cases of double-rotation.  */
+    if (cmp1 < 0 && cmp2 < 0)
+      {
+	rotate_left (&n->left, c, c->left);
+	rotate_left (&sp->root, n, n->left);
+      }
+    else if (cmp1 > 0 && cmp2 > 0)
+      {
+	rotate_right (&n->right, c, c->right);
+	rotate_right (&sp->root, n, n->right);
+      }
+    else if (cmp1 < 0 && cmp2 > 0)
+      {
+	rotate_right (&n->left, c, c->right);
+	rotate_left (&sp->root, n, n->left);
+      }
+    else if (cmp1 > 0 && cmp2 < 0)
+      {
+	rotate_left (&n->right, c, c->left);
+	rotate_right (&sp->root, n, n->right);
+      }
+  } while (1);
+}
+
+/* Insert a new NODE into SP.  The NODE shouldn't exist in the tree.  */
+
+attribute_hidden void
+splay_tree_insert (splay_tree sp, splay_tree_node node)
+{
+  int comparison = 0;
+
+  splay_tree_splay (sp, &node->key);
+
+  if (sp->root)
+    comparison = splay_compare (&sp->root->key, &node->key);
+
+  if (sp->root && comparison == 0)
+    gomp_fatal ("Duplicate node");
+  else
+    {
+      /* Insert it at the root.  */
+      if (sp->root == NULL)
+	node->left = node->right = NULL;
+      else if (comparison < 0)
+	{
+	  node->left = sp->root;
+	  node->right = node->left->right;
+	  node->left->right = NULL;
+	}
+      else
+	{
+	  node->right = sp->root;
+	  node->left = node->right->left;
+	  node->right->left = NULL;
+	}
+
+      sp->root = node;
+    }
+}
+
+/* Remove node with KEY from SP.  It is not an error if it did not exist.  */
+
+attribute_hidden void
+splay_tree_remove (splay_tree sp, splay_tree_key key)
+{
+  splay_tree_splay (sp, key);
+
+  if (sp->root && splay_compare (&sp->root->key, key) == 0)
+    {
+      splay_tree_node left, right;
+
+      left = sp->root->left;
+      right = sp->root->right;
+
+      /* One of the children is now the root.  Doesn't matter much
+	 which, so long as we preserve the properties of the tree.  */
+      if (left)
+	{
+	  sp->root = left;
+
+	  /* If there was a right child as well, hang it off the
+	     right-most leaf of the left child.  */
+	  if (right)
+	    {
+	      while (left->right)
+		left = left->right;
+	      left->right = right;
+	    }
+	}
+      else
+	sp->root = right;
+    }
+}
+
+/* Lookup KEY in SP, returning NODE if present, and NULL
+   otherwise.  */
+
+attribute_hidden splay_tree_key
+splay_tree_lookup (splay_tree sp, splay_tree_key key)
+{
+  splay_tree_splay (sp, key);
+
+  if (sp->root && splay_compare (&sp->root->key, key) == 0)
+    return &sp->root->key;
+  else
+    return NULL;
+}
diff --git a/libgomp/splay-tree.h b/libgomp/splay-tree.h
index 1296be65f87..085021cf3d6 100644
--- a/libgomp/splay-tree.h
+++ b/libgomp/splay-tree.h
@@ -43,6 +43,9 @@ typedef struct splay_tree_key_s *splay_tree_key;
    The major feature of splay trees is that all basic tree operations
    are amortized O(log n) time for a tree with n nodes.  */
 
+#ifndef _SPLAY_TREE_H
+#define _SPLAY_TREE_H 1
+
 /* The nodes in the splay tree.  */
 struct splay_tree_node_s {
   struct splay_tree_key_s key;
@@ -56,177 +59,8 @@ struct splay_tree_s {
   splay_tree_node root;
 };
 
-/* Rotate the edge joining the left child N with its parent P.  PP is the
-   grandparents' pointer to P.  */
-
-static inline void
-rotate_left (splay_tree_node *pp, splay_tree_node p, splay_tree_node n)
-{
-  splay_tree_node tmp;
-  tmp = n->right;
-  n->right = p;
-  p->left = tmp;
-  *pp = n;
-}
-
-/* Rotate the edge joining the right child N with its parent P.  PP is the
-   grandparents' pointer to P.  */
-
-static inline void
-rotate_right (splay_tree_node *pp, splay_tree_node p, splay_tree_node n)
-{
-  splay_tree_node tmp;
-  tmp = n->left;
-  n->left = p;
-  p->right = tmp;
-  *pp = n;
-}
-
-/* Bottom up splay of KEY.  */
-
-static void
-splay_tree_splay (splay_tree sp, splay_tree_key key)
-{
-  if (sp->root == NULL)
-    return;
-
-  do {
-    int cmp1, cmp2;
-    splay_tree_node n, c;
-
-    n = sp->root;
-    cmp1 = splay_compare (key, &n->key);
-
-    /* Found.  */
-    if (cmp1 == 0)
-      return;
-
-    /* Left or right?  If no child, then we're done.  */
-    if (cmp1 < 0)
-      c = n->left;
-    else
-      c = n->right;
-    if (!c)
-      return;
-
-    /* Next one left or right?  If found or no child, we're done
-       after one rotation.  */
-    cmp2 = splay_compare (key, &c->key);
-    if (cmp2 == 0
-	|| (cmp2 < 0 && !c->left)
-	|| (cmp2 > 0 && !c->right))
-      {
-	if (cmp1 < 0)
-	  rotate_left (&sp->root, n, c);
-	else
-	  rotate_right (&sp->root, n, c);
-	return;
-      }
-
-    /* Now we have the four cases of double-rotation.  */
-    if (cmp1 < 0 && cmp2 < 0)
-      {
-	rotate_left (&n->left, c, c->left);
-	rotate_left (&sp->root, n, n->left);
-      }
-    else if (cmp1 > 0 && cmp2 > 0)
-      {
-	rotate_right (&n->right, c, c->right);
-	rotate_right (&sp->root, n, n->right);
-      }
-    else if (cmp1 < 0 && cmp2 > 0)
-      {
-	rotate_right (&n->left, c, c->right);
-	rotate_left (&sp->root, n, n->left);
-      }
-    else if (cmp1 > 0 && cmp2 < 0)
-      {
-	rotate_left (&n->right, c, c->left);
-	rotate_right (&sp->root, n, n->right);
-      }
-  } while (1);
-}
-
-/* Insert a new NODE into SP.  The NODE shouldn't exist in the tree.  */
-
-static void
-splay_tree_insert (splay_tree sp, splay_tree_node node)
-{
-  int comparison = 0;
-
-  splay_tree_splay (sp, &node->key);
-
-  if (sp->root)
-    comparison = splay_compare (&sp->root->key, &node->key);
-
-  if (sp->root && comparison == 0)
-    abort ();
-  else
-    {
-      /* Insert it at the root.  */
-      if (sp->root == NULL)
-	node->left = node->right = NULL;
-      else if (comparison < 0)
-	{
-	  node->left = sp->root;
-	  node->right = node->left->right;
-	  node->left->right = NULL;
-	}
-      else
-	{
-	  node->right = sp->root;
-	  node->left = node->right->left;
-	  node->right->left = NULL;
-	}
-
-      sp->root = node;
-    }
-}
-
-/* Remove node with KEY from SP.  It is not an error if it did not exist.  */
-
-static void
-splay_tree_remove (splay_tree sp, splay_tree_key key)
-{
-  splay_tree_splay (sp, key);
-
-  if (sp->root && splay_compare (&sp->root->key, key) == 0)
-    {
-      splay_tree_node left, right;
-
-      left = sp->root->left;
-      right = sp->root->right;
-
-      /* One of the children is now the root.  Doesn't matter much
-	 which, so long as we preserve the properties of the tree.  */
-      if (left)
-	{
-	  sp->root = left;
-
-	  /* If there was a right child as well, hang it off the
-	     right-most leaf of the left child.  */
-	  if (right)
-	    {
-	      while (left->right)
-		left = left->right;
-	      left->right = right;
-	    }
-	}
-      else
-	sp->root = right;
-    }
-}
-
-/* Lookup KEY in SP, returning NODE if present, and NULL
-   otherwise.  */
-
-static splay_tree_key
-splay_tree_lookup (splay_tree sp, splay_tree_key key)
-{
-  splay_tree_splay (sp, key);
+extern splay_tree_key splay_tree_lookup (splay_tree, splay_tree_key);
+extern void splay_tree_insert (splay_tree, splay_tree_node);
+extern void splay_tree_remove (splay_tree, splay_tree_key);
 
-  if (sp->root && splay_compare (&sp->root->key, key) == 0)
-    return &sp->root->key;
-  else
-    return NULL;
-}
+#endif /* _SPLAY_TREE_H */
diff --git a/libgomp/target.c b/libgomp/target.c
index ec097de9ca2..83ad51108d0 100644
--- a/libgomp/target.c
+++ b/libgomp/target.c
@@ -27,11 +27,14 @@
 
 #include "config.h"
 #include "libgomp.h"
-#include "libgomp_target.h"
+#include "oacc-plugin.h"
+#include "oacc-int.h"
+#include "gomp-constants.h"
 #include <limits.h>
 #include <stdbool.h>
 #include <stdlib.h>
 #include <string.h>
+#include <assert.h>
 
 #ifdef PLUGIN_SUPPORT
 #include <dlfcn.h>
@@ -39,52 +42,9 @@
 
 static void gomp_target_init (void);
 
+/* The whole initialization code for offloading plugins is only run one.  */
 static pthread_once_t gomp_is_initialized = PTHREAD_ONCE_INIT;
 
-/* Forward declaration for a node in the tree.  */
-typedef struct splay_tree_node_s *splay_tree_node;
-typedef struct splay_tree_s *splay_tree;
-typedef struct splay_tree_key_s *splay_tree_key;
-
-struct target_mem_desc {
-  /* Reference count.  */
-  uintptr_t refcount;
-  /* All the splay nodes allocated together.  */
-  splay_tree_node array;
-  /* Start of the target region.  */
-  uintptr_t tgt_start;
-  /* End of the targer region.  */
-  uintptr_t tgt_end;
-  /* Handle to free.  */
-  void *to_free;
-  /* Previous target_mem_desc.  */
-  struct target_mem_desc *prev;
-  /* Number of items in following list.  */
-  size_t list_count;
-
-  /* Corresponding target device descriptor.  */
-  struct gomp_device_descr *device_descr;
-
-  /* List of splay keys to remove (or decrease refcount)
-     at the end of region.  */
-  splay_tree_key list[];
-};
-
-struct splay_tree_key_s {
-  /* Address of the host object.  */
-  uintptr_t host_start;
-  /* Address immediately after the host object.  */
-  uintptr_t host_end;
-  /* Descriptor of the target memory.  */
-  struct target_mem_desc *tgt;
-  /* Offset from tgt->tgt_start to the start of the target object.  */
-  uintptr_t tgt_offset;
-  /* Reference count.  */
-  uintptr_t refcount;
-  /* True if data should be copied from device to host at the end.  */
-  bool copy_from;
-};
-
 /* This structure describes an offload image.
    It contains type of the target device, pointer to host table descriptor, and
    pointer to target data.  */
@@ -106,9 +66,12 @@ static struct gomp_device_descr *devices;
 /* Total number of available devices.  */
 static int num_devices;
 
+/* Number of GOMP_OFFLOAD_CAP_OPENMP_400 devices.  */
+static int num_devices_openmp;
+
 /* The comparison function.  */
 
-static int
+attribute_hidden int
 splay_compare (splay_tree_key x, splay_tree_key y)
 {
   if (x->host_start == x->host_end
@@ -123,54 +86,23 @@ splay_compare (splay_tree_key x, splay_tree_key y)
 
 #include "splay-tree.h"
 
-/* This structure describes accelerator device.
-   It contains ID-number of the device, its type, function handlers for
-   interaction with the device, and information about mapped memory.  */
-struct gomp_device_descr
+attribute_hidden void
+gomp_init_targets_once (void)
 {
-  /* This is the ID number of device.  It could be specified in DEVICE-clause of
-     TARGET construct.  */
-  int id;
-
-  /* This is the ID number of device among devices of the same type.  */
-  int target_id;
-
-  /* This is the TYPE of device.  */
-  enum offload_target_type type;
-
-  /* Set to true when device is initialized.  */
-  bool is_initialized;
-
-  /* Function handlers.  */
-  int (*get_type_func) (void);
-  int (*get_num_devices_func) (void);
-  void (*register_image_func) (void *, void *);
-  void (*init_device_func) (int);
-  int (*get_table_func) (int, void *);
-  void *(*alloc_func) (int, size_t);
-  void (*free_func) (int, void *);
-  void *(*host2dev_func) (int, void *, const void *, size_t);
-  void *(*dev2host_func) (int, void *, const void *, size_t);
-  void (*run_func) (int, void *, void *);
-
-  /* Splay tree containing information about mapped memory regions.  */
-  struct splay_tree_s dev_splay_tree;
-
-  /* Mutex for operating with the splay tree and other shared structures.  */
-  gomp_mutex_t dev_env_lock;
-};
+  (void) pthread_once (&gomp_is_initialized, gomp_target_init);
+}
 
 attribute_hidden int
 gomp_get_num_devices (void)
 {
-  (void) pthread_once (&gomp_is_initialized, gomp_target_init);
-  return num_devices;
+  gomp_init_targets_once ();
+  return num_devices_openmp;
 }
 
 static struct gomp_device_descr *
 resolve_device (int device_id)
 {
-  if (device_id == -1)
+  if (device_id == GOMP_DEVICE_ICV)
     {
       struct gomp_task_icv *icv = gomp_icv (false);
       device_id = icv->default_device_var;
@@ -190,27 +122,39 @@ static inline void
 gomp_map_vars_existing (splay_tree_key oldn, splay_tree_key newn,
 			unsigned char kind)
 {
-  if (oldn->host_start > newn->host_start
+  if ((kind & GOMP_MAP_FLAG_FORCE)
+      || oldn->host_start > newn->host_start
       || oldn->host_end < newn->host_end)
-    gomp_fatal ("Trying to map into device [%p..%p) object when"
+    gomp_fatal ("Trying to map into device [%p..%p) object when "
 		"[%p..%p) is already mapped",
 		(void *) newn->host_start, (void *) newn->host_end,
 		(void *) oldn->host_start, (void *) oldn->host_end);
   oldn->refcount++;
 }
 
-static struct target_mem_desc *
+static int
+get_kind (bool is_openacc, void *kinds, int idx)
+{
+  return is_openacc ? ((unsigned short *) kinds)[idx]
+		    : ((unsigned char *) kinds)[idx];
+}
+
+attribute_hidden struct target_mem_desc *
 gomp_map_vars (struct gomp_device_descr *devicep, size_t mapnum,
-	       void **hostaddrs, size_t *sizes, unsigned char *kinds,
-	       bool is_target)
+	       void **hostaddrs, void **devaddrs, size_t *sizes, void *kinds,
+	       bool is_openacc, bool is_target)
 {
   size_t i, tgt_align, tgt_size, not_found_cnt = 0;
+  const int rshift = is_openacc ? 8 : 3;
+  const int typemask = is_openacc ? 0xff : 0x7;
+  struct gomp_memory_mapping *mm = &devicep->mem_map;
   struct splay_tree_key_s cur_node;
   struct target_mem_desc *tgt
     = gomp_malloc (sizeof (*tgt) + sizeof (tgt->list[0]) * mapnum);
   tgt->list_count = mapnum;
   tgt->refcount = 1;
   tgt->device_descr = devicep;
+  tgt->mem_map = mm;
 
   if (mapnum == 0)
     return tgt;
@@ -224,40 +168,43 @@ gomp_map_vars (struct gomp_device_descr *devicep, size_t mapnum,
       tgt_size = mapnum * sizeof (void *);
     }
 
-  gomp_mutex_lock (&devicep->dev_env_lock);
+  gomp_mutex_lock (&mm->lock);
+
   for (i = 0; i < mapnum; i++)
     {
+      int kind = get_kind (is_openacc, kinds, i);
       if (hostaddrs[i] == NULL)
 	{
 	  tgt->list[i] = NULL;
 	  continue;
 	}
       cur_node.host_start = (uintptr_t) hostaddrs[i];
-      if ((kinds[i] & 7) != 4)
+      if (!GOMP_MAP_POINTER_P (kind & typemask))
 	cur_node.host_end = cur_node.host_start + sizes[i];
       else
 	cur_node.host_end = cur_node.host_start + sizeof (void *);
-      splay_tree_key n = splay_tree_lookup (&devicep->dev_splay_tree,
-					    &cur_node);
+      splay_tree_key n = splay_tree_lookup (&mm->splay_tree, &cur_node);
       if (n)
 	{
 	  tgt->list[i] = n;
-	  gomp_map_vars_existing (n, &cur_node, kinds[i]);
+	  gomp_map_vars_existing (n, &cur_node, kind & typemask);
 	}
       else
 	{
-	  size_t align = (size_t) 1 << (kinds[i] >> 3);
 	  tgt->list[i] = NULL;
+
+	  size_t align = (size_t) 1 << (kind >> rshift);
 	  not_found_cnt++;
 	  if (tgt_align < align)
 	    tgt_align = align;
 	  tgt_size = (tgt_size + align - 1) & ~(align - 1);
 	  tgt_size += cur_node.host_end - cur_node.host_start;
-	  if ((kinds[i] & 7) == 5)
+	  if ((kind & typemask) == GOMP_MAP_TO_PSET)
 	    {
 	      size_t j;
 	      for (j = i + 1; j < mapnum; j++)
-		if ((kinds[j] & 7) != 4)
+		if (!GOMP_MAP_POINTER_P (get_kind (is_openacc, kinds, j)
+					 & typemask))
 		  break;
 		else if ((uintptr_t) hostaddrs[j] < cur_node.host_start
 			 || ((uintptr_t) hostaddrs[j] + sizeof (void *)
@@ -272,7 +219,15 @@ gomp_map_vars (struct gomp_device_descr *devicep, size_t mapnum,
 	}
     }
 
-  if (not_found_cnt || is_target)
+  if (devaddrs)
+    {
+      if (mapnum != 1)
+        gomp_fatal ("unexpected aggregation");
+      tgt->to_free = devaddrs[0];
+      tgt->tgt_start = (uintptr_t) tgt->to_free;
+      tgt->tgt_end = tgt->tgt_start + sizes[0];
+    }
+  else if (not_found_cnt || is_target)
     {
       /* Allocate tgt_align aligned tgt_size block of memory.  */
       /* FIXME: Perhaps change interface to allocate properly aligned
@@ -304,44 +259,47 @@ gomp_map_vars (struct gomp_device_descr *devicep, size_t mapnum,
       for (i = 0; i < mapnum; i++)
 	if (tgt->list[i] == NULL)
 	  {
+	    int kind = get_kind (is_openacc, kinds, i);
 	    if (hostaddrs[i] == NULL)
 	      continue;
 	    splay_tree_key k = &array->key;
 	    k->host_start = (uintptr_t) hostaddrs[i];
-	    if ((kinds[i] & 7) != 4)
+	    if (!GOMP_MAP_POINTER_P (kind & typemask))
 	      k->host_end = k->host_start + sizes[i];
 	    else
 	      k->host_end = k->host_start + sizeof (void *);
-	    splay_tree_key n
-	      = splay_tree_lookup (&devicep->dev_splay_tree, k);
+	    splay_tree_key n = splay_tree_lookup (&mm->splay_tree, k);
 	    if (n)
 	      {
 		tgt->list[i] = n;
-		gomp_map_vars_existing (n, k, kinds[i]);
+		gomp_map_vars_existing (n, k, kind & typemask);
 	      }
 	    else
 	      {
-		size_t align = (size_t) 1 << (kinds[i] >> 3);
+		size_t align = (size_t) 1 << (kind >> rshift);
 		tgt->list[i] = k;
 		tgt_size = (tgt_size + align - 1) & ~(align - 1);
 		k->tgt = tgt;
 		k->tgt_offset = tgt_size;
 		tgt_size += k->host_end - k->host_start;
-		k->copy_from = false;
-		if ((kinds[i] & 7) == 2 || (kinds[i] & 7) == 3)
-		  k->copy_from = true;
+		k->copy_from = GOMP_MAP_COPY_FROM_P (kind & typemask);
 		k->refcount = 1;
+		k->async_refcount = 0;
 		tgt->refcount++;
 		array->left = NULL;
 		array->right = NULL;
-		splay_tree_insert (&devicep->dev_splay_tree, array);
-		switch (kinds[i] & 7)
+		splay_tree_insert (&mm->splay_tree, array);
+		switch (kind & typemask)
 		  {
-		  case 0: /* ALLOC */
-		  case 2: /* FROM */
+		  case GOMP_MAP_ALLOC:
+		  case GOMP_MAP_FROM:
+		  case GOMP_MAP_FORCE_ALLOC:
+		  case GOMP_MAP_FORCE_FROM:
 		    break;
-		  case 1: /* TO */
-		  case 3: /* TOFROM */
+		  case GOMP_MAP_TO:
+		  case GOMP_MAP_TOFROM:
+		  case GOMP_MAP_FORCE_TO:
+		  case GOMP_MAP_FORCE_TOFROM:
 		    /* FIXME: Perhaps add some smarts, like if copying
 		       several adjacent fields from host to target, use some
 		       host buffer to avoid sending each var individually.  */
@@ -351,12 +309,13 @@ gomp_map_vars (struct gomp_device_descr *devicep, size_t mapnum,
 					    (void *) k->host_start,
 					    k->host_end - k->host_start);
 		    break;
-		  case 4: /* POINTER */
+		  case GOMP_MAP_POINTER:
 		    cur_node.host_start
 		      = (uintptr_t) *(void **) k->host_start;
 		    if (cur_node.host_start == (uintptr_t) NULL)
 		      {
 			cur_node.tgt_offset = (uintptr_t) NULL;
+			/* FIXME: see above FIXME comment.  */
 			devicep->host2dev_func (devicep->target_id,
 						(void *) (tgt->tgt_start
 							  + k->tgt_offset),
@@ -367,19 +326,16 @@ gomp_map_vars (struct gomp_device_descr *devicep, size_t mapnum,
 		    /* Add bias to the pointer value.  */
 		    cur_node.host_start += sizes[i];
 		    cur_node.host_end = cur_node.host_start + 1;
-		    n = splay_tree_lookup (&devicep->dev_splay_tree,
-					   &cur_node);
+		    n = splay_tree_lookup (&mm->splay_tree, &cur_node);
 		    if (n == NULL)
 		      {
 			/* Could be possibly zero size array section.  */
 			cur_node.host_end--;
-			n = splay_tree_lookup (&devicep->dev_splay_tree,
-					       &cur_node);
+			n = splay_tree_lookup (&mm->splay_tree, &cur_node);
 			if (n == NULL)
 			  {
 			    cur_node.host_start--;
-			    n = splay_tree_lookup (&devicep->dev_splay_tree,
-						   &cur_node);
+			    n = splay_tree_lookup (&mm->splay_tree, &cur_node);
 			    cur_node.host_start++;
 			  }
 		      }
@@ -393,20 +349,24 @@ gomp_map_vars (struct gomp_device_descr *devicep, size_t mapnum,
 		       array section.  Now subtract bias to get what we want
 		       to initialize the pointer with.  */
 		    cur_node.tgt_offset -= sizes[i];
+		    /* FIXME: see above FIXME comment.  */
 		    devicep->host2dev_func (devicep->target_id,
 					    (void *) (tgt->tgt_start
 						      + k->tgt_offset),
 					    (void *) &cur_node.tgt_offset,
 					    sizeof (void *));
 		    break;
-		  case 5: /* TO_PSET */
+		  case GOMP_MAP_TO_PSET:
+		    /* FIXME: see above FIXME comment.  */
 		    devicep->host2dev_func (devicep->target_id,
 					    (void *) (tgt->tgt_start
 						      + k->tgt_offset),
 					    (void *) k->host_start,
 					    k->host_end - k->host_start);
+
 		    for (j = i + 1; j < mapnum; j++)
-		      if ((kinds[j] & 7) != 4)
+		      if (!GOMP_MAP_POINTER_P (get_kind (is_openacc, kinds, j)
+					       & typemask))
 			break;
 		      else if ((uintptr_t) hostaddrs[j] < k->host_start
 			       || ((uintptr_t) hostaddrs[j] + sizeof (void *)
@@ -421,6 +381,7 @@ gomp_map_vars (struct gomp_device_descr *devicep, size_t mapnum,
 			  if (cur_node.host_start == (uintptr_t) NULL)
 			    {
 			      cur_node.tgt_offset = (uintptr_t) NULL;
+			      /* FIXME: see above FIXME comment.  */
 			      devicep->host2dev_func (devicep->target_id,
 				 (void *) (tgt->tgt_start + k->tgt_offset
 					   + ((uintptr_t) hostaddrs[j]
@@ -433,19 +394,18 @@ gomp_map_vars (struct gomp_device_descr *devicep, size_t mapnum,
 			  /* Add bias to the pointer value.  */
 			  cur_node.host_start += sizes[j];
 			  cur_node.host_end = cur_node.host_start + 1;
-			  n = splay_tree_lookup (&devicep->dev_splay_tree,
-						 &cur_node);
+			  n = splay_tree_lookup (&mm->splay_tree, &cur_node);
 			  if (n == NULL)
 			    {
 			      /* Could be possibly zero size array section.  */
 			      cur_node.host_end--;
-			      n = splay_tree_lookup (&devicep->dev_splay_tree,
+			      n = splay_tree_lookup (&mm->splay_tree,
 						     &cur_node);
 			      if (n == NULL)
 				{
 				  cur_node.host_start--;
-				  n = splay_tree_lookup
-					(&devicep->dev_splay_tree, &cur_node);
+				  n = splay_tree_lookup (&mm->splay_tree,
+							 &cur_node);
 				  cur_node.host_start++;
 				}
 			    }
@@ -460,6 +420,7 @@ gomp_map_vars (struct gomp_device_descr *devicep, size_t mapnum,
 			     array section.  Now subtract bias to get what we
 			     want to initialize the pointer with.  */
 			  cur_node.tgt_offset -= sizes[j];
+			  /* FIXME: see above FIXME comment.  */
 			  devicep->host2dev_func (devicep->target_id,
 			     (void *) (tgt->tgt_start + k->tgt_offset
 				       + ((uintptr_t) hostaddrs[j]
@@ -468,12 +429,35 @@ gomp_map_vars (struct gomp_device_descr *devicep, size_t mapnum,
 			     sizeof (void *));
 			  i++;
 			}
-		      break;
+		    break;
+		  case GOMP_MAP_FORCE_PRESENT:
+		    {
+		      /* We already looked up the memory region above and it
+			 was missing.  */
+		      size_t size = k->host_end - k->host_start;
+		      gomp_fatal ("present clause: !acc_is_present (%p, "
+				  "%zd (0x%zx))", (void *) k->host_start,
+				  size, size);
+		    }
+		    break;
+		  case GOMP_MAP_FORCE_DEVICEPTR:
+		    assert (k->host_end - k->host_start == sizeof (void *));
+
+		    devicep->host2dev_func (devicep->target_id,
+					    (void *) (tgt->tgt_start
+						      + k->tgt_offset),
+					    (void *) k->host_start,
+					    sizeof (void *));
+		    break;
+		  default:
+		    gomp_fatal ("%s: unhandled kind 0x%.2x", __FUNCTION__,
+				kind);
 		  }
 		array++;
 	      }
 	  }
     }
+
   if (is_target)
     {
       for (i = 0; i < mapnum; i++)
@@ -483,6 +467,7 @@ gomp_map_vars (struct gomp_device_descr *devicep, size_t mapnum,
 	  else
 	    cur_node.tgt_offset = tgt->list[i]->tgt->tgt_start
 				  + tgt->list[i]->tgt_offset;
+	  /* FIXME: see above FIXME comment.  */
 	  devicep->host2dev_func (devicep->target_id,
 				  (void *) (tgt->tgt_start
 					    + i * sizeof (void *)),
@@ -491,7 +476,7 @@ gomp_map_vars (struct gomp_device_descr *devicep, size_t mapnum,
 	}
     }
 
-  gomp_mutex_unlock (&devicep->dev_env_lock);
+  gomp_mutex_unlock (&mm->lock);
   return tgt;
 }
 
@@ -506,10 +491,50 @@ gomp_unmap_tgt (struct target_mem_desc *tgt)
   free (tgt);
 }
 
-static void
-gomp_unmap_vars (struct target_mem_desc *tgt)
+/* Decrease the refcount for a set of mapped variables, and queue asychronous
+   copies from the device back to the host after any work that has been issued.
+   Because the regions are still "live", increment an asynchronous reference
+   count to indicate that they should not be unmapped from host-side data
+   structures until the asynchronous copy has completed.  */
+
+attribute_hidden void
+gomp_copy_from_async (struct target_mem_desc *tgt)
+{
+  struct gomp_device_descr *devicep = tgt->device_descr;
+  struct gomp_memory_mapping *mm = tgt->mem_map;
+  size_t i;
+
+  gomp_mutex_lock (&mm->lock);
+
+  for (i = 0; i < tgt->list_count; i++)
+    if (tgt->list[i] == NULL)
+      ;
+    else if (tgt->list[i]->refcount > 1)
+      {
+	tgt->list[i]->refcount--;
+	tgt->list[i]->async_refcount++;
+      }
+    else
+      {
+	splay_tree_key k = tgt->list[i];
+	if (k->copy_from)
+	  devicep->dev2host_func (devicep->target_id, (void *) k->host_start,
+				  (void *) (k->tgt->tgt_start + k->tgt_offset),
+				  k->host_end - k->host_start);
+      }
+
+  gomp_mutex_unlock (&mm->lock);
+}
+
+/* Unmap variables described by TGT.  If DO_COPYFROM is true, copy relevant
+   variables back from device to host: if it is false, it is assumed that this
+   has been done already, i.e. by gomp_copy_from_async above.  */
+
+attribute_hidden void
+gomp_unmap_vars (struct target_mem_desc *tgt, bool do_copyfrom)
 {
   struct gomp_device_descr *devicep = tgt->device_descr;
+  struct gomp_memory_mapping *mm = tgt->mem_map;
 
   if (tgt->list_count == 0)
     {
@@ -517,21 +542,24 @@ gomp_unmap_vars (struct target_mem_desc *tgt)
       return;
     }
 
+  gomp_mutex_lock (&mm->lock);
+
   size_t i;
-  gomp_mutex_lock (&devicep->dev_env_lock);
   for (i = 0; i < tgt->list_count; i++)
     if (tgt->list[i] == NULL)
       ;
     else if (tgt->list[i]->refcount > 1)
       tgt->list[i]->refcount--;
+    else if (tgt->list[i]->async_refcount > 0)
+      tgt->list[i]->async_refcount--;
     else
       {
 	splay_tree_key k = tgt->list[i];
-	if (k->copy_from)
+	if (k->copy_from && do_copyfrom)
 	  devicep->dev2host_func (devicep->target_id, (void *) k->host_start,
 				  (void *) (k->tgt->tgt_start + k->tgt_offset),
 				  k->host_end - k->host_start);
-	splay_tree_remove (&devicep->dev_splay_tree, k);
+	splay_tree_remove (&mm->splay_tree, k);
 	if (k->tgt->refcount > 1)
 	  k->tgt->refcount--;
 	else
@@ -542,15 +570,18 @@ gomp_unmap_vars (struct target_mem_desc *tgt)
     tgt->refcount--;
   else
     gomp_unmap_tgt (tgt);
-  gomp_mutex_unlock (&devicep->dev_env_lock);
+
+  gomp_mutex_unlock (&mm->lock);
 }
 
 static void
-gomp_update (struct gomp_device_descr *devicep, size_t mapnum,
-	     void **hostaddrs, size_t *sizes, unsigned char *kinds)
+gomp_update (struct gomp_device_descr *devicep, struct gomp_memory_mapping *mm,
+	     size_t mapnum, void **hostaddrs, size_t *sizes, void *kinds,
+	     bool is_openacc)
 {
   size_t i;
   struct splay_tree_key_s cur_node;
+  const int typemask = is_openacc ? 0xff : 0x7;
 
   if (!devicep)
     return;
@@ -558,16 +589,17 @@ gomp_update (struct gomp_device_descr *devicep, size_t mapnum,
   if (mapnum == 0)
     return;
 
-  gomp_mutex_lock (&devicep->dev_env_lock);
+  gomp_mutex_lock (&mm->lock);
   for (i = 0; i < mapnum; i++)
     if (sizes[i])
       {
 	cur_node.host_start = (uintptr_t) hostaddrs[i];
 	cur_node.host_end = cur_node.host_start + sizes[i];
-	splay_tree_key n = splay_tree_lookup (&devicep->dev_splay_tree,
+	splay_tree_key n = splay_tree_lookup (&mm->splay_tree,
 					      &cur_node);
 	if (n)
 	  {
+	    int kind = get_kind (is_openacc, kinds, i);
 	    if (n->host_start > cur_node.host_start
 		|| n->host_end < cur_node.host_end)
 	      gomp_fatal ("Trying to update [%p..%p) object when"
@@ -576,7 +608,7 @@ gomp_update (struct gomp_device_descr *devicep, size_t mapnum,
 			  (void *) cur_node.host_end,
 			  (void *) n->host_start,
 			  (void *) n->host_end);
-	    if ((kinds[i] & 7) == 1)
+	    if (GOMP_MAP_COPY_TO_P (kind & typemask))
 	      devicep->host2dev_func (devicep->target_id,
 				      (void *) (n->tgt->tgt_start
 						+ n->tgt_offset
@@ -584,7 +616,7 @@ gomp_update (struct gomp_device_descr *devicep, size_t mapnum,
 						- n->host_start),
 				      (void *) cur_node.host_start,
 				      cur_node.host_end - cur_node.host_start);
-	    else if ((kinds[i] & 7) == 2)
+	    if (GOMP_MAP_COPY_FROM_P (kind & typemask))
 	      devicep->dev2host_func (devicep->target_id,
 				      (void *) cur_node.host_start,
 				      (void *) (n->tgt->tgt_start
@@ -598,7 +630,7 @@ gomp_update (struct gomp_device_descr *devicep, size_t mapnum,
 		      (void *) cur_node.host_start,
 		      (void *) cur_node.host_end);
       }
-  gomp_mutex_unlock (&devicep->dev_env_lock);
+  gomp_mutex_unlock (&mm->lock);
 }
 
 /* This function should be called from every offload image.
@@ -620,13 +652,23 @@ GOMP_offload_register (void *host_table, enum offload_target_type target_type,
   num_offload_images++;
 }
 
-/* This function initializes the target device, specified by DEVICEP.  */
+/* This function initializes the target device, specified by DEVICEP.  DEVICEP
+   must be locked on entry, and remains locked on return.  */
 
-static void
+attribute_hidden void
 gomp_init_device (struct gomp_device_descr *devicep)
 {
   devicep->init_device_func (devicep->target_id);
+  devicep->is_initialized = true;
+}
 
+/* Initialize address mapping tables.  MM must be locked on entry, and remains
+   locked on return.  */
+
+attribute_hidden void
+gomp_init_tables (struct gomp_device_descr *devicep,
+		  struct gomp_memory_mapping *mm)
+{
   /* Get address mapping table for device.  */
   struct mapping_table *table = NULL;
   int num_entries = devicep->get_table_func (devicep->target_id, &table);
@@ -653,30 +695,63 @@ gomp_init_device (struct gomp_device_descr *devicep)
       k->tgt = tgt;
       node->left = NULL;
       node->right = NULL;
-      splay_tree_insert (&devicep->dev_splay_tree, node);
+      splay_tree_insert (&mm->splay_tree, node);
     }
 
   free (table);
-  devicep->is_initialized = true;
+  mm->is_initialized = true;
+}
+
+/* Free address mapping tables.  MM must be locked on entry, and remains locked
+   on return.  */
+
+attribute_hidden void
+gomp_free_memmap (struct gomp_memory_mapping *mm)
+{
+  while (mm->splay_tree.root)
+    {
+      struct target_mem_desc *tgt = mm->splay_tree.root->key.tgt;
+
+      splay_tree_remove (&mm->splay_tree, &mm->splay_tree.root->key);
+      free (tgt->array);
+      free (tgt);
+    }
+
+  mm->is_initialized = false;
+}
+
+/* This function de-initializes the target device, specified by DEVICEP.
+   DEVICEP must be locked on entry, and remains locked on return.  */
+
+attribute_hidden void
+gomp_fini_device (struct gomp_device_descr *devicep)
+{
+  if (devicep->is_initialized)
+    devicep->fini_device_func (devicep->target_id);
+
+  devicep->is_initialized = false;
 }
 
 /* Called when encountering a target directive.  If DEVICE
-   is -1, it means use device-var ICV.  If it is -2 (or any other value
-   larger than last available hw device, use host fallback.
-   FN is address of host code, OPENMP_TARGET contains value of the
-   __OPENMP_TARGET__ symbol in the shared library or binary that invokes
+   is GOMP_DEVICE_ICV, it means use device-var ICV.  If it is
+   GOMP_DEVICE_HOST_FALLBACK (or any value
+   larger than last available hw device), use host fallback.
+   FN is address of host code, OFFLOAD_TABLE contains value of the
+   __OFFLOAD_TABLE__ symbol in the shared library or binary that invokes
    GOMP_target.  HOSTADDRS, SIZES and KINDS are arrays
    with MAPNUM entries, with addresses of the host objects,
    sizes of the host objects (resp. for pointer kind pointer bias
    and assumed sizeof (void *) size) and kinds.  */
 
 void
-GOMP_target (int device, void (*fn) (void *), const void *openmp_target,
+GOMP_target (int device, void (*fn) (void *), const void *offload_table,
 	     size_t mapnum, void **hostaddrs, size_t *sizes,
 	     unsigned char *kinds)
 {
   struct gomp_device_descr *devicep = resolve_device (device);
-  if (devicep == NULL)
+
+  if (devicep == NULL
+      || !(devicep->capabilities & GOMP_OFFLOAD_CAP_OPENMP_400))
     {
       /* Host fallback.  */
       struct gomp_thread old_thr, *thr = gomp_thread ();
@@ -693,20 +768,38 @@ GOMP_target (int device, void (*fn) (void *), const void *openmp_target,
       return;
     }
 
-  gomp_mutex_lock (&devicep->dev_env_lock);
+  gomp_mutex_lock (&devicep->lock);
   if (!devicep->is_initialized)
     gomp_init_device (devicep);
+  gomp_mutex_unlock (&devicep->lock);
+
+  void *fn_addr;
+
+  if (devicep->capabilities & GOMP_OFFLOAD_CAP_NATIVE_EXEC)
+    fn_addr = (void *) fn;
+  else
+    {
+      struct gomp_memory_mapping *mm = &devicep->mem_map;
+      gomp_mutex_lock (&mm->lock);
+
+      if (!mm->is_initialized)
+	gomp_init_tables (devicep, mm);
+
+      struct splay_tree_key_s k;
+      k.host_start = (uintptr_t) fn;
+      k.host_end = k.host_start + 1;
+      splay_tree_key tgt_fn = splay_tree_lookup (&mm->splay_tree, &k);
+      if (tgt_fn == NULL)
+	gomp_fatal ("Target function wasn't mapped");
 
-  struct splay_tree_key_s k;
-  k.host_start = (uintptr_t) fn;
-  k.host_end = k.host_start + 1;
-  splay_tree_key tgt_fn = splay_tree_lookup (&devicep->dev_splay_tree, &k);
-  if (tgt_fn == NULL)
-    gomp_fatal ("Target function wasn't mapped");
-  gomp_mutex_unlock (&devicep->dev_env_lock);
+      gomp_mutex_unlock (&mm->lock);
+
+      fn_addr = (void *) tgt_fn->tgt->tgt_start;
+    }
 
   struct target_mem_desc *tgt_vars
-    = gomp_map_vars (devicep, mapnum, hostaddrs, sizes, kinds, true);
+    = gomp_map_vars (devicep, mapnum, hostaddrs, NULL, sizes, kinds, false,
+		     true);
   struct gomp_thread old_thr, *thr = gomp_thread ();
   old_thr = *thr;
   memset (thr, '\0', sizeof (*thr));
@@ -715,19 +808,20 @@ GOMP_target (int device, void (*fn) (void *), const void *openmp_target,
       thr->place = old_thr.place;
       thr->ts.place_partition_len = gomp_places_list_len;
     }
-  devicep->run_func (devicep->target_id, (void *) tgt_fn->tgt->tgt_start,
-		     (void *) tgt_vars->tgt_start);
+  devicep->run_func (devicep->target_id, fn_addr, (void *) tgt_vars->tgt_start);
   gomp_free_thread (thr);
   *thr = old_thr;
-  gomp_unmap_vars (tgt_vars);
+  gomp_unmap_vars (tgt_vars, true);
 }
 
 void
-GOMP_target_data (int device, const void *openmp_target, size_t mapnum,
+GOMP_target_data (int device, const void *offload_table, size_t mapnum,
 		  void **hostaddrs, size_t *sizes, unsigned char *kinds)
 {
   struct gomp_device_descr *devicep = resolve_device (device);
-  if (devicep == NULL)
+
+  if (devicep == NULL
+      || !(devicep->capabilities & GOMP_OFFLOAD_CAP_OPENMP_400))
     {
       /* Host fallback.  */
       struct gomp_task_icv *icv = gomp_icv (false);
@@ -738,20 +832,27 @@ GOMP_target_data (int device, const void *openmp_target, size_t mapnum,
 	     new #pragma omp target data, otherwise GOMP_target_end_data
 	     would get out of sync.  */
 	  struct target_mem_desc *tgt
-	    = gomp_map_vars (NULL, 0, NULL, NULL, NULL, false);
+	    = gomp_map_vars (NULL, 0, NULL, NULL, NULL, NULL, false, false);
 	  tgt->prev = icv->target_data;
 	  icv->target_data = tgt;
 	}
       return;
     }
 
-  gomp_mutex_lock (&devicep->dev_env_lock);
+  gomp_mutex_lock (&devicep->lock);
   if (!devicep->is_initialized)
     gomp_init_device (devicep);
-  gomp_mutex_unlock (&devicep->dev_env_lock);
+  gomp_mutex_unlock (&devicep->lock);
+
+  struct gomp_memory_mapping *mm = &devicep->mem_map;
+  gomp_mutex_lock (&mm->lock);
+  if (!mm->is_initialized)
+    gomp_init_tables (devicep, mm);
+  gomp_mutex_unlock (&mm->lock);
 
   struct target_mem_desc *tgt
-    = gomp_map_vars (devicep, mapnum, hostaddrs, sizes, kinds, false);
+    = gomp_map_vars (devicep, mapnum, hostaddrs, NULL, sizes, kinds, false,
+		     false);
   struct gomp_task_icv *icv = gomp_icv (true);
   tgt->prev = icv->target_data;
   icv->target_data = tgt;
@@ -765,24 +866,32 @@ GOMP_target_end_data (void)
     {
       struct target_mem_desc *tgt = icv->target_data;
       icv->target_data = tgt->prev;
-      gomp_unmap_vars (tgt);
+      gomp_unmap_vars (tgt, true);
     }
 }
 
 void
-GOMP_target_update (int device, const void *openmp_target, size_t mapnum,
+GOMP_target_update (int device, const void *offload_table, size_t mapnum,
 		    void **hostaddrs, size_t *sizes, unsigned char *kinds)
 {
   struct gomp_device_descr *devicep = resolve_device (device);
-  if (devicep == NULL)
+
+  if (devicep == NULL
+      || !(devicep->capabilities & GOMP_OFFLOAD_CAP_OPENMP_400))
     return;
 
-  gomp_mutex_lock (&devicep->dev_env_lock);
+  gomp_mutex_lock (&devicep->lock);
   if (!devicep->is_initialized)
     gomp_init_device (devicep);
-  gomp_mutex_unlock (&devicep->dev_env_lock);
+  gomp_mutex_unlock (&devicep->lock);
+
+  struct gomp_memory_mapping *mm = &devicep->mem_map;
+  gomp_mutex_lock (&mm->lock);
+  if (!mm->is_initialized)
+    gomp_init_tables (devicep, mm);
+  gomp_mutex_unlock (&mm->lock);
 
-  gomp_update (devicep, mapnum, hostaddrs, sizes, kinds);
+  gomp_update (devicep, mm, mapnum, hostaddrs, sizes, kinds, false);
 }
 
 void
@@ -808,54 +917,137 @@ static bool
 gomp_load_plugin_for_device (struct gomp_device_descr *device,
 			     const char *plugin_name)
 {
+  char *err = NULL, *last_missing = NULL;
+  int optional_present, optional_total;
+
+  /* Clear any existing error.  */
+  dlerror ();
+
   void *plugin_handle = dlopen (plugin_name, RTLD_LAZY);
   if (!plugin_handle)
-    return false;
+    {
+      err = dlerror ();
+      goto out;
+    }
 
   /* Check if all required functions are available in the plugin and store
      their handlers.  */
-#define DLSYM(f)						    \
-  do								    \
-    {								    \
-      device->f##_func = dlsym (plugin_handle, "GOMP_OFFLOAD_"#f);  \
-      if (!device->f##_func)					    \
-	return false;						    \
-    }								    \
+#define DLSYM(f)							\
+  do									\
+    {									\
+      device->f##_func = dlsym (plugin_handle, "GOMP_OFFLOAD_" #f);	\
+      err = dlerror ();							\
+      if (err != NULL)							\
+	goto out;							\
+    }									\
   while (0)
+  /* Similar, but missing functions are not an error.  */
+#define DLSYM_OPT(f, n)						\
+  do									\
+    {									\
+      char *tmp_err;							\
+      device->f##_func = dlsym (plugin_handle, "GOMP_OFFLOAD_" #n);	\
+      tmp_err = dlerror ();						\
+      if (tmp_err == NULL)						\
+        optional_present++;						\
+      else								\
+        last_missing = #n;						\
+      optional_total++;							\
+    }									\
+  while (0)
+
+  DLSYM (get_name);
+  DLSYM (get_caps);
   DLSYM (get_type);
   DLSYM (get_num_devices);
   DLSYM (register_image);
   DLSYM (init_device);
+  DLSYM (fini_device);
   DLSYM (get_table);
   DLSYM (alloc);
   DLSYM (free);
   DLSYM (dev2host);
   DLSYM (host2dev);
-  DLSYM (run);
+  device->capabilities = device->get_caps_func ();
+  if (device->capabilities & GOMP_OFFLOAD_CAP_OPENMP_400)
+    DLSYM (run);
+  if (device->capabilities & GOMP_OFFLOAD_CAP_OPENACC_200)
+    {
+      optional_present = optional_total = 0;
+      DLSYM_OPT (openacc.exec, openacc_parallel);
+      DLSYM_OPT (openacc.open_device, openacc_open_device);
+      DLSYM_OPT (openacc.close_device, openacc_close_device);
+      DLSYM_OPT (openacc.get_device_num, openacc_get_device_num);
+      DLSYM_OPT (openacc.set_device_num, openacc_set_device_num);
+      DLSYM_OPT (openacc.register_async_cleanup,
+		 openacc_register_async_cleanup);
+      DLSYM_OPT (openacc.async_test, openacc_async_test);
+      DLSYM_OPT (openacc.async_test_all, openacc_async_test_all);
+      DLSYM_OPT (openacc.async_wait, openacc_async_wait);
+      DLSYM_OPT (openacc.async_wait_async, openacc_async_wait_async);
+      DLSYM_OPT (openacc.async_wait_all, openacc_async_wait_all);
+      DLSYM_OPT (openacc.async_wait_all_async, openacc_async_wait_all_async);
+      DLSYM_OPT (openacc.async_set_async, openacc_async_set_async);
+      DLSYM_OPT (openacc.create_thread_data, openacc_create_thread_data);
+      DLSYM_OPT (openacc.destroy_thread_data, openacc_destroy_thread_data);
+      /* Require all the OpenACC handlers if we have
+	 GOMP_OFFLOAD_CAP_OPENACC_200.  */
+      if (optional_present != optional_total)
+	{
+	  err = "plugin missing OpenACC handler function";
+	  goto out;
+	}
+      optional_present = optional_total = 0;
+      DLSYM_OPT (openacc.cuda.get_current_device,
+		 openacc_get_current_cuda_device);
+      DLSYM_OPT (openacc.cuda.get_current_context,
+		 openacc_get_current_cuda_context);
+      DLSYM_OPT (openacc.cuda.get_stream, openacc_get_cuda_stream);
+      DLSYM_OPT (openacc.cuda.set_stream, openacc_set_cuda_stream);
+      /* Make sure all the CUDA functions are there if any of them are.  */
+      if (optional_present && optional_present != optional_total)
+	{
+	  err = "plugin missing OpenACC CUDA handler function";
+	  goto out;
+	}
+    }
 #undef DLSYM
+#undef DLSYM_OPT
 
-  return true;
+ out:
+  if (err != NULL)
+    {
+      gomp_error ("while loading %s: %s", plugin_name, err);
+      if (last_missing)
+        gomp_error ("missing function was %s", last_missing);
+      if (plugin_handle)
+	dlclose (plugin_handle);
+    }
+  return err == NULL;
 }
 
-/* This function finds OFFLOAD_IMAGES corresponding to DEVICE type, and
-   registers them in the plugin.  */
+/* This function adds a compatible offload image IMAGE to an accelerator device
+   DEVICE.  DEVICE must be locked on entry, and remains locked on return.  */
 
 static void
-gomp_register_images_for_device (struct gomp_device_descr *device)
+gomp_register_image_for_device (struct gomp_device_descr *device,
+				struct offload_image_descr *image)
 {
-  int i;
-  for (i = 0; i < num_offload_images; i++)
+  if (!device->offload_regions_registered
+      && (device->type == image->type
+	  || device->type == OFFLOAD_TARGET_TYPE_HOST))
     {
-      struct offload_image_descr *image = &offload_images[i];
-      if (image->type == device->type)
-	device->register_image_func (image->host_table, image->target_data);
+      device->register_image_func (image->host_table, image->target_data);
+      device->offload_regions_registered = true;
     }
 }
 
 /* This function initializes the runtime needed for offloading.
-   It parses the list of offload targets and tries to load the plugins for these
-   targets.  Result of the function is properly initialized variable NUM_DEVICES
-   and array DEVICES, containing descriptors for corresponding devices.  */
+   It parses the list of offload targets and tries to load the plugins for
+   these targets.  On return, the variables NUM_DEVICES and NUM_DEVICES_OPENMP
+   will be set, and the array DEVICES initialized, containing descriptors for
+   corresponding devices, first the GOMP_OFFLOAD_CAP_OPENMP_400 ones, follows
+   by the others.  */
 
 static void
 gomp_target_init (void)
@@ -894,6 +1086,8 @@ gomp_target_init (void)
 	    new_num_devices = current_device.get_num_devices_func ();
 	    if (new_num_devices >= 1)
 	      {
+		/* Augment DEVICES and NUM_DEVICES.  */
+
 		devices = realloc (devices, (num_devices + new_num_devices)
 				   * sizeof (struct gomp_device_descr));
 		if (!devices)
@@ -903,16 +1097,21 @@ gomp_target_init (void)
 		    break;
 		  }
 
+		current_device.name = current_device.get_name_func ();
+		/* current_device.capabilities has already been set.  */
 		current_device.type = current_device.get_type_func ();
+		current_device.mem_map.is_initialized = false;
+		current_device.mem_map.splay_tree.root = NULL;
 		current_device.is_initialized = false;
-		current_device.dev_splay_tree.root = NULL;
-		gomp_register_images_for_device (&current_device);
+		current_device.offload_regions_registered = false;
+		current_device.openacc.data_environ = NULL;
+		current_device.openacc.target_data = NULL;
 		for (i = 0; i < new_num_devices; i++)
 		  {
-		    current_device.id = num_devices + 1;
 		    current_device.target_id = i;
 		    devices[num_devices] = current_device;
-		    gomp_mutex_init (&devices[num_devices].dev_env_lock);
+		    gomp_mutex_init (&devices[num_devices].mem_map.lock);
+		    gomp_mutex_init (&devices[num_devices].lock);
 		    num_devices++;
 		  }
 	      }
@@ -923,6 +1122,41 @@ gomp_target_init (void)
       }
     while (next);
 
+  /* In DEVICES, sort the GOMP_OFFLOAD_CAP_OPENMP_400 ones first, and set
+     NUM_DEVICES_OPENMP.  */
+  struct gomp_device_descr *devices_s
+    = malloc (num_devices * sizeof (struct gomp_device_descr));
+  if (!devices_s)
+    {
+      num_devices = 0;
+      free (devices);
+      devices = NULL;
+    }
+  num_devices_openmp = 0;
+  for (i = 0; i < num_devices; i++)
+    if (devices[i].capabilities & GOMP_OFFLOAD_CAP_OPENMP_400)
+      devices_s[num_devices_openmp++] = devices[i];
+  int num_devices_after_openmp = num_devices_openmp;
+  for (i = 0; i < num_devices; i++)
+    if (!(devices[i].capabilities & GOMP_OFFLOAD_CAP_OPENMP_400))
+      devices_s[num_devices_after_openmp++] = devices[i];
+  free (devices);
+  devices = devices_s;
+
+  for (i = 0; i < num_devices; i++)
+    {
+      int j;
+
+      for (j = 0; j < num_offload_images; j++)
+	gomp_register_image_for_device (&devices[i], &offload_images[j]);
+
+      /* The 'devices' array can be moved (by the realloc call) until we have
+	 found all the plugins, so registering with the OpenACC runtime (which
+	 takes a copy of the pointer argument) must be delayed until now.  */
+      if (devices[i].capabilities & GOMP_OFFLOAD_CAP_OPENACC_200)
+	goacc_register (&devices[i]);
+    }
+
   free (offload_images);
   offload_images = NULL;
   num_offload_images = 0;
diff --git a/libgomp/testsuite/Makefile.am b/libgomp/testsuite/Makefile.am
index 9cc103a1c4d..66a9d947e04 100644
--- a/libgomp/testsuite/Makefile.am
+++ b/libgomp/testsuite/Makefile.am
@@ -12,7 +12,16 @@ _RUNTEST = $(shell if test -f $(top_srcdir)/../dejagnu/runtest; then \
 	     echo $(top_srcdir)/../dejagnu/runtest; else echo runtest; fi)
 RUNTEST = "$(_RUNTEST) $(AM_RUNTESTFLAGS)"
 
-# Used for support non-fallback offloading.
-export OFFLOAD_TARGETS = $(offload_targets)
-export OFFLOAD_ADDITIONAL_OPTIONS = $(offload_additional_options)
-export OFFLOAD_ADDITIONAL_LIB_PATHS = $(offload_additional_lib_paths)
+
+# Instead of directly in ../testsuite/libgomp-test-support.exp.in, the
+# following variables have to be "routed through" this Makefile, for expansion
+# of the several (Makefile) variables used therein.
+libgomp-test-support.exp: libgomp-test-support.pt.exp Makefile
+	cp $< $@.tmp
+	echo >> $@.tmp \
+	  'set offload_additional_options "$(offload_additional_options)"'
+	echo >> $@.tmp \
+	  'set offload_additional_lib_paths "$(offload_additional_lib_paths)"'
+	mv $@.tmp $@
+
+all-local: libgomp-test-support.exp
diff --git a/libgomp/testsuite/Makefile.in b/libgomp/testsuite/Makefile.in
index 2f845f0c7cb..352fc3fec7c 100644
--- a/libgomp/testsuite/Makefile.in
+++ b/libgomp/testsuite/Makefile.in
@@ -35,7 +35,8 @@ build_triplet = @build@
 host_triplet = @host@
 target_triplet = @target@
 subdir = testsuite
-DIST_COMMON = $(srcdir)/Makefile.in $(srcdir)/Makefile.am
+DIST_COMMON = $(srcdir)/Makefile.in $(srcdir)/Makefile.am \
+	$(srcdir)/libgomp-test-support.exp.in
 ACLOCAL_M4 = $(top_srcdir)/aclocal.m4
 am__aclocal_m4_deps = $(top_srcdir)/../config/acx.m4 \
 	$(top_srcdir)/../config/depstand.m4 \
@@ -49,12 +50,13 @@ am__aclocal_m4_deps = $(top_srcdir)/../config/acx.m4 \
 	$(top_srcdir)/../config/tls.m4 $(top_srcdir)/../ltoptions.m4 \
 	$(top_srcdir)/../ltsugar.m4 $(top_srcdir)/../ltversion.m4 \
 	$(top_srcdir)/../lt~obsolete.m4 $(top_srcdir)/acinclude.m4 \
-	$(top_srcdir)/../libtool.m4 $(top_srcdir)/configure.ac
+	$(top_srcdir)/../libtool.m4 $(top_srcdir)/plugin/configfrag.ac \
+	$(top_srcdir)/configure.ac
 am__configure_deps = $(am__aclocal_m4_deps) $(CONFIGURE_DEPENDENCIES) \
 	$(ACLOCAL_M4)
 mkinstalldirs = $(SHELL) $(top_srcdir)/../mkinstalldirs
 CONFIG_HEADER = $(top_builddir)/config.h
-CONFIG_CLEAN_FILES =
+CONFIG_CLEAN_FILES = libgomp-test-support.pt.exp
 CONFIG_CLEAN_VPATH_FILES =
 SOURCES =
 DEJATOOL = $(PACKAGE)
@@ -71,6 +73,8 @@ CCDEPMODE = @CCDEPMODE@
 CFLAGS = @CFLAGS@
 CPP = @CPP@
 CPPFLAGS = @CPPFLAGS@
+CUDA_DRIVER_INCLUDE = @CUDA_DRIVER_INCLUDE@
+CUDA_DRIVER_LIB = @CUDA_DRIVER_LIB@
 CYGPATH_W = @CYGPATH_W@
 DEFS = @DEFS@
 DEPDIR = @DEPDIR@
@@ -129,6 +133,10 @@ PACKAGE_URL = @PACKAGE_URL@
 PACKAGE_VERSION = @PACKAGE_VERSION@
 PATH_SEPARATOR = @PATH_SEPARATOR@
 PERL = @PERL@
+PLUGIN_NVPTX = @PLUGIN_NVPTX@
+PLUGIN_NVPTX_CPPFLAGS = @PLUGIN_NVPTX_CPPFLAGS@
+PLUGIN_NVPTX_LDFLAGS = @PLUGIN_NVPTX_LDFLAGS@
+PLUGIN_NVPTX_LIBS = @PLUGIN_NVPTX_LIBS@
 RANLIB = @RANLIB@
 SECTION_LDFLAGS = @SECTION_LDFLAGS@
 SED = @SED@
@@ -250,6 +258,8 @@ $(top_srcdir)/configure: @MAINTAINER_MODE_TRUE@ $(am__configure_deps)
 $(ACLOCAL_M4): @MAINTAINER_MODE_TRUE@ $(am__aclocal_m4_deps)
 	cd $(top_builddir) && $(MAKE) $(AM_MAKEFLAGS) am--refresh
 $(am__aclocal_m4_deps):
+libgomp-test-support.pt.exp: $(top_builddir)/config.status $(srcdir)/libgomp-test-support.exp.in
+	cd $(top_builddir) && $(SHELL) ./config.status $(subdir)/$@
 
 mostlyclean-libtool:
 	-rm -f *.lo
@@ -303,7 +313,7 @@ distclean-DEJAGNU:
 check-am: all-am
 	$(MAKE) $(AM_MAKEFLAGS) check-DEJAGNU
 check: check-am
-all-am: Makefile
+all-am: Makefile all-local
 installdirs:
 install: install-am
 install-exec: install-exec-am
@@ -398,23 +408,31 @@ uninstall-am:
 
 .MAKE: check-am install-am install-strip
 
-.PHONY: all all-am check check-DEJAGNU check-am clean clean-generic \
-	clean-libtool distclean distclean-DEJAGNU distclean-generic \
-	distclean-libtool dvi dvi-am html html-am info info-am install \
-	install-am install-data install-data-am install-dvi \
-	install-dvi-am install-exec install-exec-am install-html \
-	install-html-am install-info install-info-am install-man \
-	install-pdf install-pdf-am install-ps install-ps-am \
-	install-strip installcheck installcheck-am installdirs \
-	maintainer-clean maintainer-clean-generic mostlyclean \
-	mostlyclean-generic mostlyclean-libtool pdf pdf-am ps ps-am \
-	uninstall uninstall-am
-
-
-# Used for support non-fallback offloading.
-export OFFLOAD_TARGETS = $(offload_targets)
-export OFFLOAD_ADDITIONAL_OPTIONS = $(offload_additional_options)
-export OFFLOAD_ADDITIONAL_LIB_PATHS = $(offload_additional_lib_paths)
+.PHONY: all all-am all-local check check-DEJAGNU check-am clean \
+	clean-generic clean-libtool distclean distclean-DEJAGNU \
+	distclean-generic distclean-libtool dvi dvi-am html html-am \
+	info info-am install install-am install-data install-data-am \
+	install-dvi install-dvi-am install-exec install-exec-am \
+	install-html install-html-am install-info install-info-am \
+	install-man install-pdf install-pdf-am install-ps \
+	install-ps-am install-strip installcheck installcheck-am \
+	installdirs maintainer-clean maintainer-clean-generic \
+	mostlyclean mostlyclean-generic mostlyclean-libtool pdf pdf-am \
+	ps ps-am uninstall uninstall-am
+
+
+# Instead of directly in ../testsuite/libgomp-test-support.exp.in, the
+# following variables have to be "routed through" this Makefile, for expansion
+# of the several (Makefile) variables used therein.
+libgomp-test-support.exp: libgomp-test-support.pt.exp Makefile
+	cp $< $@.tmp
+	echo >> $@.tmp \
+	  'set offload_additional_options "$(offload_additional_options)"'
+	echo >> $@.tmp \
+	  'set offload_additional_lib_paths "$(offload_additional_lib_paths)"'
+	mv $@.tmp $@
+
+all-local: libgomp-test-support.exp
 
 # Tell versions [3.59,3.63) of GNU make to not export all variables.
 # Otherwise a system limit (for SysV at least) may be exceeded.
diff --git a/libgomp/testsuite/lib/libgomp.exp b/libgomp/testsuite/lib/libgomp.exp
index 2d6f822adf9..5a6eec12352 100644
--- a/libgomp/testsuite/lib/libgomp.exp
+++ b/libgomp/testsuite/lib/libgomp.exp
@@ -32,6 +32,29 @@ load_gcc_lib timeout-dg.exp
 load_gcc_lib torture-options.exp
 load_gcc_lib fortran-modules.exp
 
+# Try to load a test support file, built during libgomp configuration.
+load_file libgomp-test-support.exp
+
+# Populate offload_targets_s (offloading targets separated by a space), and
+# offload_targets_s_openacc (the same, but with OpenACC names; OpenACC spells
+# some of them a little differently).
+set offload_targets_s [split $offload_targets ","]
+set offload_targets_s_openacc {}
+foreach offload_target_openacc $offload_targets_s {
+    switch $offload_target_openacc {
+	intelmic {
+	    # Skip; will all FAIL because of missing
+	    # GOMP_OFFLOAD_CAP_OPENACC_200.
+	    continue
+	}
+	nvptx {
+	    set offload_target_openacc "nvidia"
+	}
+    }
+    lappend offload_targets_s_openacc "$offload_target_openacc"
+}
+lappend offload_targets_s_openacc "host"
+
 set dg-do-what-default run
 
 #
@@ -108,13 +131,9 @@ proc libgomp_init { args } {
     # Compute what needs to be put into LD_LIBRARY_PATH
     set always_ld_library_path ".:${blddir}/.libs"
 
-    # Get offload-related variables from environment (exported by Makefile)
-    set offload_targets [getenv OFFLOAD_TARGETS]
-    set offload_additional_options [getenv OFFLOAD_ADDITIONAL_OPTIONS]
-    set offload_additional_lib_paths [getenv OFFLOAD_ADDITIONAL_LIB_PATHS]
-
     # Add liboffloadmic build directory in LD_LIBRARY_PATH to support
     # non-fallback testing for Intel MIC targets
+    global offload_targets
     if { [string match "*,intelmic,*" ",$offload_targets,"] } {
 	append always_ld_library_path ":${blddir}/../liboffloadmic/.libs"
 	append always_ld_library_path ":${blddir}/../liboffloadmic/plugin/.libs"
@@ -122,6 +141,7 @@ proc libgomp_init { args } {
 	append always_ld_library_path ":${blddir}/../libstdc++-v3/src/.libs"
     }
 
+    global offload_additional_lib_paths
     if { $offload_additional_lib_paths != "" } {
 	append always_ld_library_path "${offload_additional_lib_paths}"
     }
@@ -158,9 +178,29 @@ proc libgomp_init { args } {
         lappend ALWAYS_CFLAGS "additional_flags=-B${blddir}/.libs"
         lappend ALWAYS_CFLAGS "additional_flags=-I${blddir}"
         lappend ALWAYS_CFLAGS "ldflags=-L${blddir}/.libs"
+	# The top-level include directory, for gomp-constants.h.
+	lappend ALWAYS_CFLAGS "additional_flags=-I${srcdir}/../../include"
     }
     lappend ALWAYS_CFLAGS "additional_flags=-I${srcdir}/.."
 
+    # For build-tree testing, also consider the CUDA paths used for builing.
+    # For installed testing, we assume all that to be provided in the sysroot.
+    if { $blddir != "" } {
+	global cuda_driver_include
+	global cuda_driver_lib
+	if { $cuda_driver_include != "" } {
+	    # Stop gfortran from freaking out:
+	    # Warning: Nonexistent include directory "[...]"
+	    if {[file exists $cuda_driver_include]} {
+		lappend ALWAYS_CFLAGS "additional_flags=-I$cuda_driver_include"
+	    }
+	}
+	if { $cuda_driver_lib != "" } {
+	    lappend ALWAYS_CFLAGS "additional_flags=-L$cuda_driver_lib"
+	    append always_ld_library_path ":$cuda_driver_lib"
+	}
+    }
+
     # We use atomic operations in the testcases to validate results.
     if { ([istarget i?86-*-*] || [istarget x86_64-*-*])
 	 && [check_effective_target_ia32] } {
@@ -191,6 +231,7 @@ proc libgomp_init { args } {
 
     # Used for support non-fallback offloading.
     # Help GCC to find target mkoffload.
+    global offload_additional_options
     if { $offload_additional_options != "" } {
 	lappend ALWAYS_CFLAGS "additional_flags=${offload_additional_options}"
     }
@@ -278,3 +319,29 @@ proc check_effective_target_offload_device { } {
 	}
     } ]
 }
+
+# Return 1 if at least one nvidia board is present.
+
+proc check_effective_target_openacc_nvidia_accel_present { } {
+    return [check_runtime openacc_nvidia_accel_present {
+	#include <openacc.h>
+	int main () {
+	return !(acc_get_num_devices (acc_device_nvidia) > 0);
+	}
+    } "" ]
+}
+
+# Return 1 if at least one nvidia board is present, and the nvidia device type
+# is selected by default by means of setting the environment variable
+# ACC_DEVICE_TYPE.
+
+proc check_effective_target_openacc_nvidia_accel_selected { } {
+    if { ![check_effective_target_openacc_nvidia_accel_present] } {
+	return 0;
+    }
+    global offload_target_openacc
+    if { $offload_target_openacc == "nvidia" } {
+        return 1;
+    }
+    return 0;
+}
diff --git a/libgomp/testsuite/libgomp-test-support.exp.in b/libgomp/testsuite/libgomp-test-support.exp.in
new file mode 100644
index 00000000000..764bec06c29
--- /dev/null
+++ b/libgomp/testsuite/libgomp-test-support.exp.in
@@ -0,0 +1,4 @@
+set cuda_driver_include "@CUDA_DRIVER_INCLUDE@"
+set cuda_driver_lib "@CUDA_DRIVER_LIB@"
+
+set offload_targets "@offload_targets@"
diff --git a/libgomp/testsuite/libgomp.oacc-c++/c++.exp b/libgomp/testsuite/libgomp.oacc-c++/c++.exp
new file mode 100644
index 00000000000..f486f9b97ba
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c++/c++.exp
@@ -0,0 +1,107 @@
+# This whole file adapted from libgomp.c++/c++.exp.
+
+load_lib libgomp-dg.exp
+load_gcc_lib gcc-dg.exp
+
+global shlib_ext
+
+set shlib_ext [get_shlib_extension]
+set lang_link_flags "-lstdc++"
+set lang_test_file_found 0
+set lang_library_path "../libstdc++-v3/src/.libs"
+if [info exists lang_include_flags] then {
+    unset lang_include_flags
+}
+
+# Initialize dg.
+dg-init
+
+# Turn on OpenACC.
+lappend ALWAYS_CFLAGS "additional_flags=-fopenacc"
+
+# Switch into C++ mode.  Otherwise, the libgomp.oacc-c-c++-common/*.c
+# files would be compiled as C files.
+set SAVE_GCC_UNDER_TEST "$GCC_UNDER_TEST"
+set GCC_UNDER_TEST "$GCC_UNDER_TEST -x c++"
+
+set blddir [lookfor_file [get_multilibs] libgomp]
+
+
+if { $blddir != "" } {
+    # Look for a static libstdc++ first.
+    if [file exists "${blddir}/${lang_library_path}/libstdc++.a"] {
+        set lang_test_file "${lang_library_path}/libstdc++.a"
+        set lang_test_file_found 1
+        # We may have a shared only build, so look for a shared libstdc++.
+    } elseif [file exists "${blddir}/${lang_library_path}/libstdc++.${shlib_ext}"] {
+        set lang_test_file "${lang_library_path}/libstdc++.${shlib_ext}"
+        set lang_test_file_found 1
+    } else {
+        puts "No libstdc++ library found, will not execute c++ tests"
+    }
+} elseif { [info exists GXX_UNDER_TEST] } {
+    set lang_test_file_found 1
+    # Needs to exist for libgomp.exp.
+    set lang_test_file ""
+} else {
+    puts "GXX_UNDER_TEST not defined, will not execute c++ tests"
+}
+
+if { $lang_test_file_found } {
+    # Gather a list of all tests.
+    set tests [lsort [concat \
+			  [find $srcdir/$subdir *.C] \
+			  [find $srcdir/$subdir/../libgomp.oacc-c-c++-common *.c]]]
+
+    if { $blddir != "" } {
+        set ld_library_path "$always_ld_library_path:${blddir}/${lang_library_path}"
+    } else {
+        set ld_library_path "$always_ld_library_path"
+    }
+    append ld_library_path [gcc-set-multilib-library-path $GCC_UNDER_TEST]
+    set_ld_library_path_env_vars
+
+    set flags_file "${blddir}/../libstdc++-v3/scripts/testsuite_flags"
+    if { [file exists $flags_file] } {
+	set libstdcxx_includes [exec sh $flags_file --build-includes]
+    } else {
+	set libstdcxx_includes ""
+    }
+
+    # Test OpenACC with available accelerators.
+    foreach offload_target_openacc $offload_targets_s_openacc {
+	set tagopt "-DACC_DEVICE_TYPE_$offload_target_openacc=1"
+
+	switch $offload_target_openacc {
+	    host {
+		set acc_mem_shared 1
+	    }
+	    host_nonshm {
+		set acc_mem_shared 0
+	    }
+	    nvidia {
+		# Copy ptx file (TEMPORARY)
+		remote_download host $srcdir/libgomp.oacc-c-c++-common/subr.ptx
+
+		# Where timer.h lives
+		lappend ALWAYS_CFLAGS "additional_flags=-I${srcdir}/libgomp.oacc-c-c++-common"
+
+		set acc_mem_shared 0
+	    }
+	    default {
+		set acc_mem_shared 0
+	    }
+	}
+	set tagopt "$tagopt -DACC_MEM_SHARED=$acc_mem_shared"
+
+	setenv ACC_DEVICE_TYPE $offload_target_openacc
+
+	dg-runtest $tests "$tagopt" $libstdcxx_includes
+    }
+}
+
+# See above.
+set GCC_UNDER_TEST "$SAVE_GCC_UNDER_TEST"
+
+# All done.
+dg-finish
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/abort-1.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/abort-1.c
new file mode 100644
index 00000000000..f88b9e3c7f5
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/abort-1.c
@@ -0,0 +1,17 @@
+/* { dg-do run } */
+/* { dg-shouldfail "" { *-*-* } { "*" } { "" } } */
+
+#include <stdlib.h>
+
+int
+main (void)
+{
+
+#pragma acc parallel
+  {
+    abort ();
+  }
+
+  return 0;
+}
+
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/abort-2.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/abort-2.c
new file mode 100644
index 00000000000..debb81e78f8
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/abort-2.c
@@ -0,0 +1,17 @@
+/* { dg-do run } */
+
+#include <stdlib.h>
+
+int
+main (int argc, char **argv)
+{
+
+#pragma acc parallel
+  {
+    if (argc != 1)
+      abort ();
+  }
+
+  return 0;
+}
+
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/abort-3.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/abort-3.c
new file mode 100644
index 00000000000..be7aaa86551
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/abort-3.c
@@ -0,0 +1,17 @@
+/* { dg-do run } */
+/* { dg-shouldfail "" { *-*-* } { "*" } { "" } } */
+
+#include <stdlib.h>
+
+int
+main (void)
+{
+
+#pragma acc kernels
+  {
+    abort ();
+  }
+
+  return 0;
+}
+
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/abort-4.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/abort-4.c
new file mode 100644
index 00000000000..c29ca3fb118
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/abort-4.c
@@ -0,0 +1,17 @@
+/* { dg-do run } */
+
+#include <stdlib.h>
+
+int
+main (int argc, char **argv)
+{
+
+#pragma acc kernels
+  {
+    if (argc != 1)
+      abort ();
+  }
+
+  return 0;
+}
+
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/acc_on_device-1.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/acc_on_device-1.c
new file mode 100644
index 00000000000..81ea47689a9
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/acc_on_device-1.c
@@ -0,0 +1,75 @@
+/* Disable the acc_on_device builtin; we want to test the libgomp library
+   function.  */
+/* { dg-additional-options "-fno-builtin-acc_on_device" } */
+
+#include <stdlib.h>
+#include <openacc.h>
+
+int
+main (int argc, char *argv[])
+{
+  /* Host.  */
+
+  {
+    if (!acc_on_device (acc_device_none))
+      abort ();
+    if (!acc_on_device (acc_device_host))
+      abort ();
+    if (acc_on_device (acc_device_host_nonshm))
+      abort ();
+    if (acc_on_device (acc_device_not_host))
+      abort ();
+    if (acc_on_device (acc_device_nvidia))
+      abort ();
+  }
+
+
+  /* Host via offloading fallback mode.  */
+
+#pragma acc parallel if(0)
+  {
+    if (!acc_on_device (acc_device_none))
+      abort ();
+    if (!acc_on_device (acc_device_host))
+      abort ();
+    if (acc_on_device (acc_device_host_nonshm))
+      abort ();
+    if (acc_on_device (acc_device_not_host))
+      abort ();
+    if (acc_on_device (acc_device_nvidia))
+      abort ();
+  }
+
+
+#if !ACC_DEVICE_TYPE_host
+
+  /* Offloaded.  */
+
+#pragma acc parallel
+  {
+    if (acc_on_device (acc_device_none))
+      abort ();
+    if (acc_on_device (acc_device_host))
+      abort ();
+#if ACC_DEVICE_TYPE_host_nonshm
+    if (!acc_on_device (acc_device_host_nonshm))
+      abort ();
+#else
+    if (acc_on_device (acc_device_host_nonshm))
+      abort ();
+#endif
+    if (!acc_on_device (acc_device_not_host))
+      abort ();
+#if ACC_DEVICE_TYPE_nvidia
+    if (!acc_on_device (acc_device_nvidia))
+      abort ();
+#else
+    if (acc_on_device (acc_device_nvidia))
+      abort ();
+#endif
+  }
+
+#endif
+
+  return 0;
+}
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/asyncwait-1.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/asyncwait-1.c
new file mode 100644
index 00000000000..22cef6db29f
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/asyncwait-1.c
@@ -0,0 +1,466 @@
+/* { dg-do run { target openacc_nvidia_accel_selected } } */
+/* { dg-additional-options "-lcuda" } */
+
+#include <openacc.h>
+#include <stdlib.h>
+#include "cuda.h"
+
+#include <stdio.h>
+#include <sys/time.h>
+
+int
+main (int argc, char **argv)
+{
+    CUresult r;
+    CUstream stream1;
+    int N = 128; //1024 * 1024;
+    float *a, *b, *c, *d, *e;
+    int i;
+    int nbytes;
+
+    acc_init (acc_device_nvidia);
+
+    nbytes = N * sizeof (float);
+
+    a = (float *) malloc (nbytes);
+    b = (float *) malloc (nbytes);
+    c = (float *) malloc (nbytes);
+    d = (float *) malloc (nbytes);
+    e = (float *) malloc (nbytes);
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 3.0;
+        b[i] = 0.0;
+    }
+
+#pragma acc data copy (a[0:N]) copy (b[0:N]) copyin (N)
+    {
+
+#pragma acc parallel async
+    {
+        int ii;
+
+        for (ii = 0; ii < N; ii++)
+            b[ii] = a[ii];
+    }
+
+#pragma acc wait
+
+    }
+
+    for (i = 0; i < N; i++)
+    {
+        if (a[i] != 3.0)
+            abort ();
+
+        if (b[i] != 3.0)
+            abort ();
+    }
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 2.0;
+        b[i] = 0.0;
+    }
+
+#pragma acc data copy (a[0:N]) copy (b[0:N]) copyin (N)
+    {
+
+#pragma acc parallel async (1)
+    {
+        int ii;
+
+        for (ii = 0; ii < N; ii++)
+            b[ii] = a[ii];
+    }
+
+#pragma acc wait (1)
+
+    }
+
+    for (i = 0; i < N; i++)
+    {
+        if (a[i] != 2.0)
+            abort ();
+
+        if (b[i] != 2.0)
+            abort ();
+    }
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 3.0;
+        b[i] = 0.0;
+        c[i] = 0.0;
+        d[i] = 0.0;
+    }
+
+#pragma acc data copy (a[0:N]) copy (b[0:N]) copy (c[0:N]) copy (d[0:N]) copyin (N)
+    {
+
+#pragma acc parallel async (1)
+    {
+        int ii;
+
+        for (ii = 0; ii < N; ii++)
+            b[ii] = (a[ii] * a[ii] * a[ii]) / a[ii];
+    }
+
+#pragma acc parallel async (1)
+    {
+        int ii;
+
+        for (ii = 0; ii < N; ii++)
+            c[ii] = (a[ii] + a[ii] + a[ii] + a[ii]) / a[ii];
+    }
+
+
+#pragma acc parallel async (1)
+    {
+        int ii;
+
+        for (ii = 0; ii < N; ii++)
+            d[ii] = ((a[ii] * a[ii] + a[ii]) / a[ii]) - a[ii];
+    }
+
+#pragma acc wait (1)
+
+    }
+
+    for (i = 0; i < N; i++)
+    {
+        if (a[i] != 3.0)
+            abort ();
+
+        if (b[i] != 9.0)
+            abort ();
+
+        if (c[i] != 4.0)
+            abort ();
+
+        if (d[i] != 1.0)
+            abort ();
+    }
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 2.0;
+        b[i] = 0.0;
+        c[i] = 0.0;
+        d[i] = 0.0;
+        e[i] = 0.0;
+    }
+
+#pragma acc data copy (a[0:N], b[0:N], c[0:N], d[0:N], e[0:N]) copyin (N)
+    {
+
+#pragma acc parallel async (1)
+    {
+        int ii;
+
+        for (ii = 0; ii < N; ii++)
+            b[ii] = (a[ii] * a[ii] * a[ii]) / a[ii];
+    }
+
+#pragma acc parallel async (1)
+    {
+        int ii;
+
+        for (ii = 0; ii < N; ii++)
+            c[ii] = (a[ii] + a[ii] + a[ii] + a[ii]) / a[ii];
+    }
+
+#pragma acc parallel async (1)
+    {
+        int ii;
+
+        for (ii = 0; ii < N; ii++)
+            d[ii] = ((a[ii] * a[ii] + a[ii]) / a[ii]) - a[ii];
+    }
+
+#pragma acc parallel wait (1) async (1)
+    {
+        int ii;
+
+        for (ii = 0; ii < N; ii++)
+            e[ii] = a[ii] + b[ii] + c[ii] + d[ii];
+    }
+
+#pragma acc wait (1)
+
+    }
+
+    for (i = 0; i < N; i++)
+    {
+        if (a[i] != 2.0)
+            abort ();
+
+        if (b[i] != 4.0)
+            abort ();
+
+        if (c[i] != 4.0)
+            abort ();
+
+        if (d[i] != 1.0)
+            abort ();
+
+        if (e[i] != 11.0)
+            abort ();
+    }
+
+
+    r = cuStreamCreate (&stream1, CU_STREAM_NON_BLOCKING);
+    if (r != CUDA_SUCCESS)
+    {
+        fprintf (stderr, "cuStreamCreate failed: %d\n", r);
+        abort ();
+    }
+
+    acc_set_cuda_stream (1, stream1);
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 5.0;
+        b[i] = 0.0;
+    }
+
+#pragma acc data copy (a[0:N], b[0:N]) copyin (N)
+    {
+
+#pragma acc parallel async (1)
+    {
+        int ii;
+
+        for (ii = 0; ii < N; ii++)
+            b[ii] = a[ii];
+    }
+
+#pragma acc wait (1)
+
+    }
+
+    for (i = 0; i < N; i++)
+    {
+        if (a[i] != 5.0)
+            abort ();
+
+        if (b[i] != 5.0)
+            abort ();
+    }
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 7.0;
+        b[i] = 0.0;
+        c[i] = 0.0;
+        d[i] = 0.0;
+    }
+
+#pragma acc data copy (a[0:N]) copy (b[0:N]) copy (c[0:N]) copy (d[0:N]) copyin (N)
+    {
+
+#pragma acc parallel async (1)
+    {
+        int ii;
+
+        for (ii = 0; ii < N; ii++)
+            b[ii] = (a[ii] * a[ii] * a[ii]) / a[ii];
+    }
+
+#pragma acc parallel async (1)
+    {
+        int ii;
+
+        for (ii = 0; ii < N; ii++)
+            c[ii] = (a[ii] + a[ii] + a[ii] + a[ii]) / a[ii];
+    }
+
+#pragma acc parallel async (1)
+    {
+        int ii;
+
+        for (ii = 0; ii < N; ii++)
+            d[ii] = ((a[ii] * a[ii] + a[ii]) / a[ii]) - a[ii];
+    }
+
+#pragma acc wait (1)
+
+    }
+
+    for (i = 0; i < N; i++)
+    {
+        if (a[i] != 7.0)
+            abort ();
+
+        if (b[i] != 49.0)
+            abort ();
+
+        if (c[i] != 4.0)
+            abort ();
+
+        if (d[i] != 1.0)
+            abort ();
+    }
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 3.0;
+        b[i] = 0.0;
+        c[i] = 0.0;
+        d[i] = 0.0;
+        e[i] = 0.0;
+    }
+
+#pragma acc data copy (a[0:N], b[0:N], c[0:N], d[0:N], e[0:N]) copyin (N)
+    {
+
+#pragma acc parallel async (1)
+    {
+        int ii;
+
+        for (ii = 0; ii < N; ii++)
+            b[ii] = (a[ii] * a[ii] * a[ii]) / a[ii];
+    }
+
+#pragma acc parallel async (1)
+    {
+        int ii;
+
+        for (ii = 0; ii < N; ii++)
+            c[ii] = (a[ii] + a[ii] + a[ii] + a[ii]) / a[ii];
+    }
+
+#pragma acc parallel async (1)
+    {
+        int ii;
+
+        for (ii = 0; ii < N; ii++)
+            d[ii] = ((a[ii] * a[ii] + a[ii]) / a[ii]) - a[ii];
+    }
+
+#pragma acc parallel wait (1) async (1)
+    {
+        int ii;
+
+        for (ii = 0; ii < N; ii++)
+            e[ii] = a[ii] + b[ii] + c[ii] + d[ii];
+    }
+
+#pragma acc wait (1)
+
+    }
+
+    for (i = 0; i < N; i++)
+    {
+        if (a[i] != 3.0)
+            abort ();
+
+        if (b[i] != 9.0)
+            abort ();
+
+        if (c[i] != 4.0)
+            abort ();
+
+        if (d[i] != 1.0)
+            abort ();
+
+        if (e[i] != 17.0)
+            abort ();
+    }
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 4.0;
+        b[i] = 0.0;
+        c[i] = 0.0;
+        d[i] = 0.0;
+        e[i] = 0.0;
+    }
+
+#pragma acc data copyin (a[0:N], b[0:N], c[0:N]) copyin (N)
+    {
+
+#pragma acc parallel async (1)
+    {
+        int ii;
+
+        for (ii = 0; ii < N; ii++)
+            b[ii] = (a[ii] * a[ii] * a[ii]) / a[ii];
+    }
+
+#pragma acc parallel async (1)
+    {
+        int ii;
+
+        for (ii = 0; ii < N; ii++)
+            c[ii] = (a[ii] + a[ii] + a[ii] + a[ii]) / a[ii];
+    }
+
+#pragma acc update host (a[0:N], b[0:N], c[0:N]) wait (1)
+
+    }
+
+    for (i = 0; i < N; i++)
+    {
+        if (a[i] != 4.0)
+            abort ();
+
+        if (b[i] != 16.0)
+            abort ();
+
+        if (c[i] != 4.0)
+            abort ();
+    }
+
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 5.0;
+        b[i] = 0.0;
+        c[i] = 0.0;
+        d[i] = 0.0;
+        e[i] = 0.0;
+    }
+
+#pragma acc data copyin (a[0:N], b[0:N], c[0:N]) copyin (N)
+    {
+
+#pragma acc parallel async (1)
+    {
+        int ii;
+
+        for (ii = 0; ii < N; ii++)
+            b[ii] = (a[ii] * a[ii] * a[ii]) / a[ii];
+    }
+
+#pragma acc parallel async (1)
+    {
+        int ii;
+
+        for (ii = 0; ii < N; ii++)
+            c[ii] = (a[ii] + a[ii] + a[ii] + a[ii]) / a[ii];
+    }
+
+#pragma acc update host (a[0:N], b[0:N], c[0:N]) async (1)
+
+#pragma acc wait (1)
+
+    }
+
+    for (i = 0; i < N; i++)
+    {
+        if (a[i] != 5.0)
+            abort ();
+
+        if (b[i] != 25.0)
+            abort ();
+
+        if (c[i] != 4.0)
+            abort ();
+    }
+
+    acc_shutdown (acc_device_nvidia);
+
+    return 0;
+}
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/cache-1.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/cache-1.c
new file mode 100644
index 00000000000..3f1f0bb3764
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/cache-1.c
@@ -0,0 +1,48 @@
+int
+main (int argc, char **argv)
+{
+#define N   2
+    int a[N], b[N];
+    int i;
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 3;
+        b[i] = 0;
+    }
+
+#pragma acc parallel copyin (a[0:N]) copyout (b[0:N])
+{
+    int ii;
+
+    for (ii = 0; ii < N; ii++)
+    {
+        const int idx = ii;
+        int n = 1;
+        const int len = n;
+
+#pragma acc cache (a[0:N])
+
+#pragma acc cache (a[0:N], b[0:N])
+
+#pragma acc cache (a[0])
+
+#pragma acc cache (a[0], a[1], b[0:N])
+
+#pragma acc cache (a[idx])
+
+#pragma acc cache (a[idx:len])
+
+        b[ii] = a[ii];
+    }
+}
+
+
+    for (i = 0; i < N; i++)
+    {
+        if (a[i] != b[i])
+            __builtin_abort ();
+    }
+
+    return 0;
+}
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/clauses-1.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/clauses-1.c
new file mode 100644
index 00000000000..51c0cf58a90
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/clauses-1.c
@@ -0,0 +1,623 @@
+/* { dg-do run } */
+/* { dg-skip-if "" { *-*-* } { "*" } { "-DACC_MEM_SHARED=0" } } */
+
+#include <openacc.h>
+#include <string.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <stdbool.h>
+
+int
+main (int argc, char **argv)
+{
+    int N = 8;
+    float *a, *b, *c, *d;
+    int i;
+
+    a = (float *) malloc (N * sizeof (float));
+    b = (float *) malloc (N * sizeof (float));
+    c = (float *) malloc (N * sizeof (float));
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 3.0;
+        b[i] = 0.0;
+    }
+
+#pragma acc parallel copyin (a[0:N]) copyout (b[0:N])
+    {
+        int ii;
+
+        for (ii = 0; ii < N; ii++)
+            b[ii] = a[ii];
+    }
+
+    for (i = 0; i < N; i++)
+    {
+        if (b[i] != 3.0)
+            abort ();
+    }
+
+    if (acc_is_present (&a[0], (N * sizeof (float))))
+      abort ();
+
+    if (acc_is_present (&b[0], (N * sizeof (float))))
+      abort ();
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 5.0;
+        b[i] = 1.0;
+    }
+
+#pragma acc parallel copyin (a[0:N]) copyout (b[0:N])
+    {
+        int ii;
+
+        for (ii = 0; ii < N; ii++)
+            b[ii] = a[ii];
+    }
+
+    for (i = 0; i < N; i++)
+    {
+        if (b[i] != 5.0)
+            abort ();
+    }
+
+    if (acc_is_present (&a[0], (N * sizeof (float))))
+      abort ();
+
+    if (acc_is_present (&b[0], (N * sizeof (float))))
+      abort ();
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 6.0;
+        b[i] = 0.0;
+    }
+
+    d = (float *) acc_copyin (&a[0], N * sizeof (float));
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 9.0;
+    }
+
+#pragma acc parallel present_or_copyin (a[0:N]) copyout (b[0:N])
+    {
+        int ii;
+
+        for (ii = 0; ii < N; ii++)
+            b[ii] = a[ii];
+    }
+
+    for (i = 0; i < N; i++)
+    {
+        if (b[i] != 6.0)
+            abort ();
+    }
+
+    if (!acc_is_present (&a[0], (N * sizeof (float))))
+      abort ();
+
+    if (acc_is_present (&b[0], (N * sizeof (float))))
+      abort ();
+
+    acc_free (d);
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 6.0;
+        b[i] = 0.0;
+    }
+
+#pragma acc parallel copyin (a[0:N]) present_or_copyout (b[0:N])
+    {
+        int ii;
+
+        for (ii = 0; ii < N; ii++)
+            b[ii] = a[ii];
+    }
+
+    for (i = 0; i < N; i++)
+    {
+        if (b[i] != 6.0)
+            abort ();
+    }
+
+    if (acc_is_present (&a[0], (N * sizeof (float))))
+      abort ();
+
+    if (acc_is_present (&b[0], (N * sizeof (float))))
+      abort ();
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 5.0;
+        b[i] = 2.0;
+    }
+
+    d = (float *) acc_copyin (&b[0], N * sizeof (float));
+
+#pragma acc parallel copyin (a[0:N]) present_or_copyout (b[0:N])
+    {
+        int ii;
+
+        for (ii = 0; ii < N; ii++)
+            b[ii] = a[ii];
+    }
+
+    for (i = 0; i < N; i++)
+    {
+        if (a[i] != 5.0)
+            abort ();
+
+        if (b[i] != 2.0)
+            abort ();
+    }
+
+    if (acc_is_present (&a[0], (N * sizeof (float))))
+      abort ();
+
+    if (!acc_is_present (&b[0], (N * sizeof (float))))
+      abort ();
+
+    acc_free (d);
+
+    if (acc_is_present (&b[0], (N * sizeof (float))))
+      abort ();
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 3.0;
+        b[i] = 4.0;
+    }
+
+#pragma acc parallel copy (a[0:N]) copyout (b[0:N])
+    {
+        int ii;
+
+        for (ii = 0; ii < N; ii++)
+        {
+            a[ii] = a[ii] + 1;
+            b[ii] = a[ii] + 2;
+        }
+    }
+
+    for (i = 0; i < N; i++)
+    {
+        if (a[i] != 4.0)
+            abort ();
+
+        if (b[i] != 6.0)
+            abort ();
+    }
+
+    if (acc_is_present (&a[0], (N * sizeof (float))))
+      abort ();
+
+    if (acc_is_present (&b[0], (N * sizeof (float))))
+      abort ();
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 4.0;
+        b[i] = 7.0;
+    }
+
+#pragma acc parallel present_or_copy (a[0:N]) present_or_copy (b[0:N])
+    {
+        int ii;
+
+        for (ii = 0; ii < N; ii++)
+        {
+            a[ii] = a[ii] + 1;
+            b[ii] = b[ii] + 2;
+        }
+    }
+
+    for (i = 0; i < N; i++)
+    {
+        if (a[i] != 5.0)
+            abort ();
+
+        if (b[i] != 9.0)
+            abort ();
+    }
+
+    if (acc_is_present (&a[0], (N * sizeof (float))))
+      abort ();
+
+    if (acc_is_present (&b[0], (N * sizeof (float))))
+      abort ();
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 3.0;
+        b[i] = 7.0;
+    }
+
+    d = (float *) acc_copyin (&a[0], N * sizeof (float));
+    d = (float *) acc_copyin (&b[0], N * sizeof (float));
+
+#pragma acc parallel present_or_copy (a[0:N]) present_or_copy (b[0:N])
+    {
+        int ii;
+
+        for (ii = 0; ii < N; ii++)
+        {
+            a[ii] = a[ii] + 1;
+            b[ii] = b[ii] + 2;
+        }
+    }
+
+    for (i = 0; i < N; i++)
+    {
+        if (a[i] != 3.0)
+            abort ();
+
+        if (b[i] != 7.0)
+            abort ();
+    }
+
+    if (!acc_is_present (&a[0], (N * sizeof (float))))
+      abort ();
+
+    if (!acc_is_present (&b[0], (N * sizeof (float))))
+      abort ();
+
+    d = (float *) acc_deviceptr (&a[0]);
+    acc_unmap_data (&a[0]);
+    acc_free (d);
+
+    d = (float *) acc_deviceptr (&b[0]);
+    acc_unmap_data (&b[0]);
+    acc_free (d);
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 3.0;
+        b[i] = 7.0;
+    }
+
+#pragma acc parallel copyin (a[0:N]) create (c[0:N]) copyout (b[0:N])
+    {
+        int ii;
+
+        for (ii = 0; ii < N; ii++)
+        {
+            c[ii] = a[ii];
+            b[ii] = c[ii];
+        }
+    }
+
+    for (i = 0; i < N; i++)
+    {
+        if (a[i] != 3.0)
+            abort ();
+
+        if (b[i] != 3.0)
+            abort ();
+    }
+
+    if (acc_is_present (&a[0], (N * sizeof (float))))
+      abort ();
+
+    if (acc_is_present (&b[0], (N * sizeof (float))))
+      abort ();
+
+    if (acc_is_present (&c[0], (N * sizeof (float))))
+      abort ();
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 4.0;
+        b[i] = 8.0;
+    }
+
+#pragma acc parallel copyin (a[0:N]) present_or_create (c[0:N]) copyout (b[0:N])
+    {
+        int ii;
+
+        for (ii = 0; ii < N; ii++)
+        {
+            c[ii] = a[ii];
+            b[ii] = c[ii];
+        }
+    }
+
+    for (i = 0; i < N; i++)
+    {
+        if (a[i] != 4.0)
+            abort ();
+
+        if (b[i] != 4.0)
+            abort ();
+    }
+
+    if (acc_is_present (&a[0], (N * sizeof (float))))
+      abort ();
+
+    if (acc_is_present (&b[0], (N * sizeof (float))))
+      abort ();
+
+    if (acc_is_present (&c[0], (N * sizeof (float))))
+      abort ();
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 2.0;
+        b[i] = 5.0;
+    }
+
+    d = (float *) acc_malloc (N * sizeof (float));
+    acc_map_data (c, d, N * sizeof (float));
+
+#pragma acc parallel copyin (a[0:N]) present_or_create (c[0:N]) copyout (b[0:N])
+    {
+        int ii;
+
+        for (ii = 0; ii < N; ii++)
+        {
+            c[ii] = a[ii];
+            b[ii] = c[ii];
+        }
+    }
+
+    for (i = 0; i < N; i++)
+    {
+        if (a[i] != 2.0)
+            abort ();
+
+        if (b[i] != 2.0)
+            abort ();
+    }
+
+    if (acc_is_present (a, (N * sizeof (float))))
+      abort ();
+
+    if (acc_is_present (b, (N * sizeof (float))))
+      abort ();
+
+    if (!acc_is_present (c, (N * sizeof (float))))
+      abort ();
+
+    d = (float *) acc_deviceptr (c);
+
+    acc_unmap_data (c);
+
+    acc_free (d);
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 4.0;
+        b[i] = 8.0;
+    }
+
+    d = (float *) acc_malloc (N * sizeof (float));
+    acc_map_data (c, d, N * sizeof (float));
+
+#pragma acc parallel copyin (a[0:N]) present (c[0:N]) copyout (b[0:N])
+    {
+        int ii;
+
+        for (ii = 0; ii < N; ii++)
+        {
+            c[ii] = a[ii];
+            b[ii] = c[ii];
+        }
+    }
+
+    for (i = 0; i < N; i++)
+    {
+        if (a[i] != 4.0)
+            abort ();
+
+        if (b[i] != 4.0)
+            abort ();
+    }
+
+    if (acc_is_present (a, (N * sizeof (float))))
+      abort ();
+
+    if (acc_is_present (b, (N * sizeof (float))))
+      abort ();
+
+    if (!acc_is_present (c, (N * sizeof (float))))
+      abort ();
+
+    acc_unmap_data (c);
+
+    acc_free (d);
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 4.0;
+        b[i] = 8.0;
+    }
+
+    acc_copyin (a, N * sizeof (float));
+
+    d = (float *) acc_malloc (N * sizeof (float));
+    acc_map_data (b, d, N * sizeof (float));
+
+    d = (float *) acc_malloc (N * sizeof (float));
+    acc_map_data (c, d, N * sizeof (float));
+
+#pragma acc parallel present (a[0:N]) present (c[0:N]) present (b[0:N])
+    {
+        int ii;
+
+        for (ii = 0; ii < N; ii++)
+        {
+            c[ii] = a[ii];
+            b[ii] = c[ii];
+        }
+    }
+
+    if (!acc_is_present (a, (N * sizeof (float))))
+      abort ();
+
+    if (!acc_is_present (b, (N * sizeof (float))))
+      abort ();
+
+    if (!acc_is_present (c, (N * sizeof (float))))
+      abort ();
+
+    acc_copyout (b, N * sizeof (float));
+
+    for (i = 0; i < N; i++)
+    {
+        if (a[i] != 4.0)
+            abort ();
+
+        if (b[i] != 4.0)
+            abort ();
+    }
+
+    d = (float *) acc_deviceptr (a);
+
+    acc_unmap_data (a);
+
+    acc_free (d);
+
+    d = (float *) acc_deviceptr (c);
+
+    acc_unmap_data (c);
+
+    acc_free (d);
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 3.0;
+        b[i] = 6.0;
+    }
+
+    d = (float *) acc_malloc (N * sizeof (float));
+
+#pragma acc parallel copyin (a[0:N]) deviceptr (d) copyout (b[0:N])
+    {
+        int ii;
+
+        for (ii = 0; ii < N; ii++)
+        {
+            d[ii] = a[ii];
+            b[ii] = d[ii];
+        }
+    }
+
+    for (i = 0; i < N; i++)
+    {
+        if (a[i] != 3.0)
+            abort ();
+
+        if (b[i] != 3.0)
+            abort ();
+    }
+
+    if (acc_is_present (a, (N * sizeof (float))))
+      abort ();
+
+    if (acc_is_present (b, (N * sizeof (float))))
+      abort ();
+
+    acc_free (d);
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 6.0;
+        b[i] = 0.0;
+    }
+
+    d = (float *) acc_copyin (&a[0], N * sizeof (float));
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 9.0;
+    }
+
+#pragma acc parallel pcopyin (a[0:N]) copyout (b[0:N])
+    {
+        int ii;
+
+        for (ii = 0; ii < N; ii++)
+            b[ii] = a[ii];
+    }
+
+    for (i = 0; i < N; i++)
+    {
+        if (b[i] != 6.0)
+            abort ();
+    }
+
+    if (!acc_is_present (&a[0], (N * sizeof (float))))
+      abort ();
+
+    if (acc_is_present (&b[0], (N * sizeof (float))))
+      abort ();
+
+    acc_free (d);
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 6.0;
+        b[i] = 0.0;
+    }
+
+#pragma acc parallel copyin (a[0:N]) pcopyout (b[0:N])
+    {
+        int ii;
+
+        for (ii = 0; ii < N; ii++)
+            b[ii] = a[ii];
+    }
+
+    for (i = 0; i < N; i++)
+    {
+        if (b[i] != 6.0)
+            abort ();
+    }
+
+    if (acc_is_present (&a[0], (N * sizeof (float))))
+      abort ();
+
+    if (acc_is_present (&b[0], (N * sizeof (float))))
+      abort ();
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 5.0;
+        b[i] = 7.0;
+    }
+
+#pragma acc parallel copyin (a[0:N]) pcreate (c[0:N]) copyout (b[0:N])
+    {
+        int ii;
+
+        for (ii = 0; ii < N; ii++)
+        {
+            c[ii] = a[ii];
+            b[ii] = c[ii];
+        }
+    }
+
+    for (i = 0; i < N; i++)
+    {
+        if (a[i] != 5.0)
+            abort ();
+
+        if (b[i] != 5.0)
+            abort ();
+    }
+
+    if (acc_is_present (&a[0], (N * sizeof (float))))
+      abort ();
+
+    if (acc_is_present (&b[0], (N * sizeof (float))))
+      abort ();
+
+    if (acc_is_present (&c[0], (N * sizeof (float))))
+      abort ();
+
+    return 0;
+}
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/clauses-2.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/clauses-2.c
new file mode 100644
index 00000000000..8dc45cb6287
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/clauses-2.c
@@ -0,0 +1,67 @@
+/* { dg-do run } */
+/* { dg-skip-if "" { *-*-* } { "*" } { "-DACC_MEM_SHARED=0" } } */
+
+#include <openacc.h>
+#include <string.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <stdbool.h>
+
+int
+main (int argc, char **argv)
+{
+    int N = 8;
+    float *a, *b, *c, *d;
+    int i;
+
+    a = (float *) malloc (N * sizeof (float));
+    b = (float *) malloc (N * sizeof (float));
+    c = (float *) malloc (N * sizeof (float));
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 2.0;
+        b[i] = 5.0;
+    }
+
+    d = (float *) acc_malloc (N * sizeof (float));
+    acc_map_data (c, d, N * sizeof (float));
+
+#pragma acc parallel copyin (a[0:N]) present_or_create (c[0:N+1]) copyout (b[0:N])
+    {
+        int ii;
+
+        for (ii = 0; ii < N; ii++)
+        {
+            c[ii] = a[ii];
+            b[ii] = c[ii];
+        }
+    }
+
+    for (i = 0; i < N; i++)
+    {
+        if (a[i] != 2.0)
+            abort ();
+
+        if (b[i] != 2.0)
+            abort ();
+    }
+
+    if (acc_is_present (a, (N * sizeof (float))))
+      abort ();
+
+    if (acc_is_present (b, (N * sizeof (float))))
+      abort ();
+
+    if (!acc_is_present (c, (N * sizeof (float))))
+      abort ();
+
+    d = (float *) acc_deviceptr (c);
+
+    acc_unmap_data (c);
+
+    acc_free (d);
+
+    return 0;
+}
+/* { dg-shouldfail "libgomp: \[\h+,\d+\] is not mapped" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/collapse-1.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/collapse-1.c
new file mode 100644
index 00000000000..80fed6ca550
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/collapse-1.c
@@ -0,0 +1,31 @@
+/* { dg-do run } */
+
+#include <string.h>
+#include <stdlib.h>
+
+int
+main (void)
+{
+  int i, j, k, l = 0;
+  int a[3][3][3];
+
+  memset (a, '\0', sizeof (a));
+  #pragma acc parallel
+  #pragma acc loop collapse(4 - 1)
+    for (i = 0; i < 2; i++)
+      for (j = 0; j < 2; j++)
+	for (k = 0; k < 2; k++)
+	  a[i][j][k] = i + j * 4 + k * 16;
+  #pragma acc parallel
+    {
+      #pragma acc loop collapse(2) reduction(|:l)
+	for (i = 0; i < 2; i++)
+	  for (j = 0; j < 2; j++)
+	    for (k = 0; k < 2; k++)
+	      if (a[i][j][k] != i + j * 4 + k * 16)
+		l = 1;
+    }
+  if (l)
+    abort ();
+  return 0;
+}
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/collapse-2.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/collapse-2.c
new file mode 100644
index 00000000000..44a77f7665d
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/collapse-2.c
@@ -0,0 +1,37 @@
+/* { dg-do run } */
+
+#include <stdlib.h>
+
+int
+main (void)
+{
+  int i, j, k, l = 0, f = 0, x = 0;
+  int m1 = 4, m2 = -5, m3 = 17;
+
+  #pragma acc parallel
+  #pragma acc loop collapse(3) reduction(+:l)
+    for (i = -2; i < m1; i++)
+      for (j = m2; j < -2; j++)
+	{
+	  for (k = 13; k < m3; k++)
+	    {
+              if ((i + 2) * 12 + (j + 5) * 4 + (k - 13) !=  9 + f++)
+		l++;
+	    }
+	}
+
+    for (i = -2; i < m1; i++)
+      for (j = m2; j < -2; j++)
+	{
+	  for (k = 13; k < m3; k++)
+	    {
+              if ((i + 2) * 12 + (j + 5) * 4 + (k - 13) !=  9 + f++)
+		x++;
+	    }
+	}
+
+  if (l != x)
+    abort ();
+
+  return 0;
+}
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/collapse-3.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/collapse-3.c
new file mode 100644
index 00000000000..a5be7287d22
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/collapse-3.c
@@ -0,0 +1,40 @@
+/* { dg-do run } */
+/* { dg-options "-O2" } */
+
+#include <string.h>
+#include <stdlib.h>
+#include <stdio.h>
+
+int
+main (void)
+{
+  int i2, l = 0, r = 0;
+  int a[3][3][3];
+
+  memset (a, '\0', sizeof (a));
+  #pragma acc parallel
+  #pragma acc loop collapse(4 - 1)
+    for (int i = 0; i < 2; i++)
+      for (int j = 0; j < 2; j++)
+	for (int k = 0; k < 2; k++)
+	  a[i][j][k] = i + j * 4 + k * 16;
+#pragma acc parallel
+    {
+      #pragma acc loop collapse(2) reduction(|:l)
+	for (i2 = 0; i2 < 2; i2++)
+	  for (int j = 0; j < 2; j++)
+	    for (int k = 0; k < 2; k++)
+	      if (a[i2][j][k] != i2 + j * 4 + k * 16)
+		l += 1;
+    }
+
+    for (i2 = 0; i2 < 2; i2++)
+      for (int j = 0; j < 2; j++)
+	for (int k = 0; k < 2; k++)
+	  if (a[i2][j][k] != i2 + j * 4 + k * 16)
+	    r += 1;
+
+  if (l != r)
+    abort ();
+  return 0;
+}
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/collapse-4.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/collapse-4.c
new file mode 100644
index 00000000000..52dd4353246
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/collapse-4.c
@@ -0,0 +1,27 @@
+/* { dg-do run } */
+
+#include <string.h>
+
+int
+main (void)
+{
+  int l = 0;
+  int b[3][3];
+  int i, j;
+
+  memset (b, '\0', sizeof (b));
+
+#pragma acc parallel copy(b[0:3][0:3]) copy(l)
+    {
+#pragma acc loop collapse(2) reduction(+:l)
+	for (i = 0; i < 2; i++)
+	  for (j = 0; j < 2; j++)
+	    if (b[i][j] != 16)
+		  l += 1;
+    }
+
+  if (l != 2 * 2)
+    __builtin_abort();
+
+  return 0;
+}
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/context-1.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/context-1.c
new file mode 100644
index 00000000000..dabc7063c68
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/context-1.c
@@ -0,0 +1,213 @@
+/* { dg-do run { target openacc_nvidia_accel_selected } } */
+/* { dg-additional-options "-lcuda -lcublas -lcudart" } */
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <cuda.h>
+#include <cuda_runtime_api.h>
+#include <cublas_v2.h>
+#include <openacc.h>
+
+void
+saxpy (int n, float a, float *x, float *y)
+{
+    int i;
+
+    for (i = 0; i < n; i++)
+    {
+        y[i] = a * x[i] + y[i];
+    }
+}
+
+void
+context_check (CUcontext ctx1)
+{
+    CUcontext ctx2, ctx3;
+    CUresult r;
+
+    r = cuCtxGetCurrent (&ctx2);
+    if (r != CUDA_SUCCESS)
+    {
+        fprintf (stderr, "cuCtxGetCurrent failed: %d\n", r);
+        exit (EXIT_FAILURE);
+    }
+
+    if (ctx1 != ctx2)
+    {
+        fprintf (stderr, "new context established\n");
+        exit (EXIT_FAILURE);
+    }
+
+    ctx3 = (CUcontext) acc_get_current_cuda_context ();
+
+    if (ctx1 != ctx3)
+    {
+        fprintf (stderr, "acc_get_current_cuda_context returned wrong value\n");
+        exit (EXIT_FAILURE);
+    }
+
+    return;
+}
+
+int
+main (int argc, char **argv)
+{
+    cublasStatus_t s;
+    cudaError_t e;
+    cublasHandle_t h;
+    CUcontext pctx, ctx;
+    CUresult r;
+    int dev;
+    int i;
+    const int N = 256;
+    float *h_X, *h_Y1, *h_Y2;
+    float *d_X,*d_Y;
+    float alpha = 2.0f;
+    float error_norm;
+    float ref_norm;
+
+    /* Test 1 - cuBLAS creates, OpenACC shares.  */
+
+    s = cublasCreate (&h);
+    if (s != CUBLAS_STATUS_SUCCESS)
+    {
+        fprintf (stderr, "cublasCreate failed: %d\n", s);
+        exit (EXIT_FAILURE);
+    }
+
+    r = cuCtxGetCurrent (&pctx);
+    if (r != CUDA_SUCCESS)
+    {
+        fprintf (stderr, "cuCtxGetCurrent failed: %d\n", r);
+        exit (EXIT_FAILURE);
+    }
+
+    e = cudaGetDevice (&dev);
+    if (e != cudaSuccess)
+    {
+        fprintf (stderr, "cudaGetDevice failed: %d\n", e);
+        exit (EXIT_FAILURE);
+    }
+
+    acc_set_device_num (dev, acc_device_nvidia);
+
+    h_X = (float *) malloc (N * sizeof (float));
+    if (!h_X)
+    {
+        fprintf (stderr, "malloc failed: for h_X\n");
+        exit (EXIT_FAILURE);
+    }
+
+    h_Y1 = (float *) malloc (N * sizeof (float));
+    if (!h_Y1)
+    {
+        fprintf (stderr, "malloc failed: for h_Y1\n");
+        exit (EXIT_FAILURE);
+    }
+
+    h_Y2 = (float *) malloc (N * sizeof (float));
+    if (!h_Y2)
+    {
+        fprintf (stderr, "malloc failed: for h_Y2\n");
+        exit (EXIT_FAILURE);
+    }
+
+    for (i = 0; i < N; i++)
+    {
+        h_X[i] = rand () / (float) RAND_MAX;
+        h_Y2[i] = h_Y1[i] = rand () / (float) RAND_MAX;
+    }
+
+    d_X = (float *) acc_copyin (&h_X[0], N * sizeof (float));
+    if (d_X == NULL)
+    {
+        fprintf (stderr, "copyin error h_X\n");
+        exit (EXIT_FAILURE);
+    }
+
+    context_check (pctx);
+
+    d_Y = (float *) acc_copyin (&h_Y1[0], N * sizeof (float));
+    if (d_Y == NULL)
+    {
+        fprintf (stderr, "copyin error h_Y1\n");
+        exit (EXIT_FAILURE);
+    }
+
+    context_check (pctx);
+
+    s = cublasSaxpy (h, N, &alpha, d_X, 1, d_Y, 1);
+    if (s != CUBLAS_STATUS_SUCCESS)
+    {
+        fprintf (stderr, "cublasSaxpy failed: %d\n", s);
+        exit (EXIT_FAILURE);
+    }
+
+    context_check (pctx);
+
+    acc_memcpy_from_device (&h_Y1[0], d_Y, N * sizeof (float));
+
+    context_check (pctx);
+
+    saxpy (N, alpha, h_X, h_Y2);
+
+    error_norm = 0;
+    ref_norm = 0;
+
+    for (i = 0; i < N; ++i)
+    {
+        float diff;
+
+        diff = h_Y1[i] - h_Y2[i];
+        error_norm += diff * diff;
+        ref_norm += h_Y2[i] * h_Y2[i];
+    }
+
+    error_norm = (float) sqrt ((double) error_norm);
+    ref_norm = (float) sqrt ((double) ref_norm);
+
+    if ((fabs (ref_norm) < 1e-7) || ((error_norm / ref_norm) >= 1e-6f))
+    {
+        fprintf (stderr, "math error\n");
+        exit (EXIT_FAILURE);
+    }
+
+    free (h_X);
+    free (h_Y1);
+    free (h_Y2);
+
+    acc_free (d_X);
+    acc_free (d_Y);
+
+    context_check (pctx);
+
+    s = cublasDestroy (h);
+    if (s != CUBLAS_STATUS_SUCCESS)
+    {
+        fprintf (stderr, "cublasDestroy failed: %d\n", s);
+        exit (EXIT_FAILURE);
+    }
+
+    acc_shutdown (acc_device_nvidia);
+
+    r = cuCtxGetCurrent (&ctx);
+    if (r != CUDA_SUCCESS)
+    {
+        fprintf (stderr, "cuCtxGetCurrent failed: %d\n", r);
+        exit (EXIT_FAILURE);
+    }
+
+    if (!ctx)
+    {
+        fprintf (stderr, "Expected context\n");
+        exit (EXIT_FAILURE);
+    }
+
+    if (pctx != ctx)
+    {
+        fprintf (stderr, "Unexpected new context\n");
+        exit (EXIT_FAILURE);
+    }
+
+    return EXIT_SUCCESS;
+}
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/context-2.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/context-2.c
new file mode 100644
index 00000000000..6a52f746dcb
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/context-2.c
@@ -0,0 +1,223 @@
+/* { dg-do run { target openacc_nvidia_accel_selected } } */
+/* { dg-additional-options "-lcuda -lcublas -lcudart" } */
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <cuda.h>
+#include <cuda_runtime_api.h>
+#include <cublas_v2.h>
+#include <openacc.h>
+
+void
+saxpy (int n, float a, float *x, float *y)
+{
+    int i;
+
+    for (i = 0; i < n; i++)
+    {
+        y[i] = a * x[i] + y[i];
+    }
+}
+
+void
+context_check (CUcontext ctx1)
+{
+    CUcontext ctx2, ctx3;
+    CUresult r;
+
+    r = cuCtxGetCurrent (&ctx2);
+    if (r != CUDA_SUCCESS)
+    {
+        fprintf (stderr, "cuCtxGetCurrent failed: %d\n", r);
+        exit (EXIT_FAILURE);
+    }
+
+    if (ctx1 != ctx2)
+    {
+        fprintf (stderr, "new context established\n");
+        exit (EXIT_FAILURE);
+    }
+
+    ctx3 = (CUcontext) acc_get_current_cuda_context ();
+
+    if (ctx1 != ctx3)
+    {
+        fprintf (stderr, "acc_get_current_cuda_context returned wrong value\n");
+        exit (EXIT_FAILURE);
+    }
+
+    return;
+}
+
+int
+main (int argc, char **argv)
+{
+    cublasStatus_t s;
+    cudaError_t e;
+    cublasHandle_t h;
+    CUcontext pctx, ctx;
+    CUresult r;
+    int dev;
+    int i;
+    const int N = 256;
+    float *h_X, *h_Y1, *h_Y2;
+    float *d_X,*d_Y;
+    float alpha = 2.0f;
+    float error_norm;
+    float ref_norm;
+
+    /* Test 2 - cuBLAS creates, OpenACC shares.  */
+
+    s = cublasCreate (&h);
+    if (s != CUBLAS_STATUS_SUCCESS)
+    {
+        fprintf (stderr, "cublasCreate failed: %d\n", s);
+        exit (EXIT_FAILURE);
+    }
+
+    r = cuCtxGetCurrent (&pctx);
+    if (r != CUDA_SUCCESS)
+    {
+        fprintf (stderr, "cuCtxGetCurrent failed: %d\n", r);
+        exit (EXIT_FAILURE);
+    }
+
+    e = cudaGetDevice (&dev);
+    if (e != cudaSuccess)
+    {
+        fprintf (stderr, "cudaGetDevice failed: %d\n", e);
+        exit (EXIT_FAILURE);
+    }
+
+    acc_set_device_num (dev, acc_device_nvidia);
+
+    h_X = (float *) malloc (N * sizeof (float));
+    if (h_X == 0)
+    {
+        fprintf (stderr, "malloc failed: for h_X\n");
+        exit (EXIT_FAILURE);
+    }
+
+    h_Y1 = (float *) malloc (N * sizeof (float));
+    if (h_Y1 == 0)
+    {
+        fprintf (stderr, "malloc failed: for h_Y1\n");
+        exit (EXIT_FAILURE);
+    }
+
+    h_Y2 = (float *) malloc (N * sizeof (float));
+    if (h_Y2 == 0)
+    {
+        fprintf (stderr, "malloc failed: for h_Y2\n");
+        exit (EXIT_FAILURE);
+    }
+
+    for (i = 0; i < N; i++)
+    {
+        h_X[i] = rand () / (float) RAND_MAX;
+        h_Y2[i] = h_Y1[i] = rand () / (float) RAND_MAX;
+    }
+
+    d_X = (float *) acc_copyin (&h_X[0], N * sizeof (float));
+    if (d_X == NULL)
+    {
+        fprintf (stderr, "copyin error h_X\n");
+        exit (EXIT_FAILURE);
+    }
+
+    context_check (pctx);
+
+    d_Y = (float *) acc_copyin (&h_Y1[0], N * sizeof (float));
+    if (d_Y == NULL)
+    {
+        fprintf (stderr, "copyin error h_Y1\n");
+        exit (EXIT_FAILURE);
+    }
+
+    context_check (pctx);
+
+    s = cublasSaxpy (h, N, &alpha, d_X, 1, d_Y, 1);
+    if (s != CUBLAS_STATUS_SUCCESS)
+    {
+        fprintf (stderr, "cublasSaxpy failed: %d\n", s);
+        exit (EXIT_FAILURE);
+    }
+
+    context_check (pctx);
+
+    acc_memcpy_from_device (&h_Y1[0], d_Y, N * sizeof (float));
+
+    context_check (pctx);
+
+#pragma acc parallel present (h_X[0:N]), copy (h_Y2[0:N]) copyin (alpha)
+    {
+        int i;
+
+        for (i = 0; i < N; i++)
+        {
+            h_Y2[i] = alpha * h_X[i] + h_Y2[i];
+        }
+    }
+
+    context_check (pctx);
+
+    error_norm = 0;
+    ref_norm = 0;
+
+    for (i = 0; i < N; ++i)
+    {
+        float diff;
+
+        diff = h_Y1[i] - h_Y2[i];
+        error_norm += diff * diff;
+        ref_norm += h_Y2[i] * h_Y2[i];
+    }
+
+    error_norm = (float) sqrt ((double) error_norm);
+    ref_norm = (float) sqrt ((double) ref_norm);
+
+    if ((fabs (ref_norm) < 1e-7) || ((error_norm / ref_norm) >= 1e-6f))
+    {
+        fprintf (stderr, "math error\n");
+        exit (EXIT_FAILURE);
+    }
+
+    free (h_X);
+    free (h_Y1);
+    free (h_Y2);
+
+    acc_free (d_X);
+    acc_free (d_Y);
+
+    context_check (pctx);
+
+    s = cublasDestroy (h);
+    if (s != CUBLAS_STATUS_SUCCESS)
+    {
+        fprintf (stderr, "cublasDestroy failed: %d\n", s);
+        exit (EXIT_FAILURE);
+    }
+
+    acc_shutdown (acc_device_nvidia);
+
+    r = cuCtxGetCurrent (&ctx);
+    if (r != CUDA_SUCCESS)
+    {
+        fprintf (stderr, "cuCtxGetCurrent failed: %d\n", r);
+        exit (EXIT_FAILURE);
+    }
+
+    if (!ctx)
+    {
+        fprintf (stderr, "Expected context\n");
+        exit (EXIT_FAILURE);
+    }
+
+    if (pctx != ctx)
+    {
+        fprintf (stderr, "Unexpected new context\n");
+        exit (EXIT_FAILURE);
+    }
+
+    return EXIT_SUCCESS;
+}
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/context-3.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/context-3.c
new file mode 100644
index 00000000000..ccd276cd98f
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/context-3.c
@@ -0,0 +1,200 @@
+/* { dg-do run { target openacc_nvidia_accel_selected } } */
+/* { dg-additional-options "-lcuda -lcublas -lcudart" } */
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <cuda.h>
+#include <cuda_runtime_api.h>
+#include <cublas_v2.h>
+#include <openacc.h>
+
+void
+saxpy (int n, float a, float *x, float *y)
+{
+    int i;
+
+    for (i = 0; i < n; i++)
+    {
+        y[i] = a * x[i] + y[i];
+    }
+}
+
+void
+context_check (CUcontext ctx1)
+{
+    CUcontext ctx2, ctx3;
+    CUresult r;
+
+    r = cuCtxGetCurrent (&ctx2);
+    if (r != CUDA_SUCCESS)
+    {
+        fprintf (stderr, "cuCtxGetCurrent failed: %d\n", r);
+        exit (EXIT_FAILURE);
+    }
+
+    if (ctx1 != ctx2)
+    {
+        fprintf (stderr, "new context established\n");
+        exit (EXIT_FAILURE);
+    }
+
+    ctx3 = (CUcontext) acc_get_current_cuda_context ();
+
+    if (ctx1 != ctx3)
+    {
+        fprintf (stderr, "acc_get_current_cuda_context returned wrong value\n");
+        exit (EXIT_FAILURE);
+    }
+
+    return;
+}
+
+int
+main (int argc, char **argv)
+{
+    cublasStatus_t s;
+    cublasHandle_t h;
+    CUcontext pctx;
+    CUresult r;
+    int i;
+    const int N = 256;
+    float *h_X, *h_Y1, *h_Y2;
+    float *d_X,*d_Y;
+    float alpha = 2.0f;
+    float error_norm;
+    float ref_norm;
+
+    /* Test 3 - OpenACC creates, cuBLAS shares.  */
+
+    acc_set_device_num (0, acc_device_nvidia);
+
+    r = cuCtxGetCurrent (&pctx);
+    if (r != CUDA_SUCCESS)
+    {
+        fprintf (stderr, "cuCtxGetCurrent failed: %d\n", r);
+        exit (EXIT_FAILURE);
+    }
+
+    h_X = (float *) malloc (N * sizeof (float));
+    if (h_X == 0)
+    {
+        fprintf (stderr, "malloc failed: for h_X\n");
+        exit (EXIT_FAILURE);
+    }
+
+    h_Y1 = (float *) malloc (N * sizeof (float));
+    if (h_Y1 == 0)
+    {
+        fprintf (stderr, "malloc failed: for h_Y1\n");
+        exit (EXIT_FAILURE);
+    }
+
+    h_Y2 = (float *) malloc (N * sizeof (float));
+    if (h_Y2 == 0)
+    {
+        fprintf (stderr, "malloc failed: for h_Y2\n");
+        exit (EXIT_FAILURE);
+    }
+
+    for (i = 0; i < N; i++)
+    {
+        h_X[i] = rand () / (float) RAND_MAX;
+        h_Y2[i] = h_Y1[i] = rand () / (float) RAND_MAX;
+    }
+
+    d_X = (float *) acc_copyin (&h_X[0], N * sizeof (float));
+    if (d_X == NULL)
+    {
+        fprintf (stderr, "copyin error h_X\n");
+        exit (EXIT_FAILURE);
+    }
+
+    d_Y = (float *) acc_copyin (&h_Y1[0], N * sizeof (float));
+    if (d_Y == NULL)
+    {
+        fprintf (stderr, "copyin error h_Y1\n");
+        exit (EXIT_FAILURE);
+    }
+
+    context_check (pctx);
+
+    s = cublasCreate (&h);
+    if (s != CUBLAS_STATUS_SUCCESS)
+    {
+        fprintf (stderr, "cublasCreate failed: %d\n", s);
+        exit (EXIT_FAILURE);
+    }
+
+    context_check (pctx);
+
+    s = cublasSaxpy (h, N, &alpha, d_X, 1, d_Y, 1);
+    if (s != CUBLAS_STATUS_SUCCESS)
+    {
+        fprintf (stderr, "cublasSaxpy failed: %d\n", s);
+        exit (EXIT_FAILURE);
+    }
+
+    context_check (pctx);
+
+    acc_memcpy_from_device (&h_Y1[0], d_Y, N * sizeof (float));
+
+    context_check (pctx);
+
+    saxpy (N, alpha, h_X, h_Y2);
+
+    error_norm = 0;
+    ref_norm = 0;
+
+    for (i = 0; i < N; ++i)
+    {
+        float diff;
+
+        diff = h_Y1[i] - h_Y2[i];
+        error_norm += diff * diff;
+        ref_norm += h_Y2[i] * h_Y2[i];
+    }
+
+    error_norm = (float) sqrt ((double) error_norm);
+    ref_norm = (float) sqrt ((double) ref_norm);
+
+    if ((fabs (ref_norm) < 1e-7) || ((error_norm / ref_norm) >= 1e-6f))
+    {
+        fprintf (stderr, "math error\n");
+        exit (EXIT_FAILURE);
+    }
+
+    free (h_X);
+    free (h_Y1);
+    free (h_Y2);
+
+    acc_free (d_X);
+    acc_free (d_Y);
+
+    context_check (pctx);
+
+    s = cublasDestroy (h);
+    if (s != CUBLAS_STATUS_SUCCESS)
+    {
+        fprintf (stderr, "cublasDestroy failed: %d\n", s);
+        exit (EXIT_FAILURE);
+    }
+
+    context_check (pctx);
+
+    acc_shutdown (acc_device_nvidia);
+
+    r = cuCtxGetCurrent (&pctx);
+    if (r != CUDA_SUCCESS)
+    {
+        fprintf (stderr, "cuCtxGetCurrent failed: %d\n", r);
+        exit (EXIT_FAILURE);
+    }
+
+    if (pctx)
+    {
+        fprintf (stderr, "Unexpected context\n");
+        exit (EXIT_FAILURE);
+    }
+
+    return EXIT_SUCCESS;
+}
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/context-4.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/context-4.c
new file mode 100644
index 00000000000..71365e8ed32
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/context-4.c
@@ -0,0 +1,213 @@
+/* { dg-do run { target openacc_nvidia_accel_selected } } */
+/* { dg-additional-options "-lcuda -lcublas -lcudart" } */
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <cuda.h>
+#include <cuda_runtime_api.h>
+#include <cublas_v2.h>
+#include <openacc.h>
+
+void
+saxpy (int n, float a, float *x, float *y)
+{
+    int i;
+
+    for (i = 0; i < n; i++)
+    {
+        y[i] = a * x[i] + y[i];
+    }
+}
+
+void
+context_check (CUcontext ctx1)
+{
+    CUcontext ctx2, ctx3;
+    CUresult r;
+
+    r = cuCtxGetCurrent (&ctx2);
+    if (r != CUDA_SUCCESS)
+    {
+        fprintf (stderr, "cuCtxGetCurrent failed: %d\n", r);
+        exit (EXIT_FAILURE);
+    }
+
+    if (ctx1 != ctx2)
+    {
+        fprintf (stderr, "new context established\n");
+        exit (EXIT_FAILURE);
+    }
+
+    ctx3 = (CUcontext) acc_get_current_cuda_context ();
+
+    if (ctx1 != ctx3)
+    {
+        fprintf (stderr, "acc_get_current_cuda_context returned wrong value\n");
+        exit (EXIT_FAILURE);
+    }
+
+    return;
+}
+
+int
+main (int argc, char **argv)
+{
+    cublasStatus_t s;
+    cublasHandle_t h;
+    CUcontext pctx;
+    CUresult r;
+    int i;
+    const int N = 256;
+    float *h_X, *h_Y1, *h_Y2;
+    float *d_X,*d_Y;
+    float alpha = 2.0f;
+    float error_norm;
+    float ref_norm;
+
+    /* Test 4 - OpenACC creates, cuBLAS shares.  */
+
+    acc_set_device_num (0, acc_device_nvidia);
+
+    r = cuCtxGetCurrent (&pctx);
+    if (r != CUDA_SUCCESS)
+    {
+        fprintf (stderr, "cuCtxGetCurrent failed: %d\n", r);
+        exit (EXIT_FAILURE);
+    }
+
+    h_X = (float *) malloc (N * sizeof (float));
+    if (h_X == 0)
+    {
+        fprintf (stderr, "malloc failed: for h_X\n");
+        exit (EXIT_FAILURE);
+    }
+
+    h_Y1 = (float *) malloc (N * sizeof (float));
+    if (h_Y1 == 0)
+    {
+        fprintf (stderr, "malloc failed: for h_Y1\n");
+        exit (EXIT_FAILURE);
+    }
+
+    h_Y2 = (float *) malloc (N * sizeof (float));
+    if (h_Y2 == 0)
+    {
+        fprintf (stderr, "malloc failed: for h_Y2\n");
+        exit (EXIT_FAILURE);
+    }
+
+    for (i = 0; i < N; i++)
+    {
+        h_X[i] = rand () / (float) RAND_MAX;
+        h_Y2[i] = h_Y1[i] = rand () / (float) RAND_MAX;
+    }
+
+#pragma acc parallel copyin (h_X[0:N]), copy (h_Y2[0:N]) copy (alpha)
+    {
+        int i;
+
+        for (i = 0; i < N; i++)
+        {
+            h_Y2[i] = alpha * h_X[i] + h_Y2[i];
+        }
+    }
+
+    r = cuCtxGetCurrent (&pctx);
+    if (r != CUDA_SUCCESS)
+    {
+        fprintf (stderr, "cuCtxGetCurrent failed: %d\n", r);
+        exit (EXIT_FAILURE);
+    }
+
+    d_X = (float *) acc_copyin (&h_X[0], N * sizeof (float));
+    if (d_X == NULL)
+    {
+        fprintf (stderr, "copyin error h_Y1\n");
+        exit (EXIT_FAILURE);
+    }
+
+    d_Y = (float *) acc_copyin (&h_Y1[0], N * sizeof (float));
+    if (d_Y == NULL)
+    {
+        fprintf (stderr, "copyin error h_Y1\n");
+        exit (EXIT_FAILURE);
+    }
+
+    s = cublasCreate (&h);
+    if (s != CUBLAS_STATUS_SUCCESS)
+    {
+        fprintf (stderr, "cublasCreate failed: %d\n", s);
+        exit (EXIT_FAILURE);
+    }
+
+    context_check (pctx);
+
+    s = cublasSaxpy (h, N, &alpha, d_X, 1, d_Y, 1);
+    if (s != CUBLAS_STATUS_SUCCESS)
+    {
+        fprintf (stderr, "cublasSaxpy failed: %d\n", s);
+        exit (EXIT_FAILURE);
+    }
+
+    context_check (pctx);
+
+    acc_memcpy_from_device (&h_Y1[0], d_Y, N * sizeof (float));
+
+    context_check (pctx);
+
+    error_norm = 0;
+    ref_norm = 0;
+
+    for (i = 0; i < N; ++i)
+    {
+        float diff;
+
+        diff = h_Y1[i] - h_Y2[i];
+        error_norm += diff * diff;
+        ref_norm += h_Y2[i] * h_Y2[i];
+    }
+
+    error_norm = (float) sqrt ((double) error_norm);
+    ref_norm = (float) sqrt ((double) ref_norm);
+
+    if ((fabs (ref_norm) < 1e-7) || ((error_norm / ref_norm) >= 1e-6f))
+    {
+        fprintf (stderr, "math error\n");
+        exit (EXIT_FAILURE);
+    }
+
+    free (h_X);
+    free (h_Y1);
+    free (h_Y2);
+
+    acc_free (d_X);
+    acc_free (d_Y);
+
+    context_check (pctx);
+
+    s = cublasDestroy (h);
+    if (s != CUBLAS_STATUS_SUCCESS)
+    {
+        fprintf (stderr, "cublasDestroy failed: %d\n", s);
+        exit (EXIT_FAILURE);
+    }
+
+    context_check (pctx);
+
+    acc_shutdown (acc_device_nvidia);
+
+    r = cuCtxGetCurrent (&pctx);
+    if (r != CUDA_SUCCESS)
+    {
+        fprintf (stderr, "cuCtxGetCurrent failed: %d\n", r);
+        exit (EXIT_FAILURE);
+    }
+
+    if (pctx)
+    {
+        fprintf (stderr, "Unexpected context\n");
+        exit (EXIT_FAILURE);
+    }
+
+    return EXIT_SUCCESS;
+}
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/data-1.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/data-1.c
new file mode 100644
index 00000000000..e7564cc12b2
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/data-1.c
@@ -0,0 +1,188 @@
+/* { dg-do run } */
+
+#include <stdlib.h>
+#include <openacc.h>
+
+int i;
+
+int
+is_mapped (void *p, size_t n)
+{
+#if ACC_MEM_SHARED
+  return 1;
+#else
+  return acc_is_present (p, n);
+#endif
+}
+
+int main(void)
+{
+  int j;
+
+  i = -1;
+  j = -2;
+#pragma acc data copyin (i, j)
+  {
+    if (!is_mapped (&i, sizeof (i)) || !is_mapped (&j, sizeof (j)))
+      abort ();
+    if (i != -1 || j != -2)
+      abort ();
+    i = 2;
+    j = 1;
+    if (i != 2 || j != 1)
+      abort ();
+  }
+  if (i != 2 || j != 1)
+    abort ();
+
+  i = -1;
+  j = -2;
+#pragma acc data copyout (i, j)
+  {
+    if (!is_mapped (&i, sizeof (i)) || !is_mapped (&j, sizeof (j)))
+      abort ();
+    if (i != -1 || j != -2)
+      abort ();
+    i = 2;
+    j = 1;
+    if (i != 2 || j != 1)
+      abort ();
+
+#pragma acc parallel present (i, j)
+    {
+      i = 4;
+      j = 2;
+    }
+  }
+  if (i != 4 || j != 2)
+    abort ();
+
+  i = -1;
+  j = -2;
+#pragma acc data create (i, j)
+  {
+    if (!is_mapped (&i, sizeof (i)) || !is_mapped (&j, sizeof (j)))
+      abort ();
+    if (i != -1 || j != -2)
+      abort ();
+    i = 2;
+    j = 1;
+    if (i != 2 || j != 1)
+      abort ();
+  }
+  if (i != 2 || j != 1)
+    abort ();
+
+  i = -1;
+  j = -2;
+#pragma acc data present_or_copyin (i, j)
+  {
+    if (!is_mapped (&i, sizeof (i)) || !is_mapped (&j, sizeof (j)))
+      abort ();
+    if (i != -1 || j != -2)
+      abort ();
+    i = 2;
+    j = 1;
+    if (i != 2 || j != 1)
+      abort ();
+  }
+  if (i != 2 || j != 1)
+    abort ();
+
+  i = -1;
+  j = -2;
+#pragma acc data present_or_copyout (i, j)
+  {
+    if (!is_mapped (&i, sizeof (i)) || !is_mapped (&j, sizeof (j)))
+      abort ();
+    if (i != -1 || j != -2)
+      abort ();
+    i = 2;
+    j = 1;
+    if (i != 2 || j != 1)
+      abort ();
+
+#pragma acc parallel present (i, j)
+    {
+      i = 4;
+      j = 2;
+    }
+  }
+  if (i != 4 || j != 2)
+    abort ();
+
+  i = -1;
+  j = -2;
+#pragma acc data present_or_copy (i, j)
+  {
+    if (!is_mapped (&i, sizeof (i)) || !is_mapped (&j, sizeof (j)))
+      abort ();
+    if (i != -1 || j != -2)
+      abort ();
+    i = 2;
+    j = 1;
+    if (i != 2 || j != 1)
+      abort ();
+  }
+#if ACC_MEM_SHARED
+  if (i != 2 || j != 1)
+    abort ();
+#else
+  if (i != -1 || j != -2)
+    abort ();
+#endif
+
+  i = -1;
+  j = -2;
+#pragma acc data present_or_create (i, j)
+  {
+    if (!is_mapped (&i, sizeof (i)) || !is_mapped (&j, sizeof (j)))
+      abort ();
+    i = 2;
+    j = 1;
+    if (i != 2 || j != 1)
+      abort ();
+  }
+
+  if (i != 2 || j != 1)
+    abort ();
+
+  i = -1;
+  j = -2;
+#pragma acc data copyin (i, j)
+  {
+#pragma acc data present (i, j)
+    {
+      if (!is_mapped (&i, sizeof (i)) || !is_mapped (&j, sizeof (j)))
+        abort ();
+      if (i != -1 || j != -2)
+        abort ();
+      i = 2;
+      j = 1;
+      if (i != 2 || j != 1)
+        abort ();
+    }
+  }
+  if (i != 2 || j != 1)
+    abort ();
+
+  i = -1;
+  j = -2;
+#pragma acc data
+  {
+#if !ACC_MEM_SHARED
+    if (is_mapped (&i, sizeof (i)) || is_mapped (&j, sizeof (j)))
+      abort ();
+#endif
+    if (i != -1 || j != -2)
+      abort ();
+    i = 2;
+    j = 1;
+    if (i != 2 || j != 1)
+      abort ();
+  }
+  if (i != 2 || j != 1)
+    abort ();
+
+  return 0;
+}
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/data-2.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/data-2.c
new file mode 100644
index 00000000000..f867a66e5b7
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/data-2.c
@@ -0,0 +1,162 @@
+/* { dg-do run } */
+
+#include <stdlib.h>
+
+int
+main (int argc, char **argv)
+{
+  int N = 128; //1024 * 1024;
+  float *a, *b, *c, *d, *e;
+  int i;
+  int nbytes;
+
+  nbytes = N * sizeof (float);
+
+  a = (float *) malloc (nbytes);
+  b = (float *) malloc (nbytes);
+  c = (float *) malloc (nbytes);
+  d = (float *) malloc (nbytes);
+  e = (float *) malloc (nbytes);
+
+  for (i = 0; i < N; i++)
+    {
+      a[i] = 3.0;
+      b[i] = 0.0;
+    }
+
+#pragma acc enter data copyin (a[0:N]) copyin (b[0:N]) copyin (N) async
+#pragma acc parallel async wait
+#pragma acc loop
+  for (i = 0; i < N; i++)
+    b[i] = a[i];
+
+#pragma acc exit data copyout (a[0:N]) copyout (b[0:N]) wait async
+#pragma acc wait
+
+  for (i = 0; i < N; i++)
+    {
+      if (a[i] != 3.0)
+	abort ();
+
+      if (b[i] != 3.0)
+	abort ();
+    }
+
+  for (i = 0; i < N; i++)
+    {
+      a[i] = 2.0;
+      b[i] = 0.0;
+    }
+
+#pragma acc enter data copyin (a[0:N]) copyin (b[0:N]) copyin (N) async (1)
+#pragma acc parallel async (1)
+#pragma acc loop
+  for (i = 0; i < N; i++)
+    b[i] = a[i];
+
+#pragma acc exit data copyout (a[0:N]) copyout (b[0:N]) wait (1) async (1)
+#pragma acc wait (1)
+
+  for (i = 0; i < N; i++)
+    {
+      if (a[i] != 2.0)
+	abort ();
+
+      if (b[i] != 2.0)
+	abort ();
+    }
+
+  for (i = 0; i < N; i++)
+    {
+      a[i] = 3.0;
+      b[i] = 0.0;
+      c[i] = 0.0;
+      d[i] = 0.0;
+    }
+
+#pragma acc enter data copyin (a[0:N]) copyin (b[0:N]) copyin (c[0:N]) copyin (d[0:N]) copyin (N) async (1)
+
+#pragma acc parallel async (1) wait (1)
+#pragma acc loop
+  for (i = 0; i < N; i++)
+    b[i] = (a[i] * a[i] * a[i]) / a[i];
+
+#pragma acc parallel async (2) wait (1)
+#pragma acc loop
+  for (i = 0; i < N; i++)
+    c[i] = (a[i] + a[i] + a[i] + a[i]) / a[i];
+
+#pragma acc parallel async (3) wait (1)
+#pragma acc loop
+  for (i = 0; i < N; i++)
+    d[i] = ((a[i] * a[i] + a[i]) / a[i]) - a[i];
+
+#pragma acc exit data copyout (a[0:N]) copyout (b[0:N]) copyout (c[0:N]) copyout (d[0:N]) wait (1, 2, 3) async (1)
+#pragma acc wait (1)
+
+  for (i = 0; i < N; i++)
+    {
+      if (a[i] != 3.0)
+	abort ();
+
+      if (b[i] != 9.0)
+	abort ();
+
+      if (c[i] != 4.0)
+	abort ();
+
+      if (d[i] != 1.0)
+	abort ();
+    }
+
+  for (i = 0; i < N; i++)
+    {
+      a[i] = 2.0;
+      b[i] = 0.0;
+      c[i] = 0.0;
+      d[i] = 0.0;
+      e[i] = 0.0;
+    }
+
+#pragma acc enter data copyin (a[0:N]) copyin (b[0:N]) copyin (c[0:N]) copyin (d[0:N]) copyin (e[0:N]) copyin (N) async (1)
+
+#pragma acc parallel async (1) wait (1)
+  for (int ii = 0; ii < N; ii++)
+    b[ii] = (a[ii] * a[ii] * a[ii]) / a[ii];
+
+#pragma acc parallel async (2) wait (1)
+  for (int ii = 0; ii < N; ii++)
+    c[ii] = (a[ii] + a[ii] + a[ii] + a[ii]) / a[ii];
+
+#pragma acc parallel async (3) wait (1)
+  for (int ii = 0; ii < N; ii++)
+    d[ii] = ((a[ii] * a[ii] + a[ii]) / a[ii]) - a[ii];
+
+#pragma acc parallel wait (1) async (4)
+  for (int ii = 0; ii < N; ii++)
+    e[ii] = a[ii] + b[ii] + c[ii] + d[ii];
+
+#pragma acc exit data copyout (a[0:N]) copyout (b[0:N]) copyout (c[0:N]) copyout (d[0:N]) copyout (e[0:N]) wait (1, 2, 3, 4) async (1)
+#pragma acc wait (1)
+
+
+  for (i = 0; i < N; i++)
+    {
+      if (a[i] != 2.0)
+	abort ();
+
+      if (b[i] != 4.0)
+	abort ();
+
+      if (c[i] != 4.0)
+	abort ();
+
+      if (d[i] != 1.0)
+	abort ();
+
+      if (e[i] != 11.0)
+	abort ();
+    }
+
+  return 0;
+}
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/data-3.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/data-3.c
new file mode 100644
index 00000000000..747109f6136
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/data-3.c
@@ -0,0 +1,166 @@
+/* { dg-do run } */
+
+#include <stdlib.h>
+
+int
+main (int argc, char **argv)
+{
+  int N = 128; //1024 * 1024;
+  float *a, *b, *c, *d, *e;
+  int i;
+  int nbytes;
+
+  nbytes = N * sizeof (float);
+
+  a = (float *) malloc (nbytes);
+  b = (float *) malloc (nbytes);
+  c = (float *) malloc (nbytes);
+  d = (float *) malloc (nbytes);
+  e = (float *) malloc (nbytes);
+
+  for (i = 0; i < N; i++)
+    {
+      a[i] = 3.0;
+      b[i] = 0.0;
+    }
+
+#pragma acc enter data copyin (a[0:N]) copyin (b[0:N]) copyin (N) async
+#pragma acc parallel async wait
+#pragma acc loop
+  for (i = 0; i < N; i++)
+    b[i] = a[i];
+
+#pragma acc update host (a[0:N], b[0:N]) async wait
+#pragma acc wait
+
+  for (i = 0; i < N; i++)
+    {
+      if (a[i] != 3.0)
+	abort ();
+
+      if (b[i] != 3.0)
+	abort ();
+    }
+
+  for (i = 0; i < N; i++)
+    {
+      a[i] = 2.0;
+      b[i] = 0.0;
+    }
+
+#pragma acc update device (a[0:N], b[0:N]) async (1)
+#pragma acc parallel async (1)
+#pragma acc loop
+  for (i = 0; i < N; i++)
+    b[i] = a[i];
+
+#pragma acc update host (a[0:N], b[0:N]) async (1) wait (1)
+#pragma acc wait (1)
+
+  for (i = 0; i < N; i++)
+    {
+      if (a[i] != 2.0)
+	abort ();
+
+      if (b[i] != 2.0)
+	abort ();
+    }
+
+  for (i = 0; i < N; i++)
+    {
+      a[i] = 3.0;
+      b[i] = 0.0;
+      c[i] = 0.0;
+      d[i] = 0.0;
+    }
+
+#pragma acc update device (a[0:N]) async (1)
+#pragma acc update device (b[0:N]) async (2)
+#pragma acc enter data copyin (c[0:N], d[0:N]) async (3)
+
+#pragma acc parallel async (1) wait (1,2)
+#pragma acc loop
+  for (i = 0; i < N; i++)
+    b[i] = (a[i] * a[i] * a[i]) / a[i];
+
+#pragma acc parallel async (2) wait (1,3)
+#pragma acc loop
+  for (i = 0; i < N; i++)
+    c[i] = (a[i] + a[i] + a[i] + a[i]) / a[i];
+
+#pragma acc parallel async (3) wait (1,3)
+#pragma acc loop
+  for (i = 0; i < N; i++)
+    d[i] = ((a[i] * a[i] + a[i]) / a[i]) - a[i];
+
+#pragma acc update host (a[0:N], b[0:N], c[0:N], d[0:N]) async (1) wait (1,2,3)
+#pragma acc wait (1)
+
+  for (i = 0; i < N; i++)
+    {
+      if (a[i] != 3.0)
+	abort ();
+
+      if (b[i] != 9.0)
+	abort ();
+
+      if (c[i] != 4.0)
+	abort ();
+
+      if (d[i] != 1.0)
+	abort ();
+    }
+
+  for (i = 0; i < N; i++)
+    {
+      a[i] = 2.0;
+      b[i] = 0.0;
+      c[i] = 0.0;
+      d[i] = 0.0;
+      e[i] = 0.0;
+    }
+
+#pragma acc update device (a[0:N], b[0:N], c[0:N], d[0:N]) async (1)
+#pragma acc enter data copyin (e[0:N]) async (5)
+
+#pragma acc parallel async (1) wait (1)
+  for (int ii = 0; ii < N; ii++)
+    b[ii] = (a[ii] * a[ii] * a[ii]) / a[ii];
+
+#pragma acc parallel async (2) wait (1)
+  for (int ii = 0; ii < N; ii++)
+    c[ii] = (a[ii] + a[ii] + a[ii] + a[ii]) / a[ii];
+
+#pragma acc parallel async (3) wait (1)
+  for (int ii = 0; ii < N; ii++)
+    d[ii] = ((a[ii] * a[ii] + a[ii]) / a[ii]) - a[ii];
+
+#pragma acc parallel wait (1,5) async (4)
+  for (int ii = 0; ii < N; ii++)
+    e[ii] = a[ii] + b[ii] + c[ii] + d[ii];
+
+#pragma acc exit data copyout (a[0:N]) copyout (b[0:N]) copyout (c[0:N]) copyout (d[0:N]) copyout (e[0:N]) wait (1, 2, 3, 4) async (1)
+#pragma acc exit data delete (N)
+#pragma acc wait (1)
+
+
+  for (i = 0; i < N; i++)
+    {
+      if (a[i] != 2.0)
+	abort ();
+
+      if (b[i] != 4.0)
+	abort ();
+
+      if (c[i] != 4.0)
+	abort ();
+
+      if (d[i] != 1.0)
+	abort ();
+
+      if (e[i] != 11.0)
+	abort ();
+    }
+
+  return 0;
+}
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/data-already-1.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/data-already-1.c
new file mode 100644
index 00000000000..83c0a42c512
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/data-already-1.c
@@ -0,0 +1,19 @@
+/* { dg-skip-if "" { *-*-* } { "*" } { "-DACC_MEM_SHARED=0" } } */
+
+#include <openacc.h>
+
+int
+main (int argc, char *argv[])
+{
+  int i;
+
+  acc_copyin (&i, sizeof i);
+
+#pragma acc data copy (i)
+  ++i;
+
+  return 0;
+}
+
+/* { dg-shouldfail "" }
+   { dg-output "Trying to map into device .* object when .* is already mapped" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/data-already-2.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/data-already-2.c
new file mode 100644
index 00000000000..137d8ce09da
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/data-already-2.c
@@ -0,0 +1,16 @@
+/* { dg-skip-if "" { *-*-* } { "*" } { "-DACC_MEM_SHARED=0" } } */
+
+int
+main (int argc, char *argv[])
+{
+  int i;
+
+#pragma acc data present_or_copy (i)
+#pragma acc data copyout (i)
+  ++i;
+
+  return 0;
+}
+
+/* { dg-shouldfail "" }
+   { dg-output "Trying to map into device .* object when .* is already mapped" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/data-already-3.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/data-already-3.c
new file mode 100644
index 00000000000..b993b78c5ac
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/data-already-3.c
@@ -0,0 +1,17 @@
+/* { dg-skip-if "" { *-*-* } { "*" } { "-DACC_MEM_SHARED=0" } } */
+
+#include <openacc.h>
+
+int
+main (int argc, char *argv[])
+{
+  int i;
+
+#pragma acc data present_or_copy (i)
+  acc_copyin (&i, sizeof i);
+
+  return 0;
+}
+
+/* { dg-shouldfail "" }
+   { dg-output "already mapped to" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/data-already-4.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/data-already-4.c
new file mode 100644
index 00000000000..82523f42f1a
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/data-already-4.c
@@ -0,0 +1,17 @@
+/* { dg-skip-if "" { *-*-* } { "*" } { "-DACC_MEM_SHARED=0" } } */
+
+#include <openacc.h>
+
+int
+main (int argc, char *argv[])
+{
+  int i;
+
+  acc_present_or_copyin (&i, sizeof i);
+  acc_copyin (&i, sizeof i);
+
+  return 0;
+}
+
+/* { dg-shouldfail "" }
+   { dg-output "already mapped to" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/data-already-5.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/data-already-5.c
new file mode 100644
index 00000000000..4961fe54406
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/data-already-5.c
@@ -0,0 +1,17 @@
+/* { dg-skip-if "" { *-*-* } { "*" } { "-DACC_MEM_SHARED=0" } } */
+
+#include <openacc.h>
+
+int
+main (int argc, char *argv[])
+{
+  int i;
+
+#pragma acc enter data create (i)
+  acc_copyin (&i, sizeof i);
+
+  return 0;
+}
+
+/* { dg-shouldfail "" }
+   { dg-output "already mapped to" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/data-already-6.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/data-already-6.c
new file mode 100644
index 00000000000..77b56a93281
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/data-already-6.c
@@ -0,0 +1,17 @@
+/* { dg-skip-if "" { *-*-* } { "*" } { "-DACC_MEM_SHARED=0" } } */
+
+#include <openacc.h>
+
+int
+main (int argc, char *argv[])
+{
+  int i;
+
+  acc_present_or_copyin (&i, sizeof i);
+#pragma acc enter data create (i)
+
+  return 0;
+}
+
+/* { dg-shouldfail "" }
+   { dg-output "already mapped to" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/data-already-7.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/data-already-7.c
new file mode 100644
index 00000000000..b08417bd6ac
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/data-already-7.c
@@ -0,0 +1,17 @@
+/* { dg-skip-if "" { *-*-* } { "*" } { "-DACC_MEM_SHARED=0" } } */
+
+#include <openacc.h>
+
+int
+main (int argc, char *argv[])
+{
+  int i;
+
+#pragma acc enter data create (i)
+  acc_create (&i, sizeof i);
+
+  return 0;
+}
+
+/* { dg-shouldfail "" }
+   { dg-output "already mapped to" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/data-already-8.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/data-already-8.c
new file mode 100644
index 00000000000..a50f7de3a65
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/data-already-8.c
@@ -0,0 +1,16 @@
+/* { dg-skip-if "" { *-*-* } { "*" } { "-DACC_MEM_SHARED=0" } } */
+
+int
+main (int argc, char *argv[])
+{
+  int i;
+
+#pragma acc data create (i)
+#pragma acc parallel copyin (i)
+  ++i;
+
+  return 0;
+}
+
+/* { dg-shouldfail "" }
+   { dg-output "Trying to map into device .* object when .* is already mapped" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/deviceptr-1.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/deviceptr-1.c
new file mode 100644
index 00000000000..e271a3770a1
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/deviceptr-1.c
@@ -0,0 +1,32 @@
+/* { dg-do run } */
+
+#include <stdlib.h>
+
+int main (void)
+{
+  void *a, *a_1, *a_2;
+
+#define A (void *) 0x123
+  a = A;
+
+#pragma acc data copyout (a_1, a_2)
+#pragma acc kernels deviceptr (a)
+  {
+    a_1 = a;
+    a_2 = &a;
+  }
+
+  if (a != A)
+    abort ();
+  if (a_1 != a)
+    abort ();
+#if ACC_MEM_SHARED
+  if (a_2 != &a)
+    abort ();
+#else
+  if (a_2 == &a)
+    abort ();
+#endif
+
+  return 0;
+}
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/if-1.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/if-1.c
new file mode 100644
index 00000000000..184b3554bf5
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/if-1.c
@@ -0,0 +1,613 @@
+/* { dg-do run } */
+/* { dg-additional-options "-fno-builtin-acc_on_device" } */
+
+#include <openacc.h>
+#include <stdlib.h>
+#include <stdbool.h>
+
+#define N   32
+
+int
+main(int argc, char **argv)
+{
+    float *a, *b, *d_a, *d_b, exp, exp2;
+    int i;
+    const int one = 1;
+    const int zero = 0;
+    int n;
+
+    a = (float *) malloc (N * sizeof (float));
+    b = (float *) malloc (N * sizeof (float));
+    d_a = (float *) acc_malloc (N * sizeof (float));
+    d_b = (float *) acc_malloc (N * sizeof (float));
+
+    for (i = 0; i < N; i++)
+        a[i] = 4.0;
+
+#pragma acc parallel copyin(a[0:N]) copyout(b[0:N]) if(1)
+    {
+        int ii;
+
+        for (ii = 0; ii < N; ii++)
+        {
+            if (acc_on_device (acc_device_host))
+                b[ii] = a[ii] + 1;
+            else
+                b[ii] = a[ii];
+        }
+    }
+
+#if ACC_MEM_SHARED
+    exp = 5.0;
+#else
+    exp = 4.0;
+#endif
+
+    for (i = 0; i < N; i++)
+    {
+        if (b[i] != exp)
+            abort();
+    }
+
+    for (i = 0; i < N; i++)
+        a[i] = 16.0;
+
+#pragma acc parallel if(0)
+    {
+        int ii;
+
+        for (ii = 0; ii < N; ii++)
+        {
+            if (acc_on_device (acc_device_host))
+                b[ii] = a[ii] + 1;
+            else
+                b[ii] = a[ii];
+        }
+    }
+
+    for (i = 0; i < N; i++)
+    {
+        if (b[i] != 17.0)
+            abort();
+    }
+
+    for (i = 0; i < N; i++)
+        a[i] = 8.0;
+
+#pragma acc parallel copyin(a[0:N]) copyout(b[0:N]) if(one)
+    {
+        int ii;
+
+        for (ii = 0; ii < N; ii++)
+        {
+            if (acc_on_device (acc_device_host))
+                b[ii] = a[ii] + 1;
+            else
+                b[ii] = a[ii];
+        }
+    }
+
+#if ACC_MEM_SHARED
+    exp = 9.0;
+#else
+    exp = 8.0;
+#endif
+
+    for (i = 0; i < N; i++)
+    {
+        if (b[i] != exp)
+            abort();
+    }
+
+    for (i = 0; i < N; i++)
+        a[i] = 22.0;
+
+#pragma acc parallel if(zero)
+    {
+        int ii;
+
+        for (ii = 0; ii < N; ii++)
+        {
+            if (acc_on_device (acc_device_host))
+                b[ii] = a[ii] + 1;
+            else
+                b[ii] = a[ii];
+        }
+    }
+
+    for (i = 0; i < N; i++)
+    {
+        if (b[i] != 23.0)
+            abort();
+    }
+
+    for (i = 0; i < N; i++)
+        a[i] = 16.0;
+
+#pragma acc parallel copyin(a[0:N]) copyout(b[0:N]) if(true)
+    {
+        int ii;
+
+        for (ii = 0; ii < N; ii++)
+        {
+            if (acc_on_device (acc_device_host))
+                b[ii] = a[ii] + 1;
+            else
+                b[ii] = a[ii];
+        }
+    }
+
+#if ACC_MEM_SHARED
+    exp = 17.0;
+#else
+    exp = 16.0;
+#endif
+
+    for (i = 0; i < N; i++)
+    {
+        if (b[i] != exp)
+            abort();
+    }
+
+    for (i = 0; i < N; i++)
+        a[i] = 76.0;
+
+#pragma acc parallel if(false)
+    {
+        int ii;
+
+        for (ii = 0; ii < N; ii++)
+        {
+            if (acc_on_device (acc_device_host))
+                b[ii] = a[ii] + 1;
+            else
+                b[ii] = a[ii];
+        }
+    }
+
+    for (i = 0; i < N; i++)
+    {
+        if (b[i] != 77.0)
+            abort();
+    }
+
+    for (i = 0; i < N; i++)
+        a[i] = 22.0;
+
+    n = 1;
+
+#pragma acc parallel copyin(a[0:N]) copyout(b[0:N]) if(n)
+    {
+        int ii;
+
+        for (ii = 0; ii < N; ii++)
+        {
+            if (acc_on_device (acc_device_host))
+                b[ii] = a[ii] + 1;
+            else
+                b[ii] = a[ii];
+        }
+    }
+
+#if ACC_MEM_SHARED
+    exp = 23.0;
+#else
+    exp = 22.0;
+#endif
+
+    for (i = 0; i < N; i++)
+    {
+        if (b[i] != exp)
+            abort();
+    }
+
+    for (i = 0; i < N; i++)
+        a[i] = 18.0;
+
+    n = 0;
+
+#pragma acc parallel if(n)
+    {
+        int ii;
+
+        for (ii = 0; ii < N; ii++)
+        {
+            if (acc_on_device (acc_device_host))
+                b[ii] = a[ii] + 1;
+            else
+                b[ii] = a[ii];
+        }
+    }
+
+    for (i = 0; i < N; i++)
+    {
+        if (b[i] != 19.0)
+            abort();
+    }
+
+    for (i = 0; i < N; i++)
+        a[i] = 49.0;
+
+    n = 1;
+
+#pragma acc parallel copyin(a[0:N]) copyout(b[0:N]) if(n + n)
+    {
+        int ii;
+
+        for (ii = 0; ii < N; ii++)
+        {
+            if (acc_on_device (acc_device_host))
+                b[ii] = a[ii] + 1;
+            else
+                b[ii] = a[ii];
+        }
+    }
+
+#if ACC_MEM_SHARED
+    exp = 50.0;
+#else
+    exp = 49.0;
+#endif
+
+    for (i = 0; i < N; i++)
+    {
+        if (b[i] != exp)
+            abort();
+    }
+
+    for (i = 0; i < N; i++)
+        a[i] = 38.0;
+
+    n = 0;
+
+#pragma acc parallel if(n + n)
+    {
+        int ii;
+
+        for (ii = 0; ii < N; ii++)
+        {
+            if (acc_on_device (acc_device_host))
+                b[ii] = a[ii] + 1;
+            else
+                b[ii] = a[ii];
+        }
+    }
+
+    for (i = 0; i < N; i++)
+    {
+        if (b[i] != 39.0)
+            abort();
+    }
+
+    for (i = 0; i < N; i++)
+        a[i] = 91.0;
+
+#pragma acc parallel copyin(a[0:N]) copyout(b[0:N]) if(-2)
+    {
+        int ii;
+
+        for (ii = 0; ii < N; ii++)
+        {
+            if (acc_on_device (acc_device_host))
+                b[ii] = a[ii] + 1;
+            else
+                b[ii] = a[ii];
+        }
+    }
+
+#if ACC_MEM_SHARED
+    exp = 92.0;
+#else
+    exp = 91.0;
+#endif
+
+    for (i = 0; i < N; i++)
+    {
+        if (b[i] != exp)
+            abort();
+    }
+
+    for (i = 0; i < N; i++)
+        a[i] = 43.0;
+
+#pragma acc parallel copyin(a[0:N]) copyout(b[0:N]) if(one == 1)
+    {
+        int ii;
+
+        for (ii = 0; ii < N; ii++)
+        {
+            if (acc_on_device (acc_device_host))
+                b[ii] = a[ii] + 1;
+            else
+                b[ii] = a[ii];
+        }
+    }
+
+#if ACC_MEM_SHARED
+    exp = 44.0;
+#else
+    exp = 43.0;
+#endif
+
+    for (i = 0; i < N; i++)
+    {
+        if (b[i] != exp)
+            abort();
+    }
+
+    for (i = 0; i < N; i++)
+        a[i] = 87.0;
+
+#pragma acc parallel if(one == 0)
+    {
+        int ii;
+
+        for (ii = 0; ii < N; ii++)
+        {
+            if (acc_on_device (acc_device_host))
+                b[ii] = a[ii] + 1;
+            else
+                b[ii] = a[ii];
+        }
+    }
+
+    for (i = 0; i < N; i++)
+    {
+        if (b[i] != 88.0)
+            abort();
+    }
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 3.0;
+        b[i] = 9.0;
+    }
+
+#if ACC_MEM_SHARED
+    exp = 0.0;
+    exp2 = 0.0;
+#else
+    acc_map_data (a, d_a, N * sizeof (float));
+    acc_map_data (b, d_b, N * sizeof (float));
+    exp = 3.0;
+    exp2 = 9.0;
+#endif
+
+#pragma acc update device(a[0:N], b[0:N]) if(1)
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 0.0;
+        b[i] = 0.0;
+    }
+
+#pragma acc update host(a[0:N], b[0:N]) if(1)
+
+    for (i = 0; i < N; i++)
+    {
+        if (a[i] != exp)
+            abort();
+
+        if (b[i] != exp2)
+            abort();
+    }
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 6.0;
+        b[i] = 12.0;
+    }
+
+#pragma acc update device(a[0:N], b[0:N]) if(0)
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 0.0;
+        b[i] = 0.0;
+    }
+
+#pragma acc update host(a[0:N], b[0:N]) if(1)
+
+    for (i = 0; i < N; i++)
+    {
+        if (a[i] != exp)
+            abort();
+
+        if (b[i] != exp2)
+            abort();
+    }
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 26.0;
+        b[i] = 21.0;
+    }
+
+#pragma acc update device(a[0:N], b[0:N]) if(1)
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 0.0;
+        b[i] = 0.0;
+    }
+
+#pragma acc update host(a[0:N], b[0:N]) if(0)
+
+    for (i = 0; i < N; i++)
+    {
+        if (a[i] != 0.0)
+            abort();
+
+        if (b[i] != 0.0)
+            abort();
+    }
+
+#if !ACC_MEM_SHARED
+    acc_unmap_data (a);
+    acc_unmap_data (b);
+#endif
+
+    acc_free (d_a);
+    acc_free (d_b);
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 4.0;
+        b[i] = 0.0;
+    }
+
+#pragma acc data copyin(a[0:N]) copyout(b[0:N]) if(1)
+{
+#pragma acc parallel present(a[0:N])
+    {
+        int ii;
+
+        for (ii = 0; ii < N; ii++)
+        {
+            b[ii] = a[ii];
+        }
+    }
+}
+
+    for (i = 0; i < N; i++)
+    {
+        if (b[i] != 4.0)
+            abort();
+    }
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 8.0;
+        b[i] = 1.0;
+    }
+
+#pragma acc data copyin(a[0:N]) copyout(b[0:N]) if(0)
+{
+#if !ACC_MEM_SHARED
+    if (acc_is_present (a, N * sizeof (float)))
+        abort ();
+#endif
+
+#if !ACC_MEM_SHARED
+    if (acc_is_present (b, N * sizeof (float)))
+        abort ();
+#endif
+}
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 18.0;
+        b[i] = 21.0;
+    }
+
+#pragma acc data copyin(a[0:N]) if(1)
+{
+#if !ACC_MEM_SHARED
+    if (!acc_is_present (a, N * sizeof (float)))
+        abort ();
+#endif
+
+#pragma acc data copyout(b[0:N]) if(0)
+    {
+#if !ACC_MEM_SHARED
+        if (acc_is_present (b, N * sizeof (float)))
+            abort ();
+#endif
+
+#pragma acc data copyout(b[0:N]) if(1)
+        {
+#pragma acc parallel present(a[0:N]) present(b[0:N])
+            {
+                int ii;
+
+                for (ii = 0; ii < N; ii++)
+                {
+                    b[ii] = a[ii];
+                }
+            }
+        }
+
+#if !ACC_MEM_SHARED
+        if (acc_is_present (b, N * sizeof (float)))
+            abort ();
+#endif
+    }
+}
+
+    for (i = 0; i < N; i++)
+    {
+        if (b[i] != 18.0)
+            abort ();
+    }
+
+#pragma acc enter data copyin (b[0:N]) if (0)
+
+#if !ACC_MEM_SHARED
+    if (acc_is_present (b, N * sizeof (float)))
+	abort ();
+#endif
+
+#pragma acc exit data delete (b[0:N]) if (0)
+
+#pragma acc enter data copyin (b[0:N]) if (1)
+
+#if !ACC_MEM_SHARED
+    if (!acc_is_present (b, N * sizeof (float)))
+	abort ();
+#endif
+
+#pragma acc exit data delete (b[0:N]) if (1)
+
+#if !ACC_MEM_SHARED
+    if (acc_is_present (b, N * sizeof (float)))
+	abort ();
+#endif
+
+#pragma acc enter data copyin (b[0:N]) if (zero)
+
+#if !ACC_MEM_SHARED
+    if (acc_is_present (b, N * sizeof (float)))
+	abort ();
+#endif
+
+#pragma acc exit data delete (b[0:N]) if (zero)
+
+#pragma acc enter data copyin (b[0:N]) if (one)
+
+#if !ACC_MEM_SHARED
+    if (!acc_is_present (b, N * sizeof (float)))
+	abort ();
+#endif
+
+#pragma acc exit data delete (b[0:N]) if (one)
+
+#if !ACC_MEM_SHARED
+    if (acc_is_present (b, N * sizeof (float)))
+	abort ();
+#endif
+
+#pragma acc enter data copyin (b[0:N]) if (one == 0)
+
+#if !ACC_MEM_SHARED
+    if (acc_is_present (b, N * sizeof (float)))
+	abort ();
+#endif
+
+#pragma acc exit data delete (b[0:N]) if (one == 0)
+
+#pragma acc enter data copyin (b[0:N]) if (one == 1)
+
+#if !ACC_MEM_SHARED
+    if (!acc_is_present (b, N * sizeof (float)))
+	abort ();
+#endif
+
+#pragma acc exit data delete (b[0:N]) if (one == 1)
+
+#if !ACC_MEM_SHARED
+    if (acc_is_present (b, N * sizeof (float)))
+	abort ();
+#endif
+
+    return 0;
+}
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/kernels-1.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/kernels-1.c
new file mode 100644
index 00000000000..3acfdf55ef6
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/kernels-1.c
@@ -0,0 +1,184 @@
+/* { dg-do run } */
+
+#include <stdlib.h>
+
+int i;
+
+int main (void)
+{
+  int j, v;
+
+#if 0
+  i = -1;
+  j = -2;
+  v = 0;
+#pragma acc kernels /* copyout */ present_or_copyout (v) copyin (i, j)
+  {
+    if (i != -1 || j != -2)
+      abort ();
+    i = 2;
+    j = 1;
+    if (i != 2 || j != 1)
+      abort ();
+    v = 1;
+  }
+  if (v != 1 || i != -1 || j != -2)
+    abort ();
+
+  i = -1;
+  j = -2;
+  v = 0;
+#pragma acc kernels /* copyout */ present_or_copyout (v) copyout (i, j)
+  {
+    i = 2;
+    j = 1;
+    if (i != 2 || j != 1)
+      abort ();
+    v = 1;
+  }
+  if (v != 1 || i != 2 || j != 1)
+    abort ();
+
+  i = -1;
+  j = -2;
+  v = 0;
+#pragma acc kernels /* copyout */ present_or_copyout (v) copy (i, j)
+  {
+    if (i != -1 || j != -2)
+      abort ();
+    i = 2;
+    j = 1;
+    if (i != 2 || j != 1)
+      abort ();
+    v = 1;
+  }
+  if (v != 1 || i != 2 || j != 1)
+    abort ();
+
+  i = -1;
+  j = -2;
+  v = 0;
+#pragma acc kernels /* copyout */ present_or_copyout (v) create (i, j)
+  {
+    i = 2;
+    j = 1;
+    if (i != 2 || j != 1)
+      abort ();
+    v = 1;
+  }
+  if (v != 1 || i != -1 || j != -2)
+    abort ();
+#endif
+
+  i = -1;
+  j = -2;
+  v = 0;
+#pragma acc kernels /* copyout */ present_or_copyout (v) present_or_copyin (i, j)
+  {
+    if (i != -1 || j != -2)
+      abort ();
+    i = 2;
+    j = 1;
+    if (i != 2 || j != 1)
+      abort ();
+    v = 1;
+  }
+  if (v != 1)
+    abort ();
+#if ACC_MEM_SHARED
+  if (i != 2 || j != 1)
+    abort ();
+#else
+  if (i != -1 || j != -2)
+    abort ();
+#endif
+
+  i = -1;
+  j = -2;
+  v = 0;
+#pragma acc kernels /* copyout */ present_or_copyout (v) present_or_copyout (i, j)
+  {
+    i = 2;
+    j = 1;
+    if (i != 2 || j != 1)
+      abort ();
+    v = 1;
+  }
+  if (v != 1 || i != 2 || j != 1)
+    abort ();
+
+  i = -1;
+  j = -2;
+  v = 0;
+#pragma acc kernels /* copyout */ present_or_copyout (v) present_or_copy (i, j)
+  {
+    if (i != -1 || j != -2)
+      abort ();
+    i = 2;
+    j = 1;
+    if (i != 2 || j != 1)
+      abort ();
+    v = 1;
+  }
+  if (v != 1 || i != 2 || j != 1)
+    abort ();
+
+  i = -1;
+  j = -2;
+  v = 0;
+#pragma acc kernels /* copyout */ present_or_copyout (v) present_or_create (i, j)
+  {
+    i = 2;
+    j = 1;
+    if (i != 2 || j != 1)
+      abort ();
+    v = 1;
+  }
+  if (v != 1)
+    abort ();
+#if ACC_MEM_SHARED
+  if (i != 2 || j != 1)
+    abort ();
+#else
+  if (i != -1 || j != -2)
+    abort ();
+#endif
+
+#if 0
+  i = -1;
+  j = -2;
+  v = 0;
+#pragma acc kernels /* copyout */ present_or_copyout (v) present (i, j)
+  {
+    if (i != -1 || j != -2)
+      abort ();
+    i = 2;
+    j = 1;
+    if (i != 2 || j != 1)
+      abort ();
+    v = 1;
+  }
+  if (v != 1 || i != 2 || j != 1)
+    abort ();
+#endif
+
+#if 0
+  i = -1;
+  j = -2;
+  v = 0;
+#pragma acc kernels /* copyout */ present_or_copyout (v)
+  {
+    if (i != -1 || j != -2)
+      abort ();
+    i = 2;
+    j = 1;
+    if (i != 2 || j != 1)
+      abort ();
+    v = 1;
+  }
+  if (v != 1 || i != 2 || j != 1)
+    abort ();
+#endif
+
+  return 0;
+}
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/kernels-empty.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/kernels-empty.c
new file mode 100644
index 00000000000..a68a7cdb032
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/kernels-empty.c
@@ -0,0 +1,6 @@
+int
+main (void)
+{
+#pragma acc kernels
+  ;
+}
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-1.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-1.c
new file mode 100644
index 00000000000..17129d8bebd
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-1.c
@@ -0,0 +1,24 @@
+/* { dg-do run } */
+
+#include <openacc.h>
+
+int
+main (int argc, char **argv)
+{
+  acc_device_t devtype = acc_device_host;
+
+#if ACC_DEVICE_TYPE_nvidia
+  devtype = acc_device_nvidia;
+
+  if (acc_get_num_devices (devtype) == 0)
+    return 0;
+#endif
+
+  acc_init (devtype);
+
+  acc_init (devtype);
+
+  return 0;
+}
+
+/* { dg-shouldfail "libgomp: device already active" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-10.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-10.c
new file mode 100644
index 00000000000..cf1af8c142f
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-10.c
@@ -0,0 +1,58 @@
+/* { dg-do run } */
+
+#include <stdlib.h>
+#include <openacc.h>
+
+int
+main (int argc, char **argv)
+{
+  void *d;
+  acc_device_t devtype = acc_device_host;
+
+#if ACC_DEVICE_TYPE_nvidia
+  devtype = acc_device_nvidia;
+
+  if (acc_get_num_devices (acc_device_nvidia) == 0)
+    return 0;
+#endif
+
+  acc_init (devtype);
+
+  d = acc_malloc (0);
+  if (d != NULL)
+    abort ();
+
+  acc_free (0);
+
+  acc_shutdown (devtype);
+
+  acc_set_device_type (devtype);
+
+  d = acc_malloc (0);
+  if (d != NULL)
+    abort ();
+
+  acc_shutdown (devtype);
+
+  acc_init (devtype);
+
+  d = acc_malloc (1024);
+  if (d == NULL)
+    abort ();
+
+  acc_free (d);
+
+  acc_shutdown (devtype);
+
+  acc_set_device_type (devtype);
+
+  d = acc_malloc (1024);
+  if (d == NULL)
+    abort ();
+
+  acc_free (d);
+
+  acc_shutdown (devtype);
+
+  return 0;
+}
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-11.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-11.c
new file mode 100644
index 00000000000..eccdb8cbacf
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-11.c
@@ -0,0 +1,23 @@
+/* Only nvptx plugin does the required error checking.
+   { dg-do run { target openacc_nvidia_accel_selected } } */
+
+#include <stdlib.h>
+#include <openacc.h>
+#include <stdint.h>
+
+int
+main (int argc, char **argv)
+{
+  const int N = 512;
+  void *d;
+
+  d = acc_malloc (N);
+  if (d == NULL)
+    abort ();
+
+  acc_free ((void *)((uintptr_t) d + (uintptr_t) (N >> 1)));
+
+  return 0;
+}
+
+/* { dg-shouldfail "libgomp: mem free failed 1" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-12.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-12.c
new file mode 100644
index 00000000000..b46f59028b7
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-12.c
@@ -0,0 +1,37 @@
+/* { dg-do run } */
+/* { dg-skip-if "" { *-*-* } { "*" } { "-DACC_MEM_SHARED=0" } } */
+
+#include <string.h>
+#include <stdlib.h>
+#include <openacc.h>
+
+int
+main (int argc, char **argv)
+{
+  const int N = 256;
+  int i;
+  unsigned char *h;
+
+  h = (unsigned char *) malloc (N);
+
+  for (i = 0; i < N; i++)
+    {
+      h[i] = i;
+    }
+
+  (void) acc_copyin (h, N);
+
+  memset (h, 0, N);
+
+  acc_copyout (h, N);
+
+  for (i = 0; i < N; i++)
+    {
+      if (h[i] != i)
+	abort ();
+    }
+
+  free (h);
+
+  return 0;
+}
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-13.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-13.c
new file mode 100644
index 00000000000..7098ef3549c
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-13.c
@@ -0,0 +1,60 @@
+/* { dg-do run } */
+
+#include <stdlib.h>
+#include <openacc.h>
+
+#include <stdio.h>
+
+int
+main (int argc, char **argv)
+{
+  const int N = 256;
+  int i;
+  unsigned char *h;
+  void *d;
+
+  h = (unsigned char *) malloc (N);
+
+  for (i = 0; i < N; i++)
+    {
+      h[i] = i;
+    }
+
+  d = acc_copyin (h, N);
+
+  if (acc_is_present (h, 1) != 1)
+    abort ();
+
+  if (acc_is_present (h, N + 1) != 0)
+    abort ();
+
+  if (acc_is_present (h + 1, N) != 0)
+    abort ();
+
+  if (acc_is_present (h - 1, N) != 0)
+    abort ();
+
+  if (acc_is_present (h - 1, N - 1) != 0)
+    abort ();
+
+  if (acc_is_present (h + N, 0) != 0)
+    abort ();
+
+  if (acc_is_present (h + N, N) != 0)
+    abort ();
+
+  if (acc_is_present (0, N) != 0)
+    abort ();
+   
+  if (acc_is_present (h, 0) != 0)
+    abort ();
+
+  acc_free (d);
+
+  if (acc_is_present (h, 1) != 0)
+    abort ();
+
+  free (h);
+
+  return 0;
+}
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-14.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-14.c
new file mode 100644
index 00000000000..a9632f786ee
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-14.c
@@ -0,0 +1,61 @@
+/* { dg-do run } */
+
+#include <stdlib.h>
+#include <openacc.h>
+
+#include <stdio.h>
+
+int
+main (int argc, char **argv)
+{
+  const int N = 256;
+  int i;
+  unsigned char *h;
+  void *d;
+
+  h = (unsigned char *) malloc (N);
+
+  for (i = 0; i < N; i++)
+    {
+      h[i] = i;
+    }
+
+  d = acc_copyin (h, N);
+
+  if (acc_is_present (h, 1) != 1)
+    abort ();
+
+  if (acc_is_present (h + N - 1, 1) != 1)
+    abort ();
+
+  if (acc_is_present (h - 1, 1) != 0)
+    abort ();
+
+  if (acc_is_present (h + N, 1) != 0)
+    abort ();
+
+  for (i = 0; i < N; i++)
+    {
+      if (acc_is_present (h + i, 1) != 1)
+	abort ();
+    }
+
+  for (i = 0; i < N; i++)
+    {
+      if (acc_is_present (h + i, N - i) != 1)
+	abort ();
+    }
+
+  acc_free (d);
+
+  for (i = 0; i < N; i++)
+    {
+      if (acc_is_present (h + i, N - i) != 0)
+	abort ();
+    }
+
+
+  free (h);
+
+  return 0;
+}
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-15.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-15.c
new file mode 100644
index 00000000000..4f6a731be1d
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-15.c
@@ -0,0 +1,33 @@
+/* { dg-do run } */
+
+#include <stdlib.h>
+#include <openacc.h>
+
+int
+main (int argc, char **argv)
+{
+  const int N = 256;
+  int i;
+  unsigned char *h;
+
+  h = (unsigned char *) malloc (N);
+
+  for (i = 0; i < N; i++)
+    {
+      h[i] = i;
+    }
+
+  (void) acc_copyin (h, N);
+
+  acc_copyout (h, N);
+
+  for (i = 0; i < N; i++)
+    {
+      if (acc_is_present (h + i, 1) != 0)
+	abort ();
+    }
+
+  free (h);
+
+  return 0;
+}
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-16.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-16.c
new file mode 100644
index 00000000000..9d277ac39f7
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-16.c
@@ -0,0 +1,29 @@
+/* { dg-do run } */
+
+#include <stdlib.h>
+#include <openacc.h>
+
+int
+main (int argc, char **argv)
+{
+  const int N = 256;
+  int i;
+  unsigned char *h;
+
+  h = (unsigned char *) malloc (N);
+
+  for (i = 0; i < N; i++)
+    {
+      h[i] = i;
+    }
+
+  (void) acc_copyin (h, N);
+
+  (void) acc_copyin (h, N);
+
+  free (h);
+
+  return 0;
+}
+
+/* { dg-shouldfail "libgomp: \[\h+,\+256\] already mapped to \[\h+,\+256\]" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-17.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-17.c
new file mode 100644
index 00000000000..5ff894cb33a
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-17.c
@@ -0,0 +1,31 @@
+/* { dg-do run } */
+
+#include <stdlib.h>
+#include <openacc.h>
+
+int
+main (int argc, char **argv)
+{
+  const int N = 256;
+  int i;
+  unsigned char *h;
+
+  h = (unsigned char *) malloc (N);
+
+  for (i = 0; i < N; i++)
+    {
+      h[i] = i;
+    }
+
+  (void) acc_copyin (h, N);
+
+  acc_copyout (h, N);
+
+  acc_copyout (h, N);
+
+  free (h);
+
+  return 0;
+}
+
+/* { dg-shouldfail "libgomp: \[\h+,256\] is not mapped" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-18.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-18.c
new file mode 100644
index 00000000000..2bc32637d36
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-18.c
@@ -0,0 +1,34 @@
+/* { dg-do run } */
+
+#include <stdlib.h>
+#include <openacc.h>
+
+#include <stdio.h>
+
+int
+main (int argc, char **argv)
+{
+  const int N = 256;
+  int i;
+  unsigned char *h;
+  void *d;
+
+  h = (unsigned char *) malloc (N);
+
+  for (i = 0; i < N; i++)
+    {
+      h[i] = i;
+    }
+
+  d = acc_copyin (h, N);
+
+  acc_free (d);
+
+  acc_copyout (h, N);
+
+  free (h);
+
+  return 0;
+}
+
+/* { dg-shouldfail "libgomp: \[\h+,256\] is not mapped" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-19.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-19.c
new file mode 100644
index 00000000000..3581616d754
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-19.c
@@ -0,0 +1,60 @@
+/* { dg-do run } */
+/* { dg-skip-if "" { *-*-* } { "*" } { "-DACC_MEM_SHARED=0" } } */
+
+#include <string.h>
+#include <stdlib.h>
+#include <openacc.h>
+
+#include <stdio.h>
+
+int
+main (int argc, char **argv)
+{
+  const int N = 256;
+  int i;
+  unsigned char *h[N];
+
+  for (i = 0; i < N; i++)
+    {
+      int j;
+      unsigned char *p;
+
+      h[i] = (unsigned char *) malloc (N);
+      p = h[i];
+
+      for (j = 0; j < N; j++)
+	{
+	  p[j] = i;
+	}
+
+      (void) acc_copyin (p, N);
+    }
+
+  for (i = 0; i < N; i++)
+    {
+      memset (h[i], 0, i);
+    }
+
+  for (i = 0; i < N; i++)
+    {
+      int j;
+      unsigned char *p;
+
+      acc_copyout (h[i], N);
+
+      p = h[i];
+
+      for (j = 0; j < N; j++)
+	{
+	  if (p[j] != i)
+	    abort ();
+	}
+    }
+
+  for (i = 0; i < N; i++)
+    {
+      free (h[i]);
+    }
+
+  return 0;
+}
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-2.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-2.c
new file mode 100644
index 00000000000..9a4501f0e00
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-2.c
@@ -0,0 +1,26 @@
+/* { dg-do run } */
+
+#include <openacc.h>
+
+int
+main (int argc, char **argv)
+{
+  acc_device_t devtype = acc_device_host;
+
+#if ACC_DEVICE_TYPE_nvidia
+  devtype = acc_device_nvidia;
+
+  if (acc_get_num_devices (acc_device_nvidia) == 0)
+    return 0;
+#endif
+
+  acc_init (devtype);
+
+  acc_shutdown (devtype);
+
+  acc_shutdown (devtype);
+
+  return 0;
+}
+
+/* { dg-shouldfail "libgomp: no device initialized" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-20.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-20.c
new file mode 100644
index 00000000000..b379a8f7add
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-20.c
@@ -0,0 +1,29 @@
+/* { dg-do run } */
+
+#include <stdlib.h>
+#include <openacc.h>
+
+int
+main (int argc, char **argv)
+{
+  const int N = 256;
+  int i;
+  unsigned char *h;
+
+  h = (unsigned char *) malloc (N);
+
+  for (i = 0; i < N; i++)
+    {
+      h[i] = i;
+    }
+
+  (void) acc_copyin (h, N);
+
+  acc_copyout (h, N + 1);
+
+  free (h);
+
+  return 0;
+}
+
+/* { dg-shouldfail "libgomp: \[\h+,256\] surounds2 \[\h+,\+257\]" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-21.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-21.c
new file mode 100644
index 00000000000..3a67400a620
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-21.c
@@ -0,0 +1,29 @@
+/* { dg-do run } */
+
+#include <stdlib.h>
+#include <openacc.h>
+
+int
+main (int argc, char **argv)
+{
+  const int N = 256;
+  int i;
+  unsigned char *h;
+
+  h = (unsigned char *) malloc (N);
+
+  for (i = 0; i < N; i++)
+    {
+      h[i] = i;
+    }
+
+  (void) acc_copyin (h, N);
+
+  acc_copyout (h, 0);
+
+  free (h);
+
+  return 0;
+}
+
+/* { dg-shouldfail "libgomp: \[\h+,0\] is not mapped" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-22.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-22.c
new file mode 100644
index 00000000000..2b86da80580
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-22.c
@@ -0,0 +1,29 @@
+/* { dg-do run } */
+
+#include <stdlib.h>
+#include <openacc.h>
+
+int
+main (int argc, char **argv)
+{
+  const int N = 256;
+  int i;
+  unsigned char *h;
+
+  h = (unsigned char *) malloc (N);
+
+  for (i = 0; i < N; i++)
+    {
+      h[i] = i;
+    }
+
+  (void) acc_copyin (h, N);
+
+  acc_copyout (h + 1, N - 1);
+
+  free (h);
+
+  return 0;
+}
+
+/* { dg-shouldfail "libgomp: \[\h+,256\] surrounds2 \[\h+,\+255\]" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-23.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-23.c
new file mode 100644
index 00000000000..38f236dcea9
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-23.c
@@ -0,0 +1,39 @@
+/* { dg-do run } */
+
+#include <stdlib.h>
+#include <openacc.h>
+
+int
+main (int argc, char **argv)
+{
+  const int N = 256;
+  int i;
+  unsigned char *h1, *h2;
+
+  h1 = (unsigned char *) malloc (N);
+
+  for (i = 0; i < N; i++)
+    {
+      h1[i] = 0xab;
+    }
+
+  (void) acc_copyin (h1, N);
+
+  h2 = (unsigned char *) malloc (N);
+
+  for (i = 0; i < N; i++)
+    {
+      h2[i] = 0xde;
+    }
+
+  (void) acc_copyin (h2, N);
+
+  acc_copyout (h1, N + N);
+
+  free (h1);
+  free (h2);
+
+  return 0;
+}
+
+/* { dg-shouldfail "libgomp: \[\h+,256\] surrounds2 \[\h+,\+512\]" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-24.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-24.c
new file mode 100644
index 00000000000..d7de8e34f4c
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-24.c
@@ -0,0 +1,55 @@
+/* { dg-do run } */
+
+#include <stdlib.h>
+#include <openacc.h>
+
+int
+main (int argc, char **argv)
+{
+  const int N = 256;
+  int i;
+  unsigned char *h;
+  void *d;
+
+  h = (unsigned char *) malloc (N);
+
+  d = acc_create (h, N);
+  if (!d)
+    abort ();
+
+  for (i = 0; i < N; i++)
+    {
+      if (acc_is_present (h + i, 1) != 1)
+	abort ();
+    }
+
+  acc_delete (h, N);
+
+  for (i = 0; i < N; i++)
+    {
+      if (acc_is_present (h + i, 1) != 0)
+	abort ();
+    }
+
+  d = acc_create (h, N);
+  if (!d)
+    abort ();
+
+  for (i = 0; i < N; i++)
+    {
+      if (acc_is_present (h + i, 1) != 1)
+	abort ();
+    }
+
+  acc_delete (h, N);
+
+  for (i = 0; i < N; i++)
+    {
+      if (acc_is_present (h + i, 1) != 0)
+	abort ();
+    }
+
+  free (h);
+
+  return 0;
+}
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-25.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-25.c
new file mode 100644
index 00000000000..11458283b2c
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-25.c
@@ -0,0 +1,30 @@
+/* { dg-do run } */
+
+#include <stdlib.h>
+#include <openacc.h>
+
+int
+main (int argc, char **argv)
+{
+  const int N = 256;
+  unsigned char *h;
+  void *d;
+
+  h = (unsigned char *) malloc (N);
+
+  d = acc_create (h, N);
+  if (!d)
+    abort ();
+
+  d = acc_create (h, N);
+  if (!d)
+    abort ();
+
+  acc_delete (h, N);
+
+  free (h);
+
+  return 0;
+}
+
+/* { dg-shouldfail "libgomp: \[\h+,256\] already mapped to \[\h+,256\]" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-26.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-26.c
new file mode 100644
index 00000000000..a23f56e753f
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-26.c
@@ -0,0 +1,26 @@
+/* { dg-do run } */
+
+#include <stdlib.h>
+#include <openacc.h>
+
+int
+main (int argc, char **argv)
+{
+  const int N = 256;
+  unsigned char *h;
+  void *d;
+
+  h = (unsigned char *) malloc (N);
+
+  d = acc_create (h, 0);
+  if (!d)
+    abort ();
+
+  acc_delete (h, N);
+
+  free (h);
+
+  return 0;
+}
+
+/* { dg-shouldfail "libgomp: \[\h+,\+0\] is a bad range" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-27.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-27.c
new file mode 100644
index 00000000000..074fddb2ab7
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-27.c
@@ -0,0 +1,26 @@
+/* { dg-do run } */
+
+#include <stdlib.h>
+#include <openacc.h>
+
+int
+main (int argc, char **argv)
+{
+  const int N = 256;
+  unsigned char *h;
+  void *d;
+
+  h = (unsigned char *) malloc (N);
+
+  d = acc_create (0, N);
+  if (!d)
+    abort ();
+
+  acc_delete (h, N);
+
+  free (h);
+
+  return 0;
+}
+
+/* { dg-shouldfail "libgomp: \[\(nil\)\] is a bad range" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-28.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-28.c
new file mode 100644
index 00000000000..027f7cc6460
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-28.c
@@ -0,0 +1,26 @@
+/* { dg-do run } */
+
+#include <stdlib.h>
+#include <openacc.h>
+
+int
+main (int argc, char **argv)
+{
+  const int N = 256;
+  unsigned char *h;
+  void *d;
+
+  h = (unsigned char *) malloc (N);
+
+  d = acc_create (h, N);
+  if (!d)
+    abort ();
+
+  acc_delete (0, N);
+
+  free (h);
+
+  return 0;
+}
+
+/* { dg-shouldfail "libgomp: \[\(nil\),256\] is not mapped" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-29.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-29.c
new file mode 100644
index 00000000000..a66de0f538d
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-29.c
@@ -0,0 +1,26 @@
+/* { dg-do run } */
+
+#include <stdlib.h>
+#include <openacc.h>
+
+int
+main (int argc, char **argv)
+{
+  const int N = 256;
+  unsigned char *h;
+  void *d;
+
+  h = (unsigned char *) malloc (N);
+
+  d = acc_create (h, N);
+  if (!d)
+    abort ();
+
+  acc_delete (h, 0);
+
+  free (h);
+
+  return 0;
+}
+
+/* { dg-shouldfail "libgomp: \[\h+,0\] is not mapped" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-3.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-3.c
new file mode 100644
index 00000000000..e823a4153fa
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-3.c
@@ -0,0 +1,15 @@
+/* { dg-do run } */
+
+#include <openacc.h>
+
+int
+main (int argc, char **argv)
+{
+  acc_init (acc_device_host);
+
+  acc_shutdown (acc_device_not_host);
+
+  return 0;
+}
+
+/* { dg-shouldfail "libgomp: device 4(4) is initialized" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-30.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-30.c
new file mode 100644
index 00000000000..ce2bdb44739
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-30.c
@@ -0,0 +1,26 @@
+/* { dg-do run } */
+
+#include <stdlib.h>
+#include <openacc.h>
+
+int
+main (int argc, char **argv)
+{
+  const int N = 256;
+  unsigned char *h;
+  void *d;
+
+  h = (unsigned char *) malloc (N);
+
+  d = acc_create (h, N);
+  if (!d)
+    abort ();
+
+  acc_delete (h, N - 2);
+
+  free (h);
+
+  return 0;
+}
+
+/* { dg-shouldfail "libgomp: \[\h+,256\] surrounds2 \[\h+,\+254\]" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-31.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-31.c
new file mode 100644
index 00000000000..25ce5a95977
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-31.c
@@ -0,0 +1,27 @@
+/* { dg-do run } */
+
+#include <stdlib.h>
+#include <openacc.h>
+
+int
+main (int argc, char **argv)
+{
+  const int N = 256;
+  unsigned char *h;
+  void *d;
+
+  h = (unsigned char *) malloc (N);
+
+  d = acc_present_or_create (h, N);
+  if (!d)
+    abort ();
+
+  if (acc_is_present (h, 1) != 1)
+    abort ();
+
+  acc_delete (h, N);
+
+  free (h);
+
+  return 0;
+}
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-32.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-32.c
new file mode 100644
index 00000000000..e3f87a85672
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-32.c
@@ -0,0 +1,38 @@
+/* { dg-do run } */
+
+#include <stdlib.h>
+#include <openacc.h>
+
+int
+main (int argc, char **argv)
+{
+  const int N = 256;
+  unsigned char *h;
+  void *d1, *d2;
+
+  h = (unsigned char *) malloc (N);
+
+  d1 = acc_present_or_create (h, N);
+  if (!d1)
+    abort ();
+
+  d2 = acc_present_or_create (h, N);
+  if (!d2)
+    abort ();
+
+  if (d1 != d2)
+    abort ();
+
+  d2 = acc_pcreate (h, N);
+  if (!d2)
+    abort ();
+
+  if (d1 != d2)
+    abort ();
+
+  acc_delete (h, N);
+
+  free (h);
+
+  return 0;
+}
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-33.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-33.c
new file mode 100644
index 00000000000..4abaa02e085
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-33.c
@@ -0,0 +1,31 @@
+/* { dg-do run } */
+
+#include <stdlib.h>
+#include <openacc.h>
+
+int
+main (int argc, char **argv)
+{
+  const int N = 256;
+  unsigned char *h;
+  void *d1, *d2;
+
+  h = (unsigned char *) malloc (N);
+
+  d1 = acc_present_or_create (h, N);
+  if (!d1)
+    abort ();
+
+  d2 = acc_present_or_create (h, N - 2);
+  if (!d2)
+    abort ();
+
+  if (d1 != d2)
+    abort ();
+
+  acc_delete (h, N);
+
+  free (h);
+
+  return 0;
+}
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-34.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-34.c
new file mode 100644
index 00000000000..32d5d5117e6
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-34.c
@@ -0,0 +1,33 @@
+/* { dg-do run } */
+
+#include <stdlib.h>
+#include <openacc.h>
+
+int
+main (int argc, char **argv)
+{
+  const int N = 256;
+  unsigned char *h;
+  void *d1, *d2;
+
+  h = (unsigned char *) malloc (N);
+
+  d1 = acc_present_or_create (h, N);
+  if (!d1)
+    abort ();
+
+  d2 = acc_present_or_create (h + 2, N);
+  if (!d2)
+    abort ();
+
+  if (d1 != d2)
+    abort ();
+
+  acc_delete (h, N);
+
+  free (h);
+
+  return 0;
+}
+
+/* { dg-shouldfail "libgomp: \[\h+,\+256\] not mapped" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-35.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-35.c
new file mode 100644
index 00000000000..ca8edab52a2
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-35.c
@@ -0,0 +1,26 @@
+/* { dg-do run } */
+
+#include <stdlib.h>
+#include <openacc.h>
+
+int
+main (int argc, char **argv)
+{
+  const int N = 256;
+  unsigned char *h;
+  void *d;
+
+  h = (unsigned char *) malloc (N);
+
+  d = acc_present_or_create (0, N);
+  if (!d)
+    abort ();
+
+  acc_delete (h, N);
+
+  free (h);
+
+  return 0;
+}
+
+/* { dg-shouldfail "libgomp: \[\(nil\),+256\] is a bad range" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-36.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-36.c
new file mode 100644
index 00000000000..cb29397e14e
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-36.c
@@ -0,0 +1,26 @@
+/* { dg-do run } */
+
+#include <stdlib.h>
+#include <openacc.h>
+
+int
+main (int argc, char **argv)
+{
+  const int N = 256;
+  unsigned char *h;
+  void *d;
+
+  h = (unsigned char *) malloc (N);
+
+  d = acc_present_or_create (h, 0);
+  if (!d)
+    abort ();
+
+  acc_delete (h, N);
+
+  free (h);
+
+  return 0;
+}
+
+/* { dg-shouldfail "libgomp: \[\h+,\+0\] is a bad range" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-37.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-37.c
new file mode 100644
index 00000000000..5a7d533a670
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-37.c
@@ -0,0 +1,40 @@
+/* { dg-do run } */
+/* { dg-skip-if "" { *-*-* } { "*" } { "-DACC_MEM_SHARED=0" } } */
+
+#include <string.h>
+#include <stdlib.h>
+#include <openacc.h>
+
+int
+main (int argc, char **argv)
+{
+  const int N = 256;
+  int i;
+  unsigned char *h;
+  void *d;
+
+  h = (unsigned char *) malloc (N);
+
+  for (i = 0; i < N; i++)
+    {
+      h[i] = i;
+    }
+
+  d = acc_present_or_copyin (h, N);
+  if (!d)
+    abort ();
+
+  memset (&h[0], 0, N);
+
+  acc_copyout (h, N);
+
+  for (i = 0; i < N; i++)
+    {
+      if (h[i] != i)
+	abort ();
+    }
+
+  free (h);
+
+  return 0;
+}
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-38.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-38.c
new file mode 100644
index 00000000000..05d8498c1f9
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-38.c
@@ -0,0 +1,64 @@
+/* { dg-do run } */
+/* { dg-skip-if "" { *-*-* } { "*" } { "-DACC_MEM_SHARED=0" } } */
+
+#include <string.h>
+#include <stdlib.h>
+#include <openacc.h>
+
+int
+main (int argc, char **argv)
+{
+  const int N = 256;
+  int i;
+  unsigned char *h;
+  void *d1, *d2;
+
+  h = (unsigned char *) malloc (N);
+
+  for (i = 0; i < N; i++)
+    {
+      h[i] = i;
+    }
+
+  d1 = acc_present_or_copyin (h, N);
+  if (!d1)
+    abort ();
+
+  for (i = 0; i < N; i++)
+    {
+      h[i] = 0xab;
+    }
+
+  d2 = acc_present_or_copyin (h, N);
+  if (!d2)
+    abort ();
+
+  if (d1 != d2)
+    abort ();
+
+  memset (&h[0], 0, N);
+
+  acc_copyout (h, N);
+
+  for (i = 0; i < N; i++)
+    {
+      if (h[i] != i)
+	abort ();
+    }
+
+  d2 = acc_pcopyin (h, N);
+  if (!d2)
+    abort ();
+
+  acc_copyout (h, N);
+
+  for (i = 0; i < N; i++)
+    {
+      if (h[i] != i)
+	abort ();
+    }
+
+  free (h);
+
+  return 0;
+}
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-39.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-39.c
new file mode 100644
index 00000000000..db1e0b3ffd5
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-39.c
@@ -0,0 +1,41 @@
+/* { dg-do run } */
+
+#include <string.h>
+#include <stdlib.h>
+#include <openacc.h>
+
+int
+main (int argc, char **argv)
+{
+  const int N = 256;
+  int i;
+  unsigned char *h;
+  void *d;
+
+  h = (unsigned char *) malloc (N);
+
+  for (i = 0; i < N; i++)
+    {
+      h[i] = i;
+    }
+
+  d = acc_present_or_copyin (0, N);
+  if (!d)
+    abort ();
+
+  memset (&h[0], 0, N);
+
+  acc_copyout (h, N);
+
+  for (i = 0; i < N; i++)
+    {
+      if (h[i] != i)
+	abort ();
+    }
+
+  free (h);
+
+  return 0;
+}
+
+/* { dg-shouldfail "libgomp: \[\(nil\),+256\] is a bad range" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-4.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-4.c
new file mode 100644
index 00000000000..060275b999b
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-4.c
@@ -0,0 +1,13 @@
+/* { dg-do run } */
+
+#include <openacc.h>
+
+int
+main (int argc, char **argv)
+{
+  acc_init ((acc_device_t) 99);
+
+  return 0;
+}
+
+/* { dg-shouldfail "libgomp: device 99 is out of range" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-40.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-40.c
new file mode 100644
index 00000000000..cb6c422872e
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-40.c
@@ -0,0 +1,42 @@
+/* { dg-do run } */
+
+#include <stdlib.h>
+#include <string.h>
+#include <unistd.h>
+#include <openacc.h>
+
+int
+main (int argc, char **argv)
+{
+  const int N = 256;
+  int i;
+  unsigned char *h;
+  void *d;
+
+  h = (unsigned char *) malloc (N);
+
+  for (i = 0; i < N; i++)
+    {
+      h[i] = i;
+    }
+
+  d = acc_present_or_copyin (h, 0);
+  if (!d)
+    abort ();
+
+  memset (&h[0], 0, N);
+
+  acc_copyout (h, N);
+
+  for (i = 0; i < N; i++)
+    {
+      if (h[i] != i)
+	abort ();
+    }
+
+  free (h);
+
+  return 0;
+}
+
+/* { dg-shouldfail "libgomp: \[\h+,\+0\] is a bad range" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-41.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-41.c
new file mode 100644
index 00000000000..01c5f3cfaa3
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-41.c
@@ -0,0 +1,43 @@
+/* { dg-do run } */
+
+#include <stdlib.h>
+#include <openacc.h>
+
+int
+main (int argc, char **argv)
+{
+  const int N = 256;
+  int i;
+  unsigned char *h;
+  void *d;
+
+  h = (unsigned char *) malloc (N);
+
+  for (i = 0; i < N; i++)
+    {
+      h[i] = i;
+    }
+
+  d = acc_copyin (h, N);
+  if (!d)
+    abort ();
+
+  for (i = 0; i < N; i++)
+    {
+      h[i] = 0xab;
+    }
+
+  acc_update_device (h, N);
+
+  acc_copyout (h, N);
+
+  for (i = 0; i < N; i++)
+    {
+      if (h[i] != 0xab)
+	abort ();
+    }
+
+  free (h);
+
+  return 0;
+}
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-42.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-42.c
new file mode 100644
index 00000000000..d577fe37e53
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-42.c
@@ -0,0 +1,35 @@
+/* { dg-do run } */
+
+#include <stdlib.h>
+#include <openacc.h>
+
+int
+main (int argc, char **argv)
+{
+  const int N = 256;
+  int i;
+  unsigned char *h;
+
+  h = (unsigned char *) malloc (N);
+
+  for (i = 0; i < N; i++)
+    {
+      h[i] = i;
+    }
+
+  acc_update_device (h, N);
+
+  acc_copyout (h, N);
+
+  for (i = 0; i < N; i++)
+    {
+      if (h[i] != 0xab)
+	abort ();
+    }
+
+  free (h);
+
+  return 0;
+}
+
+/* { dg-shouldfail "libgomp: \[\h+,256\] is not mapped" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-43.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-43.c
new file mode 100644
index 00000000000..ceeb155fe44
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-43.c
@@ -0,0 +1,45 @@
+/* { dg-do run } */
+
+#include <stdlib.h>
+#include <openacc.h>
+
+int
+main (int argc, char **argv)
+{
+  const int N = 256;
+  int i;
+  unsigned char *h;
+  void *d;
+
+  h = (unsigned char *) malloc (N);
+
+  for (i = 0; i < N; i++)
+    {
+      h[i] = i;
+    }
+
+  d = acc_copyin (h, N);
+  if (!d)
+    abort ();
+
+  for (i = 0; i < N; i++)
+    {
+      h[i] = 0xab;
+    }
+
+  acc_update_device (0, N);
+
+  acc_copyout (h, N);
+
+  for (i = 0; i < N; i++)
+    {
+      if (h[i] != 0xab)
+	abort ();
+    }
+
+  free (h);
+
+  return 0;
+}
+
+/* { dg-shouldfail "libgomp: \[\(nil\),256\] is not mapped" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-44.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-44.c
new file mode 100644
index 00000000000..0cabb0d31d9
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-44.c
@@ -0,0 +1,45 @@
+/* { dg-do run } */
+
+#include <stdlib.h>
+#include <openacc.h>
+
+int
+main (int argc, char **argv)
+{
+  const int N = 256;
+  int i;
+  unsigned char *h;
+  void *d;
+
+  h = (unsigned char *) malloc (N);
+
+  for (i = 0; i < N; i++)
+    {
+      h[i] = i;
+    }
+
+  d = acc_copyin (h, N);
+  if (!d)
+    abort ();
+
+  for (i = 0; i < N; i++)
+    {
+      h[i] = 0xab;
+    }
+
+  acc_update_device (h, 0);
+
+  acc_copyout (h, N);
+
+  for (i = 0; i < N; i++)
+    {
+      if (h[i] != 0xab)
+	abort ();
+    }
+
+  free (h);
+
+  return 0;
+}
+
+/* { dg-shouldfail "libgomp: \[\h+,0\] is not mapped" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-45.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-45.c
new file mode 100644
index 00000000000..f9a629465b4
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-45.c
@@ -0,0 +1,50 @@
+/* { dg-do run } */
+/* { dg-skip-if "" { *-*-* } { "*" } { "-DACC_MEM_SHARED=0" } } */
+
+#include <stdlib.h>
+#include <openacc.h>
+
+int
+main (int argc, char **argv)
+{
+  const int N = 256;
+  int i;
+  unsigned char *h;
+  void *d;
+
+  h = (unsigned char *) malloc (N);
+
+  for (i = 0; i < N; i++)
+    {
+      h[i] = i;
+    }
+
+  d = acc_copyin (h, N);
+  if (!d)
+    abort ();
+
+  for (i = 0; i < N; i++)
+    {
+      h[i] = 0xab;
+    }
+
+  acc_update_device (h, N - 2);
+
+  acc_copyout (h, N);
+
+  for (i = 0; i < N - 2; i++)
+    {
+      if (h[i] != 0xab)
+	abort ();
+    }
+
+  for (i = N - 2; i < N; i++)
+    {
+      if (h[i] != i)
+	abort ();
+    }
+
+  free (h);
+
+  return 0;
+}
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-46.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-46.c
new file mode 100644
index 00000000000..b195725c92c
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-46.c
@@ -0,0 +1,42 @@
+/* { dg-do run } */
+/* { dg-skip-if "" { *-*-* } { "*" } { "-DACC_MEM_SHARED=0" } } */
+
+#include <string.h>
+#include <stdlib.h>
+#include <openacc.h>
+
+int
+main (int argc, char **argv)
+{
+  const int N = 256;
+  int i;
+  unsigned char *h;
+  void *d;
+
+  h = (unsigned char *) malloc (N);
+
+  for (i = 0; i < N; i++)
+    {
+      h[i] = i;
+    }
+
+  d = acc_copyin (h, N);
+  if (!d)
+    abort ();
+
+  memset (&h[0], 0, N);
+
+  acc_update_self (h, N);
+
+  for (i = 0; i < N; i++)
+    {
+      if (h[i] != i)
+	abort ();
+    }
+
+  acc_delete (h, N);
+
+  free (h);
+
+  return 0;
+}
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-47.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-47.c
new file mode 100644
index 00000000000..a7ff904c97c
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-47.c
@@ -0,0 +1,43 @@
+/* { dg-do run } */
+
+#include <string.h>
+#include <stdlib.h>
+#include <openacc.h>
+
+int
+main (int argc, char **argv)
+{
+  const int N = 256;
+  int i;
+  unsigned char *h;
+  void *d;
+
+  h = (unsigned char *) malloc (N);
+
+  for (i = 0; i < N; i++)
+    {
+      h[i] = i;
+    }
+
+  d = acc_copyin (h, N);
+  if (!d)
+    abort ();
+
+  memset (&h[0], 0, N);
+
+  acc_update_self (0, N);
+
+  for (i = 0; i < N; i++)
+    {
+      if (h[i] != i)
+	abort ();
+    }
+
+  acc_delete (h, N);
+
+  free (h);
+
+  return 0;
+}
+
+/* { dg-shouldfail "libgomp: \[\(nil\),256\] is not mapped" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-48.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-48.c
new file mode 100644
index 00000000000..01d3c6c8280
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-48.c
@@ -0,0 +1,43 @@
+/* { dg-do run } */
+
+#include <string.h>
+#include <stdlib.h>
+#include <openacc.h>
+
+int
+main (int argc, char **argv)
+{
+  const int N = 256;
+  int i;
+  unsigned char *h;
+  void *d;
+
+  h = (unsigned char *) malloc (N);
+
+  for (i = 0; i < N; i++)
+    {
+      h[i] = i;
+    }
+
+  d = acc_copyin (h, N);
+  if (!d)
+    abort ();
+
+  memset (&h[0], 0, N);
+
+  acc_update_self (h, 0);
+
+  for (i = 0; i < N; i++)
+    {
+      if (h[i] != i)
+	abort ();
+    }
+
+  acc_delete (h, N);
+
+  free (h);
+
+  return 0;
+}
+
+/* { dg-shouldfail "libgomp: \[\h+,0\] is not mapped" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-49.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-49.c
new file mode 100644
index 00000000000..a33324c31ba
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-49.c
@@ -0,0 +1,48 @@
+/* { dg-do run } */
+/* { dg-skip-if "" { *-*-* } { "*" } { "-DACC_MEM_SHARED=0" } } */
+
+#include <string.h>
+#include <stdlib.h>
+#include <openacc.h>
+
+int
+main (int argc, char **argv)
+{
+  const int N = 256;
+  int i;
+  unsigned char *h;
+  void *d;
+
+  h = (unsigned char *) malloc (N);
+
+  for (i = 0; i < N; i++)
+    {
+      h[i] = i;
+    }
+
+  d = acc_copyin (h, N);
+  if (!d)
+    abort ();
+
+  memset (&h[0], 0, N);
+
+  acc_update_self (h, N - 2);
+
+  for (i = 0; i < N - 2; i++)
+    {
+      if (h[i] != i)
+	abort ();
+    }
+
+  for (i = N - 2; i < N; i++)
+    {
+      if (h[i] != 0)
+	abort ();
+    }
+
+  acc_delete (h, N);
+
+  free (h);
+
+  return 0;
+}
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-5.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-5.c
new file mode 100644
index 00000000000..961a62c2b63
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-5.c
@@ -0,0 +1,40 @@
+/* { dg-do run } */
+
+#include <stdlib.h>
+#include <openacc.h>
+
+int
+main (int argc, char **argv)
+{
+  if (acc_get_device_type () == acc_device_default)
+    abort ();
+
+  acc_init (acc_device_default);
+
+  if (acc_get_device_type () == acc_device_default)
+    abort ();
+
+  acc_shutdown (acc_device_default);
+
+  if (acc_get_num_devices (acc_device_nvidia) != 0)
+    {
+      acc_init (acc_device_nvidia);
+
+      if (acc_get_device_type () != acc_device_nvidia)
+        abort ();
+
+      acc_shutdown (acc_device_nvidia);
+
+      acc_init (acc_device_default);
+
+      acc_set_device_type (acc_device_nvidia);
+
+      if (acc_get_device_type () != acc_device_nvidia)
+        abort ();
+
+      acc_shutdown (acc_device_nvidia);
+    }
+
+  return 0;
+
+}
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-50.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-50.c
new file mode 100644
index 00000000000..e8294e1af36
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-50.c
@@ -0,0 +1,30 @@
+/* { dg-do run } */
+/* { dg-skip-if "" { *-*-* } { "*" } { "-DACC_MEM_SHARED=0" } } */
+
+#include <stdlib.h>
+#include <openacc.h>
+
+int
+main (int argc, char **argv)
+{
+  const int N = 256;
+  unsigned char *h;
+  void *d;
+
+  h = (unsigned char *) malloc (N);
+
+  d = acc_malloc (N);
+
+  acc_map_data (h, d, N);
+
+  if (acc_is_present (h, N) != 1)
+    abort ();
+
+  acc_unmap_data (h);
+
+  acc_free (d);
+
+  free (h);
+
+  return 0;
+}
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-51.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-51.c
new file mode 100644
index 00000000000..29d28f298c0
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-51.c
@@ -0,0 +1,41 @@
+/* { dg-do run } */
+/* { dg-skip-if "" { *-*-* } { "*" } { "-DACC_MEM_SHARED=0" } } */
+
+#include <stdlib.h>
+#include <openacc.h>
+
+int
+main (int argc, char **argv)
+{
+  const int N = 256;
+  int i;
+  unsigned char *h[N];
+  void *d[N];
+
+  for (i = 0; i < N; i++)
+    {
+      h[i] = (unsigned char *) malloc (N);
+      d[i] = acc_malloc (N);
+
+      acc_map_data (h[i], d[i], N);
+    }
+
+  for (i = 0; i < N; i++)
+    {
+      if (acc_is_present (h[i], N) != 1)
+	abort ();
+    }
+
+  for (i = 0; i < N; i++)
+    {
+      acc_unmap_data (h[i]);
+
+      if (acc_is_present (h[i], N) != 0)
+	abort ();
+
+      acc_free (d[i]);
+      free (h[i]);
+    }
+
+  return 0;
+}
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-52.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-52.c
new file mode 100644
index 00000000000..780db310fa1
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-52.c
@@ -0,0 +1,28 @@
+/* { dg-do run } */
+
+#include <stdlib.h>
+#include <openacc.h>
+
+int
+main (int argc, char **argv)
+{
+  const int N = 256;
+  unsigned char *h;
+  void *d;
+
+  h = (unsigned char *) malloc (N);
+
+  d = acc_malloc (N);
+
+  acc_map_data (0, d, N);
+
+  acc_unmap_data (h);
+
+  acc_free (d);
+
+  free (h);
+
+  return 0;
+}
+
+/* { dg-shouldfail "libgomp: \[(nil),+256\]->\[\h+,\+256\] is a bad map" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-53.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-53.c
new file mode 100644
index 00000000000..657adde3b02
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-53.c
@@ -0,0 +1,28 @@
+/* { dg-do run } */
+
+#include <stdlib.h>
+#include <openacc.h>
+
+int
+main (int argc, char **argv)
+{
+  const int N = 256;
+  unsigned char *h;
+  void *d;
+
+  h = (unsigned char *) malloc (N);
+
+  d = acc_malloc (N);
+
+  acc_map_data (h, 0, N);
+
+  acc_unmap_data (h);
+
+  acc_free (d);
+
+  free (h);
+
+  return 0;
+}
+
+/* { dg-shouldfail "libgomp: \[\h+,\+256\]->\[(nil),\+256\] is a bad map" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-54.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-54.c
new file mode 100644
index 00000000000..1f3df80bd41
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-54.c
@@ -0,0 +1,28 @@
+/* { dg-do run } */
+
+#include <stdlib.h>
+#include <openacc.h>
+
+int
+main (int argc, char **argv)
+{
+  const int N = 256;
+  unsigned char *h;
+  void *d;
+
+  h = (unsigned char *) malloc (N);
+
+  d = acc_malloc (N);
+
+  acc_map_data (h, d, 0);
+
+  acc_unmap_data (h);
+
+  acc_free (d);
+
+  free (h);
+
+  return 0;
+}
+
+/* { dg-shouldfail "libgomp: \[\h+,\+0\]->\[\h+,\+0\] is a bad map" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-55.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-55.c
new file mode 100644
index 00000000000..286653f3339
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-55.c
@@ -0,0 +1,48 @@
+/* { dg-do run } */
+/* { dg-skip-if "" { *-*-* } { "*" } { "-DACC_MEM_SHARED=0" } } */
+
+#include <stdlib.h>
+#include <openacc.h>
+#include <stdint.h>
+
+int
+main (int argc, char **argv)
+{
+  const int N = 256;
+  unsigned char *h;
+  int i;
+  void *d;
+
+  h = (unsigned char *) malloc (N);
+
+  d = acc_malloc (N);
+
+  for (i = 0; i < N; i++)
+    {
+      acc_map_data ((void *)((uintptr_t) h + (uintptr_t) i),
+                    				(void *)((uintptr_t) d + (uintptr_t) i), 1);
+    }
+
+  for (i = 0; i < N; i++)
+    {
+      if (acc_is_present (h + 1, 1) != 1)
+	abort ();
+    }
+
+  for (i = 0; i < N; i++)
+    {
+      acc_unmap_data (h + i);
+    }
+
+  for (i = 0; i < N; i++)
+    {
+      if (acc_is_present (h + 1, 1) != 0)
+	abort ();
+    }
+
+  acc_free (d);
+
+  free (h);
+
+  return 0;
+}
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-56.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-56.c
new file mode 100644
index 00000000000..e3f5a80518f
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-56.c
@@ -0,0 +1,33 @@
+/* { dg-do run } */
+/* { dg-skip-if "" { *-*-* } { "*" } { "-DACC_MEM_SHARED=0" } } */
+
+#include <stdlib.h>
+#include <openacc.h>
+
+int
+main (int argc, char **argv)
+{
+  const int N = 256;
+  unsigned char *h;
+  void *d;
+
+  h = (unsigned char *) malloc (N);
+
+  d = acc_malloc (N);
+
+  acc_map_data (h, d, N >> 1);
+
+  if (acc_is_present (h, 1) != 1)
+    abort ();
+
+  if (acc_is_present (h + (N >> 1), 1) != 0)
+    abort ();
+
+  acc_unmap_data (h);
+
+  acc_free (d);
+
+  free (h);
+
+  return 0;
+}
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-57.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-57.c
new file mode 100644
index 00000000000..f9043a445df
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-57.c
@@ -0,0 +1,28 @@
+/* { dg-do run } */
+
+#include <stdlib.h>
+#include <openacc.h>
+
+int
+main (int argc, char **argv)
+{
+  const int N = 256;
+  unsigned char *h;
+  void *d;
+
+  h = (unsigned char *) malloc (N);
+
+  d = acc_malloc (N);
+
+  acc_map_data (h, d, N);
+
+  acc_unmap_data (d);
+
+  acc_free (d);
+
+  free (h);
+
+  return 0;
+}
+
+/* { dg-shouldfail "libgomp: \h+ is not a mapped block" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-58.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-58.c
new file mode 100644
index 00000000000..9d6e27d89e9
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-58.c
@@ -0,0 +1,28 @@
+/* { dg-do run } */
+
+#include <stdlib.h>
+#include <openacc.h>
+
+int
+main (int argc, char **argv)
+{
+  const int N = 256;
+  unsigned char *h;
+  void *d;
+
+  h = (unsigned char *) malloc (N);
+
+  d = acc_malloc (N);
+
+  acc_map_data (h, d, N);
+
+  acc_unmap_data (0);
+
+  acc_free (d);
+
+  free (h);
+
+  return 0;
+}
+
+/* { dg-shouldfail "libgomp: \(nil\) is not a mapped block" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-59.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-59.c
new file mode 100644
index 00000000000..2f087aedccb
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-59.c
@@ -0,0 +1,55 @@
+/* { dg-do run } */
+/* { dg-skip-if "" { *-*-* } { "*" } { "-DACC_MEM_SHARED=0" } } */
+
+#include <stdlib.h>
+#include <openacc.h>
+#include <stdint.h>
+
+int
+main (int argc, char **argv)
+{
+  const int N = 256;
+  int i;
+  unsigned char *h;
+  void *d;
+
+  h = (unsigned char *) malloc (N);
+
+  d = acc_malloc (N);
+
+  acc_map_data (h, d, N);
+
+  for (i = 0; i < N; i++)
+    {
+      if (acc_hostptr ((void *)((uintptr_t) d + (uintptr_t) i)) !=
+                            (void *)((uintptr_t) h + (uintptr_t) i))
+	abort ();
+    }
+
+  for (i = 0; i < N; i++)
+    {
+      if (acc_deviceptr ((void *)((uintptr_t) h + (uintptr_t) i)) !=
+                            (void *)((uintptr_t) d + (uintptr_t) i))
+	abort ();
+    }
+
+  acc_unmap_data (h);
+
+  for (i = 0; i < N; i++)
+    {
+      if (acc_hostptr ((void *)((uintptr_t) d + (uintptr_t) i)) != 0)
+	abort ();
+    }
+
+  for (i = 0; i < N; i++)
+    {
+      if (acc_deviceptr (h + i) != 0)
+	abort ();
+    }
+
+  acc_free (d);
+
+  free (h);
+
+  return 0;
+}
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-6.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-6.c
new file mode 100644
index 00000000000..afdd480cb71
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-6.c
@@ -0,0 +1,39 @@
+/* { dg-do run } */
+
+#include <stdlib.h>
+#include <openacc.h>
+
+int
+main (int argc, char **argv)
+{
+  int devnum;
+
+  if (acc_get_device_type () == acc_device_default)
+    abort ();
+
+  if (acc_get_num_devices (acc_device_nvidia) == 0)
+    return 0;
+
+  acc_set_device_type (acc_device_nvidia);
+
+  if (acc_get_device_type () != acc_device_nvidia)
+    abort ();
+
+  acc_shutdown (acc_device_nvidia);
+
+  acc_set_device_type (acc_device_nvidia);
+
+  if (acc_get_device_type () != acc_device_nvidia)
+    abort ();
+
+  devnum = acc_get_num_devices (acc_device_host);
+  if (devnum != 1)
+    abort ();
+
+  acc_shutdown (acc_device_nvidia);
+
+  if (acc_get_device_type () == acc_device_default)
+    abort ();
+
+  return 0;
+}
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-60.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-60.c
new file mode 100644
index 00000000000..ccae728e3dc
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-60.c
@@ -0,0 +1,54 @@
+/* { dg-do run } */
+/* { dg-skip-if "" { *-*-* } { "*" } { "-DACC_MEM_SHARED=0" } } */
+
+#include <string.h>
+#include <stdlib.h>
+#include <openacc.h>
+
+int
+main (int argc, char **argv)
+{
+  const int N = 256;
+  int i;
+  unsigned char *h;
+  void *d;
+
+  h = (unsigned char *) malloc (N);
+
+  for (i = 0; i < N; i++)
+    {
+      h[i] = i;
+    }
+
+  d = acc_malloc (N);
+
+  acc_memcpy_to_device (d, h, N);
+
+  for (i = 0; i < N; i++)
+    {
+      if (acc_is_present (h + i, 1) != 0)
+	abort ();
+    }
+
+  memset (&h[0], 0, N);
+
+  acc_memcpy_from_device (h, d, N);
+
+  for (i = 0; i < N; i++)
+    {
+      if (h[i] != i)
+	abort ();
+    }
+
+  for (i = 0; i < N; i++)
+    {
+      if (acc_is_present (h + i, 1) != 0)
+	abort ();
+    }
+
+  acc_free (d);
+
+  free (h);
+
+  return 0;
+}
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-61.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-61.c
new file mode 100644
index 00000000000..ce66ced28e8
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-61.c
@@ -0,0 +1,70 @@
+/* { dg-do run } */
+/* { dg-skip-if "" { *-*-* } { "*" } { "-DACC_MEM_SHARED=0" } } */
+
+#include <string.h>
+#include <stdlib.h>
+#include <openacc.h>
+
+int
+main (int argc, char **argv)
+{
+  const int N = 256;
+  int i;
+  unsigned char *h[N];
+  void *d[N];
+
+  for (i = 0; i < N; i++)
+    {
+      int j;
+      unsigned char *p;
+
+      h[i] = (unsigned char *) malloc (N);
+
+      p = h[i];
+
+      for (j = 0; j < N; j++)
+	{
+	  p[j] = i;
+	}
+
+      d[i] = acc_malloc (N);
+
+      acc_memcpy_to_device (d[i], h[i], N);
+
+      for (j = 0; j < N; j++)
+	{
+	  if (acc_is_present (h[i] + j, 1) != 0)
+	    abort ();
+	}
+    }
+
+  for (i = 0; i < N; i++)
+    {
+      int j;
+      unsigned char *p;
+
+      memset (h[i], 0, N);
+
+      acc_memcpy_from_device (h[i], d[i], N);
+
+      p = h[i];
+
+      for (j = 0; j < N; j++)
+	{
+	  if (p[j] != i)
+	    abort ();
+	}
+
+      for (j = 0; j < N; j++)
+	{
+	  if (acc_is_present (h[i] + j, 1) != 0)
+	    abort ();
+	}
+
+      acc_free (d[i]);
+
+      free (h[i]);
+    }
+
+  return 0;
+}
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-62.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-62.c
new file mode 100644
index 00000000000..e6178e233ed
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-62.c
@@ -0,0 +1,49 @@
+/* { dg-do run } */
+
+#include <string.h>
+#include <stdlib.h>
+#include <openacc.h>
+
+int
+main (int argc, char **argv)
+{
+  const int N = 256;
+  int i;
+  unsigned char *h;
+  void *d;
+
+  acc_init (acc_device_nvidia);
+
+  h = (unsigned char *) malloc (N);
+
+  for (i = 0; i < N; i++)
+    {
+      h[i] = i;
+    }
+
+  d = acc_malloc (N);
+
+  acc_memcpy_to_device (d, h, N);
+
+  memset (&h[0], 0, N);
+
+  acc_memcpy_to_device (d, h, N << 1);
+
+  acc_memcpy_from_device (h, d, N);
+
+  for (i = 0; i < N; i++)
+    {
+      if (h[i] != i)
+	abort ();
+    }
+
+  acc_free (d);
+
+  free (h);
+
+  acc_shutdown (acc_device_nvidia);
+
+  return 0;
+}
+
+/* { dg-shouldfail "libgomp: invalid size" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-63.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-63.c
new file mode 100644
index 00000000000..ca237ec1b60
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-63.c
@@ -0,0 +1,43 @@
+/* { dg-do run } */
+
+#include <string.h>
+#include <stdlib.h>
+#include <openacc.h>
+
+int
+main (int argc, char **argv)
+{
+  const int N = 256;
+  int i;
+  unsigned char *h;
+  void *d;
+
+  h = (unsigned char *) malloc (N);
+
+  for (i = 0; i < N; i++)
+    {
+      h[i] = i;
+    }
+
+  d = acc_malloc (N);
+
+  acc_memcpy_to_device (0, h, N);
+
+  memset (&h[0], 0, N);
+
+  acc_memcpy_from_device (h, d, N);
+
+  for (i = 0; i < N; i++)
+    {
+      if (h[i] != i)
+	abort ();
+    }
+
+  acc_free (d);
+
+  free (h);
+
+  return 0;
+}
+
+/* { dg-shouldfail "libgomp: invalid device address" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-64.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-64.c
new file mode 100644
index 00000000000..850fd2e3f91
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-64.c
@@ -0,0 +1,43 @@
+/* { dg-do run } */
+
+#include <string.h>
+#include <stdlib.h>
+#include <openacc.h>
+
+int
+main (int argc, char **argv)
+{
+  const int N = 256;
+  int i;
+  unsigned char *h;
+  void *d;
+
+  h = (unsigned char *) malloc (N);
+
+  for (i = 0; i < N; i++)
+    {
+      h[i] = i;
+    }
+
+  d = acc_malloc (N);
+
+  acc_memcpy_to_device (d, 0, N);
+
+  memset (&h[0], 0, N);
+
+  acc_memcpy_from_device (h, d, N);
+
+  for (i = 0; i < N; i++)
+    {
+      if (h[i] != i)
+	abort ();
+    }
+
+  acc_free (d);
+
+  free (h);
+
+  return 0;
+}
+
+/* { dg-shouldfail "libgomp: invalid host address" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-65.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-65.c
new file mode 100644
index 00000000000..26c8cefdc42
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-65.c
@@ -0,0 +1,43 @@
+/* { dg-do run } */
+
+#include <string.h>
+#include <stdlib.h>
+#include <openacc.h>
+
+int
+main (int argc, char **argv)
+{
+  const int N = 256;
+  int i;
+  unsigned char *h;
+  void *d;
+
+  h = (unsigned char *) malloc (N);
+
+  for (i = 0; i < N; i++)
+    {
+      h[i] = i;
+    }
+
+  d = acc_malloc (N);
+
+  acc_memcpy_to_device (d, d, N);
+
+  memset (&h[0], 0, N);
+
+  acc_memcpy_from_device (h, d, N);
+
+  for (i = 0; i < N; i++)
+    {
+      if (h[i] != i)
+	abort ();
+    }
+
+  acc_free (d);
+
+  free (h);
+
+  return 0;
+}
+
+/* { dg-shouldfail "libgomp: invalid host or device address" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-66.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-66.c
new file mode 100644
index 00000000000..398dc2a4cef
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-66.c
@@ -0,0 +1,48 @@
+/* { dg-do run } */
+/* { dg-skip-if "" { *-*-* } { "*" } { "-DACC_MEM_SHARED=0" } } */
+
+#include <string.h>
+#include <stdlib.h>
+#include <openacc.h>
+
+int
+main (int argc, char **argv)
+{
+  const int N = 256;
+  int i;
+  unsigned char *h;
+  void *d;
+
+  acc_init (acc_device_default);
+
+  h = (unsigned char *) malloc (N);
+
+  for (i = 0; i < N; i++)
+    {
+      h[i] = i;
+    }
+
+  d = acc_malloc (N);
+
+  acc_memcpy_to_device (d, h, N);
+
+  memset (&h[0], 0, N);
+
+  acc_memcpy_to_device (d, h, 0);
+
+  acc_memcpy_from_device (h, d, N);
+
+  for (i = 0; i < N; i++)
+    {
+      if (h[i] != i)
+	abort ();
+    }
+
+  acc_free (d);
+
+  free (h);
+
+  acc_shutdown (acc_device_default);
+
+  return 0;
+}
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-67.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-67.c
new file mode 100644
index 00000000000..01b8b2d89ae
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-67.c
@@ -0,0 +1,43 @@
+/* { dg-do run } */
+
+#include <string.h>
+#include <stdlib.h>
+#include <openacc.h>
+
+int
+main (int argc, char **argv)
+{
+  const int N = 256;
+  int i;
+  unsigned char *h;
+  void *d;
+
+  h = (unsigned char *) malloc (N);
+
+  for (i = 0; i < N; i++)
+    {
+      h[i] = i;
+    }
+
+  d = acc_malloc (N);
+
+  acc_memcpy_to_device (d, h, N);
+
+  memset (&h[0], 0, N);
+
+  acc_memcpy_from_device (0, d, N);
+
+  for (i = 0; i < N; i++)
+    {
+      if (h[i] != i)
+	abort ();
+    }
+
+  acc_free (d);
+
+  free (h);
+
+  return 0;
+}
+
+/* { dg-shouldfail "libgomp: invalid host address" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-68.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-68.c
new file mode 100644
index 00000000000..3ff5bd7bcf7
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-68.c
@@ -0,0 +1,43 @@
+/* { dg-do run } */
+
+#include <string.h>
+#include <stdlib.h>
+#include <openacc.h>
+
+int
+main (int argc, char **argv)
+{
+  const int N = 256;
+  int i;
+  unsigned char *h;
+  void *d;
+
+  h = (unsigned char *) malloc (N);
+
+  for (i = 0; i < N; i++)
+    {
+      h[i] = i;
+    }
+
+  d = acc_malloc (N);
+
+  acc_memcpy_to_device (d, h, N);
+
+  memset (&h[0], 0, N);
+
+  acc_memcpy_from_device (h, 0, N);
+
+  for (i = 0; i < N; i++)
+    {
+      if (h[i] != i)
+	abort ();
+    }
+
+  acc_free (d);
+
+  free (h);
+
+  return 0;
+}
+
+/* { dg-shouldfail "libgomp: invalid device address" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-69.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-69.c
new file mode 100644
index 00000000000..5462f125352
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-69.c
@@ -0,0 +1,124 @@
+/* { dg-do run { target openacc_nvidia_accel_selected } } */
+/* { dg-additional-options "-lcuda" } */
+
+#include <stdio.h>
+#include <unistd.h>
+#include <openacc.h>
+#include <cuda.h>
+
+int
+main (int argc, char **argv)
+{
+  CUdevice dev;
+  CUfunction delay;
+  CUmodule module;
+  CUresult r;
+  CUstream stream;
+  unsigned long *a, *d_a, dticks;
+  int nbytes;
+  float dtime;
+  void *kargs[2];
+  int clkrate;
+  int devnum, nprocs;
+
+  acc_init (acc_device_nvidia);
+
+  devnum = acc_get_device_num (acc_device_nvidia);
+
+  r = cuDeviceGet (&dev, devnum);
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuDeviceGet failed: %d\n", r);
+      abort ();
+    }
+
+  r =
+    cuDeviceGetAttribute (&nprocs, CU_DEVICE_ATTRIBUTE_MULTIPROCESSOR_COUNT,
+			  dev);
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuDeviceGetAttribute failed: %d\n", r);
+      abort ();
+    }
+
+  r = cuDeviceGetAttribute (&clkrate, CU_DEVICE_ATTRIBUTE_CLOCK_RATE, dev);
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuDeviceGetAttribute failed: %d\n", r);
+      abort ();
+    }
+
+  r = cuModuleLoad (&module, "subr.ptx");
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuModuleLoad failed: %d\n", r);
+      abort ();
+    }
+
+  r = cuModuleGetFunction (&delay, module, "delay");
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuModuleGetFunction failed: %d\n", r);
+      abort ();
+    }
+
+  nbytes = nprocs * sizeof (unsigned long);
+
+  dtime = 200.0;
+
+  dticks = (unsigned long) (dtime * clkrate);
+
+  a = (unsigned long *) malloc (nbytes);
+  d_a = (unsigned long *) acc_malloc (nbytes);
+
+  acc_map_data (a, d_a, nbytes);
+
+  kargs[0] = (void *) &d_a;
+  kargs[1] = (void *) &dticks;
+
+  stream = (CUstream) acc_get_cuda_stream (0);
+  if (stream != NULL)
+    abort ();
+
+  r = cuStreamCreate (&stream, CU_STREAM_DEFAULT);
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuStreamCreate failed: %d\n", r);
+      abort ();
+    }
+
+  if (!acc_set_cuda_stream (0, stream))
+    abort ();
+
+  r = cuLaunchKernel (delay, 1, 1, 1, 1, 1, 1, 0, stream, kargs, 0);
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuLaunchKernel failed: %d\n", r);
+      abort ();
+    }
+
+  if (acc_async_test (0) != 0)
+    {
+      fprintf (stderr, "asynchronous operation not running\n");
+      abort ();
+    }
+
+  sleep (1);
+
+  if (acc_async_test (0) != 1)
+    {
+      fprintf (stderr, "found asynchronous operation still running\n");
+      abort ();
+    }
+
+  acc_unmap_data (a);
+
+  free (a);
+  acc_free (d_a);
+
+  acc_shutdown (acc_device_nvidia);
+
+  exit (0);
+}
+
+/* { dg-output "" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-7.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-7.c
new file mode 100644
index 00000000000..e78734b3011
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-7.c
@@ -0,0 +1,18 @@
+/* { dg-do run } */
+
+#include <stdlib.h>
+#include <openacc.h>
+
+int
+main (int argc, char **argv)
+{
+  if (acc_get_num_devices (acc_device_none) != 0)
+    abort ();
+
+  if (acc_get_num_devices (acc_device_host) == 0)
+    abort ();
+
+  return 0;
+}
+
+/* { dg-output "" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-70.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-70.c
new file mode 100644
index 00000000000..912b266ec39
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-70.c
@@ -0,0 +1,136 @@
+/* { dg-do run { target openacc_nvidia_accel_selected } } */
+/* { dg-additional-options "-lcuda" } */
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <unistd.h>
+#include <openacc.h>
+#include <cuda.h>
+
+int
+main (int argc, char **argv)
+{
+  CUdevice dev;
+  CUfunction delay;
+  CUmodule module;
+  CUresult r;
+  const int N = 10;
+  int i;
+  CUstream streams[N];
+  unsigned long *a, *d_a, dticks;
+  int nbytes;
+  float dtime;
+  void *kargs[2];
+  int clkrate;
+  int devnum, nprocs;
+
+  acc_init (acc_device_nvidia);
+
+  devnum = acc_get_device_num (acc_device_nvidia);
+
+  r = cuDeviceGet (&dev, devnum);
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuDeviceGet failed: %d\n", r);
+      abort ();
+    }
+
+  r =
+    cuDeviceGetAttribute (&nprocs, CU_DEVICE_ATTRIBUTE_MULTIPROCESSOR_COUNT,
+			  dev);
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuDeviceGetAttribute failed: %d\n", r);
+      abort ();
+    }
+
+  r = cuDeviceGetAttribute (&clkrate, CU_DEVICE_ATTRIBUTE_CLOCK_RATE, dev);
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuDeviceGetAttribute failed: %d\n", r);
+      abort ();
+    }
+
+  r = cuModuleLoad (&module, "subr.ptx");
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuModuleLoad failed: %d\n", r);
+      abort ();
+    }
+
+  r = cuModuleGetFunction (&delay, module, "delay");
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuModuleGetFunction failed: %d\n", r);
+      abort ();
+    }
+
+  nbytes = nprocs * sizeof (unsigned long);
+
+  dtime = 200.0;
+
+  dticks = (unsigned long) (dtime * clkrate);
+
+  a = (unsigned long *) malloc (nbytes);
+  d_a = (unsigned long *) acc_malloc (nbytes);
+
+  acc_map_data (a, d_a, nbytes);
+
+  kargs[0] = (void *) &d_a;
+  kargs[1] = (void *) &dticks;
+
+  for (i = 0; i < N; i++)
+    {
+      streams[i] = (CUstream) acc_get_cuda_stream (i);
+      if (streams[i] != NULL)
+	abort ();
+
+      r = cuStreamCreate (&streams[i], CU_STREAM_DEFAULT);
+      if (r != CUDA_SUCCESS)
+	{
+	  fprintf (stderr, "cuStreamCreate failed: %d\n", r);
+	  abort ();
+	}
+
+        if (!acc_set_cuda_stream (i, streams[i]))
+	  abort ();
+    }
+
+  for (i = 0; i < N; i++)
+    {
+      r = cuLaunchKernel (delay, 1, 1, 1, 1, 1, 1, 0, streams[i], kargs, 0);
+      if (r != CUDA_SUCCESS)
+	{
+	  fprintf (stderr, "cuLaunchKernel failed: %d\n", r);
+	  abort ();
+	}
+
+      if (acc_async_test (i) != 0)
+	{
+	  fprintf (stderr, "asynchronous operation not running\n");
+	  abort ();
+	}
+    }
+
+  sleep ((int) (dtime / 1000.0f) + 1);
+
+  for (i = 0; i < N; i++)
+    {
+      if (acc_async_test (i) != 1)
+	{
+	  fprintf (stderr, "found asynchronous operation still running\n");
+	  abort ();
+	}
+    }
+
+  acc_unmap_data (a);
+
+  free (a);
+  acc_free (d_a);
+
+  acc_shutdown (acc_device_nvidia);
+
+  exit (0);
+}
+
+/* { dg-output "" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-71.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-71.c
new file mode 100644
index 00000000000..a04537916f6
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-71.c
@@ -0,0 +1,119 @@
+/* { dg-do run { target openacc_nvidia_accel_selected } } */
+/* { dg-additional-options "-lcuda" } */
+
+#include <stdio.h>
+#include <unistd.h>
+#include <openacc.h>
+#include <cuda.h>
+
+int
+main (int argc, char **argv)
+{
+  CUdevice dev;
+  CUfunction delay;
+  CUmodule module;
+  CUresult r;
+  CUstream stream;
+  unsigned long *a, *d_a, dticks;
+  int nbytes;
+  float dtime;
+  void *kargs[2];
+  int clkrate;
+  int devnum, nprocs;
+
+  acc_init (acc_device_nvidia);
+
+  devnum = acc_get_device_num (acc_device_nvidia);
+
+  r = cuDeviceGet (&dev, devnum);
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuDeviceGet failed: %d\n", r);
+      abort ();
+    }
+
+  r =
+    cuDeviceGetAttribute (&nprocs, CU_DEVICE_ATTRIBUTE_MULTIPROCESSOR_COUNT,
+			  dev);
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuDeviceGetAttribute failed: %d\n", r);
+      abort ();
+    }
+
+  r = cuDeviceGetAttribute (&clkrate, CU_DEVICE_ATTRIBUTE_CLOCK_RATE, dev);
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuDeviceGetAttribute failed: %d\n", r);
+      abort ();
+    }
+
+  r = cuModuleLoad (&module, "subr.ptx");
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuModuleLoad failed: %d\n", r);
+      abort ();
+    }
+
+  r = cuModuleGetFunction (&delay, module, "delay");
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuModuleGetFunction failed: %d\n", r);
+      abort ();
+    }
+
+  nbytes = nprocs * sizeof (unsigned long);
+
+  dtime = 200.0;
+
+  dticks = (unsigned long) (dtime * clkrate);
+
+  a = (unsigned long *) malloc (nbytes);
+  d_a = (unsigned long *) acc_malloc (nbytes);
+
+  acc_map_data (a, d_a, nbytes);
+
+  kargs[0] = (void *) &d_a;
+  kargs[1] = (void *) &dticks;
+
+  r = cuStreamCreate (&stream, CU_STREAM_DEFAULT);
+  if (r != CUDA_SUCCESS)
+	{
+	  fprintf (stderr, "cuStreamCreate failed: %d\n", r);
+	  abort ();
+	}
+
+  acc_set_cuda_stream (0, stream);
+
+  r = cuLaunchKernel (delay, 1, 1, 1, 1, 1, 1, 0, stream, kargs, 0);
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuLaunchKernel failed: %d\n", r);
+      abort ();
+    }
+
+  if (acc_async_test (1) != 0)
+    {
+      fprintf (stderr, "asynchronous operation not running\n");
+      abort ();
+    }
+
+  sleep ((int) (dtime / 1000.0f) + 1);
+
+  if (acc_async_test (1) != 1)
+    {
+      fprintf (stderr, "found asynchronous operation still running\n");
+      abort ();
+    }
+
+  acc_unmap_data (a);
+
+  free (a);
+  acc_free (d_a);
+
+  acc_shutdown (acc_device_nvidia);
+
+  return 0;
+}
+
+/* { dg-shouldfail "libgomp: unknown async \d" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-72.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-72.c
new file mode 100644
index 00000000000..e383ba04d69
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-72.c
@@ -0,0 +1,121 @@
+/* { dg-do run { target openacc_nvidia_accel_selected } } */
+/* { dg-additional-options "-lcuda" } */
+
+#include <stdio.h>
+#include <unistd.h>
+#include <stdlib.h>
+#include <openacc.h>
+#include <cuda.h>
+
+int
+main (int argc, char **argv)
+{
+  CUdevice dev;
+  CUfunction delay;
+  CUmodule module;
+  CUresult r;
+  CUstream stream;
+  unsigned long *a, *d_a, dticks;
+  int nbytes;
+  float dtime;
+  void *kargs[2];
+  int clkrate;
+  int devnum, nprocs;
+
+  acc_init (acc_device_nvidia);
+
+  devnum = acc_get_device_num (acc_device_nvidia);
+
+  r = cuDeviceGet (&dev, devnum);
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuDeviceGet failed: %d\n", r);
+      abort ();
+    }
+
+  r =
+    cuDeviceGetAttribute (&nprocs, CU_DEVICE_ATTRIBUTE_MULTIPROCESSOR_COUNT,
+			  dev);
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuDeviceGetAttribute failed: %d\n", r);
+      abort ();
+    }
+
+  r = cuDeviceGetAttribute (&clkrate, CU_DEVICE_ATTRIBUTE_CLOCK_RATE, dev);
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuDeviceGetAttribute failed: %d\n", r);
+      abort ();
+    }
+
+  r = cuModuleLoad (&module, "subr.ptx");
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuModuleLoad failed: %d\n", r);
+      abort ();
+    }
+
+  r = cuModuleGetFunction (&delay, module, "delay");
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuModuleGetFunction failed: %d\n", r);
+      abort ();
+    }
+
+  nbytes = nprocs * sizeof (unsigned long);
+
+  dtime = 200.0;
+
+  dticks = (unsigned long) (dtime * clkrate);
+
+  a = (unsigned long *) malloc (nbytes);
+  d_a = (unsigned long *) acc_malloc (nbytes);
+
+  acc_map_data (a, d_a, nbytes);
+
+  kargs[0] = (void *) &d_a;
+  kargs[1] = (void *) &dticks;
+
+  r = cuStreamCreate (&stream, CU_STREAM_DEFAULT);
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuStreamCreate failed: %d\n", r);
+      abort ();
+    }
+
+  if (!acc_set_cuda_stream (0, stream))
+    abort ();
+    
+  r = cuLaunchKernel (delay, 1, 1, 1, 1, 1, 1, 0, stream, kargs, 0);
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuLaunchKernel failed: %d\n", r);
+      abort ();
+    }
+
+  if (acc_async_test_all () != 0)
+    {
+      fprintf (stderr, "asynchronous operation not running\n");
+      abort ();
+    }
+
+  sleep ((int) (dtime / 1000.f) + 1);
+
+  if (acc_async_test_all () != 1)
+    {
+      fprintf (stderr, "found asynchronous operation still running\n");
+      abort ();
+    }
+
+  acc_unmap_data (a);
+
+  free (a);
+  acc_free (d_a);
+
+  acc_shutdown (acc_device_nvidia);
+
+  exit (0);
+}
+
+/* { dg-output "" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-73.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-73.c
new file mode 100644
index 00000000000..43a8b7e6395
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-73.c
@@ -0,0 +1,134 @@
+/* { dg-do run { target openacc_nvidia_accel_selected } } */
+/* { dg-additional-options "-lcuda" } */
+
+#include <stdio.h>
+#include <unistd.h>
+#include <stdlib.h>
+#include <openacc.h>
+#include <cuda.h>
+
+int
+main (int argc, char **argv)
+{
+  CUdevice dev;
+  CUfunction delay;
+  CUmodule module;
+  CUresult r;
+  const int N = 10;
+  int i;
+  CUstream streams[N];
+  unsigned long *a, *d_a, dticks;
+  int nbytes;
+  float dtime;
+  void *kargs[2];
+  int clkrate;
+  int devnum, nprocs;
+
+  acc_init (acc_device_nvidia);
+
+  devnum = acc_get_device_num (acc_device_nvidia);
+
+  r = cuDeviceGet (&dev, devnum);
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuDeviceGet failed: %d\n", r);
+      abort ();
+    }
+
+  r =
+    cuDeviceGetAttribute (&nprocs, CU_DEVICE_ATTRIBUTE_MULTIPROCESSOR_COUNT,
+			  dev);
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuDeviceGetAttribute failed: %d\n", r);
+      abort ();
+    }
+
+  r = cuDeviceGetAttribute (&clkrate, CU_DEVICE_ATTRIBUTE_CLOCK_RATE, dev);
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuDeviceGetAttribute failed: %d\n", r);
+      abort ();
+    }
+
+  r = cuModuleLoad (&module, "subr.ptx");
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuModuleLoad failed: %d\n", r);
+      abort ();
+    }
+
+  r = cuModuleGetFunction (&delay, module, "delay");
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuModuleGetFunction failed: %d\n", r);
+      abort ();
+    }
+
+  nbytes = nprocs * sizeof (unsigned long);
+
+  dtime = 200.0;
+
+  dticks = (unsigned long) (dtime * clkrate);
+
+  a = (unsigned long *) malloc (nbytes);
+  d_a = (unsigned long *) acc_malloc (nbytes);
+
+  acc_map_data (a, d_a, nbytes);
+
+  kargs[0] = (void *) &d_a;
+  kargs[1] = (void *) &dticks;
+
+  for (i = 0; i < N; i++)
+    {
+      streams[i] = (CUstream) acc_get_cuda_stream (i);
+      if (streams[i] != NULL)
+	abort ();
+
+      r = cuStreamCreate (&streams[i], CU_STREAM_DEFAULT);
+      if (r != CUDA_SUCCESS)
+	{
+	  fprintf (stderr, "cuStreamCreate failed: %d\n", r);
+	  abort ();
+	}
+
+        if (!acc_set_cuda_stream (i, streams[i]))
+	  abort ();
+    }
+
+  for (i = 0; i < N; i++)
+    {
+      r = cuLaunchKernel (delay, 1, 1, 1, 1, 1, 1, 0, streams[i], kargs, 0);
+      if (r != CUDA_SUCCESS)
+	{
+	  fprintf (stderr, "cuLaunchKernel failed: %d\n", r);
+	  abort ();
+	}
+
+    }
+
+  if (acc_async_test_all () != 0)
+    {
+      fprintf (stderr, "asynchronous operation not running\n");
+      abort ();
+    }
+
+  sleep ((int) (dtime / 1000.0f) + 1);
+
+  if (acc_async_test_all () != 1)
+    {
+      fprintf (stderr, "asynchronous operation not running\n");
+      abort ();
+    }
+
+  acc_unmap_data (a);
+
+  free (a);
+  acc_free (d_a);
+
+  acc_shutdown (acc_device_nvidia);
+
+  exit (0);
+}
+
+/* { dg-output "" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-74.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-74.c
new file mode 100644
index 00000000000..0726ee420c3
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-74.c
@@ -0,0 +1,139 @@
+/* { dg-do run { target openacc_nvidia_accel_selected } } */
+/* { dg-additional-options "-lcuda" } */
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <openacc.h>
+#include <cuda.h>
+#include "timer.h"
+
+int
+main (int argc, char **argv)
+{
+  CUdevice dev;
+  CUfunction delay;
+  CUmodule module;
+  CUresult r;
+  CUstream stream;
+  unsigned long *a, *d_a, dticks;
+  int nbytes;
+  float atime, dtime;
+  void *kargs[2];
+  int clkrate;
+  int devnum, nprocs;
+
+  acc_init (acc_device_nvidia);
+
+  devnum = acc_get_device_num (acc_device_nvidia);
+
+  r = cuDeviceGet (&dev, devnum);
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuDeviceGet failed: %d\n", r);
+      abort ();
+    }
+
+  r =
+    cuDeviceGetAttribute (&nprocs, CU_DEVICE_ATTRIBUTE_MULTIPROCESSOR_COUNT,
+			  dev);
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuDeviceGetAttribute failed: %d\n", r);
+      abort ();
+    }
+
+  r = cuDeviceGetAttribute (&clkrate, CU_DEVICE_ATTRIBUTE_CLOCK_RATE, dev);
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuDeviceGetAttribute failed: %d\n", r);
+      abort ();
+    }
+
+  r = cuModuleLoad (&module, "subr.ptx");
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuModuleLoad failed: %d\n", r);
+      abort ();
+    }
+
+  r = cuModuleGetFunction (&delay, module, "delay");
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuModuleGetFunction failed: %d\n", r);
+      abort ();
+    }
+
+  nbytes = nprocs * sizeof (unsigned long);
+
+  dtime = 200.0;
+
+  dticks = (unsigned long) (dtime * clkrate);
+
+  a = (unsigned long *) malloc (nbytes);
+  d_a = (unsigned long *) acc_malloc (nbytes);
+
+  acc_map_data (a, d_a, nbytes);
+
+  kargs[0] = (void *) &d_a;
+  kargs[1] = (void *) &dticks;
+
+  stream = (CUstream) acc_get_cuda_stream (0);
+  if (stream != NULL)
+    abort ();
+
+  r = cuStreamCreate (&stream, CU_STREAM_DEFAULT);
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuStreamCreate failed: %d\n", r);
+      abort ();
+    }
+
+  if (!acc_set_cuda_stream (0, stream))
+    abort ();
+
+  init_timers (1);
+
+  start_timer (0);
+
+  r = cuLaunchKernel (delay, 1, 1, 1, 1, 1, 1, 0, stream, kargs, 0);
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuLaunchKernel failed: %d\n", r);
+      abort ();
+    }
+
+  acc_wait (0);
+
+  atime = stop_timer (0);
+
+  if (atime < dtime)
+    {
+      fprintf (stderr, "actual time < delay time\n");
+      abort ();
+    }
+
+  start_timer (0);
+
+  acc_wait (0);
+
+  atime = stop_timer (0);
+
+  if (0.010 < atime)
+    {
+      fprintf (stderr, "actual time too long\n");
+      abort ();
+    }
+
+  acc_unmap_data (a);
+
+  fini_timers ();
+
+  free (a);
+  acc_free (d_a);
+
+  acc_shutdown (acc_device_nvidia);
+
+  exit (0);
+}
+
+/* { dg-output "" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-75.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-75.c
new file mode 100644
index 00000000000..19422118748
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-75.c
@@ -0,0 +1,141 @@
+/* { dg-do run { target openacc_nvidia_accel_selected } } */
+/* { dg-additional-options "-lcuda" } */
+
+#include <stdio.h>
+#include <unistd.h>
+#include <stdlib.h>
+#include <openacc.h>
+#include <cuda.h>
+#include "timer.h"
+
+int
+main (int argc, char **argv)
+{
+  CUdevice dev;
+  CUfunction delay;
+  CUmodule module;
+  CUresult r;
+  int N;
+  int i;
+  CUstream stream;
+  unsigned long *a, *d_a, dticks;
+  int nbytes;
+  float atime, dtime, hitime, lotime;
+  void *kargs[2];
+  int clkrate;
+  int devnum, nprocs;
+
+  acc_init (acc_device_nvidia);
+
+  devnum = acc_get_device_num (acc_device_nvidia);
+
+  r = cuDeviceGet (&dev, devnum);
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuDeviceGet failed: %d\n", r);
+      abort ();
+    }
+
+  r =
+    cuDeviceGetAttribute (&nprocs, CU_DEVICE_ATTRIBUTE_MULTIPROCESSOR_COUNT,
+			  dev);
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuDeviceGetAttribute failed: %d\n", r);
+      abort ();
+    }
+
+  r = cuDeviceGetAttribute (&clkrate, CU_DEVICE_ATTRIBUTE_CLOCK_RATE, dev);
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuDeviceGetAttribute failed: %d\n", r);
+      abort ();
+    }
+
+  r = cuModuleLoad (&module, "subr.ptx");
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuModuleLoad failed: %d\n", r);
+      abort ();
+    }
+
+  r = cuModuleGetFunction (&delay, module, "delay");
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuModuleGetFunction failed: %d\n", r);
+      abort ();
+    }
+
+  nbytes = nprocs * sizeof (unsigned long);
+
+  dtime = 200.0;
+
+  dticks = (unsigned long) (dtime * clkrate);
+
+  N = nprocs;
+
+  a = (unsigned long *) malloc (nbytes);
+  d_a = (unsigned long *) acc_malloc (nbytes);
+
+  acc_map_data (a, d_a, nbytes);
+
+  stream = (CUstream) acc_get_cuda_stream (0);
+  if (stream != NULL)
+    abort ();
+
+  r = cuStreamCreate (&stream, CU_STREAM_DEFAULT);
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuStreamCreate failed: %d\n", r);
+      abort ();
+    }
+
+  if (!acc_set_cuda_stream (0, stream))
+    abort ();
+
+  init_timers (1);
+
+  kargs[0] = (void *) &d_a;
+  kargs[1] = (void *) &dticks;
+
+  start_timer (0);
+
+  for (i = 0; i < N; i++)
+    {
+      r = cuLaunchKernel (delay, 1, 1, 1, 1, 1, 1, 0, stream, kargs, 0);
+      if (r != CUDA_SUCCESS)
+	{
+	  fprintf (stderr, "cuLaunchKernel failed: %d\n", r);
+	  abort ();
+	}
+
+      acc_wait (0);
+    }
+
+  atime = stop_timer (0);
+
+  hitime = dtime * N;
+  hitime += hitime * 0.02;
+
+  lotime = dtime * N;
+  lotime -= lotime * 0.02;
+
+  if (atime > hitime || atime < lotime)
+    {
+      fprintf (stderr, "actual time < delay time\n");
+      abort ();
+    }
+
+  acc_unmap_data (a);
+
+  fini_timers ();
+
+  free (a);
+  acc_free (d_a);
+
+  acc_shutdown (acc_device_nvidia);
+
+  exit (0);
+}
+
+/* { dg-output "" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-76.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-76.c
new file mode 100644
index 00000000000..11d9d621f87
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-76.c
@@ -0,0 +1,147 @@
+/* { dg-do run { target openacc_nvidia_accel_selected } } */
+/* { dg-additional-options "-lcuda" } */
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <unistd.h>
+#include <openacc.h>
+#include <cuda.h>
+#include "timer.h"
+
+int
+main (int argc, char **argv)
+{
+  CUdevice dev;
+  CUfunction delay;
+  CUmodule module;
+  CUresult r;
+  int N;
+  int i;
+  CUstream *streams;
+  unsigned long *a, *d_a, dticks;
+  int nbytes;
+  float atime, dtime, hitime, lotime;
+  void *kargs[2];
+  int clkrate;
+  int devnum, nprocs;
+
+  acc_init (acc_device_nvidia);
+
+  devnum = acc_get_device_num (acc_device_nvidia);
+
+  r = cuDeviceGet (&dev, devnum);
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuDeviceGet failed: %d\n", r);
+      abort ();
+    }
+
+  r =
+    cuDeviceGetAttribute (&nprocs, CU_DEVICE_ATTRIBUTE_MULTIPROCESSOR_COUNT,
+			  dev);
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuDeviceGetAttribute failed: %d\n", r);
+      abort ();
+    }
+
+  r = cuDeviceGetAttribute (&clkrate, CU_DEVICE_ATTRIBUTE_CLOCK_RATE, dev);
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuDeviceGetAttribute failed: %d\n", r);
+      abort ();
+    }
+
+  r = cuModuleLoad (&module, "subr.ptx");
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuModuleLoad failed: %d\n", r);
+      abort ();
+    }
+
+  r = cuModuleGetFunction (&delay, module, "delay");
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuModuleGetFunction failed: %d\n", r);
+      abort ();
+    }
+
+  nbytes = nprocs * sizeof (unsigned long);
+
+  dtime = 200.0;
+
+  dticks = (unsigned long) (dtime * clkrate);
+
+  N = nprocs;
+
+  a = (unsigned long *) malloc (nbytes);
+  d_a = (unsigned long *) acc_malloc (nbytes);
+
+  acc_map_data (a, d_a, nbytes);
+
+  streams = (CUstream *) malloc (N * sizeof (void *));
+
+  for (i = 0; i < N; i++)
+    {
+      streams[i] = (CUstream) acc_get_cuda_stream (i);
+      if (streams[i] != NULL)
+	abort ();
+
+      r = cuStreamCreate (&streams[i], CU_STREAM_DEFAULT);
+      if (r != CUDA_SUCCESS)
+	{
+	  fprintf (stderr, "cuStreamCreate failed: %d\n", r);
+	  abort ();
+	}
+
+        if (!acc_set_cuda_stream (i, streams[i]))
+	  abort ();
+    }
+
+  init_timers (1);
+
+  kargs[0] = (void *) &d_a;
+  kargs[1] = (void *) &dticks;
+
+  start_timer (0);
+
+  for (i = 0; i < N; i++)
+    {
+      r = cuLaunchKernel (delay, 1, 1, 1, 1, 1, 1, 0, streams[i], kargs, 0);
+      if (r != CUDA_SUCCESS)
+	{
+	  fprintf (stderr, "cuLaunchKernel failed: %d\n", r);
+	  abort ();
+	}
+
+      acc_wait (i);
+    }
+
+  atime = stop_timer (0);
+
+  hitime = dtime * N;
+  hitime += hitime * 0.02;
+
+  lotime = dtime * N;
+  lotime -= lotime * 0.02;
+
+  if (atime > hitime || atime < lotime)
+    {
+      fprintf (stderr, "actual time < delay time\n");
+      abort ();
+    }
+
+  acc_unmap_data (a);
+
+  fini_timers ();
+
+  free (streams);
+  free (a);
+  acc_free (d_a);
+
+  acc_shutdown (acc_device_nvidia);
+
+  exit (0);
+}
+
+/* { dg-output "" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-77.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-77.c
new file mode 100644
index 00000000000..e47212b2139
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-77.c
@@ -0,0 +1,135 @@
+/* { dg-do run { target openacc_nvidia_accel_selected } } */
+/* { dg-additional-options "-lcuda" } */
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <unistd.h>
+#include <openacc.h>
+#include <cuda.h>
+#include "timer.h"
+
+int
+main (int argc, char **argv)
+{
+  CUdevice dev;
+  CUfunction delay;
+  CUmodule module;
+  CUresult r;
+  CUstream stream;
+  unsigned long *a, *d_a, dticks;
+  int nbytes;
+  float atime, dtime;
+  void *kargs[2];
+  int clkrate;
+  int devnum, nprocs;
+
+  acc_init (acc_device_nvidia);
+
+  devnum = acc_get_device_num (acc_device_nvidia);
+
+  r = cuDeviceGet (&dev, devnum);
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuDeviceGet failed: %d\n", r);
+      abort ();
+    }
+
+  r =
+    cuDeviceGetAttribute (&nprocs, CU_DEVICE_ATTRIBUTE_MULTIPROCESSOR_COUNT,
+			  dev);
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuDeviceGetAttribute failed: %d\n", r);
+      abort ();
+    }
+
+  r = cuDeviceGetAttribute (&clkrate, CU_DEVICE_ATTRIBUTE_CLOCK_RATE, dev);
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuDeviceGetAttribute failed: %d\n", r);
+      abort ();
+    }
+
+  r = cuModuleLoad (&module, "subr.ptx");
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuModuleLoad failed: %d\n", r);
+      abort ();
+    }
+
+  r = cuModuleGetFunction (&delay, module, "delay");
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuModuleGetFunction failed: %d\n", r);
+      abort ();
+    }
+
+  nbytes = nprocs * sizeof (unsigned long);
+
+  dtime = 200.0;
+
+  dticks = (unsigned long) (dtime * clkrate);
+
+  a = (unsigned long *) malloc (nbytes);
+  d_a = (unsigned long *) acc_malloc (nbytes);
+
+  acc_map_data (a, d_a, nbytes);
+
+  kargs[0] = (void *) &d_a;
+  kargs[1] = (void *) &dticks;
+
+  r = cuStreamCreate (&stream, CU_STREAM_DEFAULT);
+  if (r != CUDA_SUCCESS)
+	{
+	  fprintf (stderr, "cuStreamCreate failed: %d\n", r);
+	  abort ();
+	}
+
+  acc_set_cuda_stream (0, stream);
+
+  init_timers (1);
+
+  start_timer (0);
+
+  r = cuLaunchKernel (delay, 1, 1, 1, 1, 1, 1, 0, stream, kargs, 0);
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuLaunchKernel failed: %d\n", r);
+      abort ();
+    }
+
+  acc_wait (1);
+
+  atime = stop_timer (0);
+
+  if (atime < dtime)
+    {
+      fprintf (stderr, "actual time < delay time\n");
+      abort ();
+    }
+
+  start_timer (0);
+
+  acc_wait (1);
+
+  atime = stop_timer (0);
+
+  if (0.010 < atime)
+    {
+      fprintf (stderr, "actual time < delay time\n");
+      abort ();
+    }
+
+  acc_unmap_data (a);
+
+  fini_timers ();
+
+  free (a);
+  acc_free (d_a);
+
+  acc_shutdown (acc_device_nvidia);
+
+  return 0;
+}
+
+/* { dg-shouldfail "libgomp: unknown async \d" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-78.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-78.c
new file mode 100644
index 00000000000..4f58fb23cfe
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-78.c
@@ -0,0 +1,140 @@
+/* { dg-do run { target openacc_nvidia_accel_selected } } */
+/* { dg-additional-options "-lcuda" } */
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <unistd.h>
+#include <openacc.h>
+#include <cuda.h>
+#include "timer.h"
+
+int
+main (int argc, char **argv)
+{
+  CUdevice dev;
+  CUfunction delay;
+  CUmodule module;
+  CUresult r;
+  CUstream stream;
+  unsigned long *a, *d_a, dticks;
+  int nbytes;
+  float atime, dtime;
+  void *kargs[2];
+  int clkrate;
+  int devnum, nprocs;
+
+  acc_init (acc_device_nvidia);
+
+  devnum = acc_get_device_num (acc_device_nvidia);
+
+  r = cuDeviceGet (&dev, devnum);
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuDeviceGet failed: %d\n", r);
+      abort ();
+    }
+
+  r =
+    cuDeviceGetAttribute (&nprocs, CU_DEVICE_ATTRIBUTE_MULTIPROCESSOR_COUNT,
+			  dev);
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuDeviceGetAttribute failed: %d\n", r);
+      abort ();
+    }
+
+  r = cuDeviceGetAttribute (&clkrate, CU_DEVICE_ATTRIBUTE_CLOCK_RATE, dev);
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuDeviceGetAttribute failed: %d\n", r);
+      abort ();
+    }
+
+  r = cuModuleLoad (&module, "subr.ptx");
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuModuleLoad failed: %d\n", r);
+      abort ();
+    }
+
+  r = cuModuleGetFunction (&delay, module, "delay");
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuModuleGetFunction failed: %d\n", r);
+      abort ();
+    }
+
+  nbytes = nprocs * sizeof (unsigned long);
+
+  dtime = 200.0;
+
+  dticks = (unsigned long) (dtime * clkrate);
+
+  a = (unsigned long *) malloc (nbytes);
+  d_a = (unsigned long *) acc_malloc (nbytes);
+
+  acc_map_data (a, d_a, nbytes);
+
+  kargs[0] = (void *) &d_a;
+  kargs[1] = (void *) &dticks;
+
+  stream = (CUstream) acc_get_cuda_stream (0);
+  if (stream != NULL)
+    abort ();
+
+  r = cuStreamCreate (&stream, CU_STREAM_DEFAULT);
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuStreamCreate failed: %d\n", r);
+      abort ();
+    }
+
+  if (!acc_set_cuda_stream (0, stream))
+    abort ();
+
+  init_timers (1);
+
+  start_timer (0);
+
+  r = cuLaunchKernel (delay, 1, 1, 1, 1, 1, 1, 0, stream, kargs, 0);
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuLaunchKernel failed: %d\n", r);
+      abort ();
+    }
+
+  acc_wait_all ();
+
+  atime = stop_timer (0);
+
+  if (atime < dtime)
+    {
+      fprintf (stderr, "actual time < delay time\n");
+      abort ();
+    }
+
+  start_timer (0);
+
+  acc_wait_all ();
+
+  atime = stop_timer (0);
+
+  if (0.010 < atime)
+    {
+      fprintf (stderr, "actual time too long\n");
+      abort ();
+    }
+
+  acc_unmap_data (a);
+
+  fini_timers ();
+
+  free (a);
+  acc_free (d_a);
+
+  acc_shutdown (acc_device_nvidia);
+
+  exit (0);
+}
+
+/* { dg-output "" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-79.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-79.c
new file mode 100644
index 00000000000..ef3df13ebc9
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-79.c
@@ -0,0 +1,167 @@
+/* { dg-do run { target openacc_nvidia_accel_selected } } */
+/* { dg-additional-options "-lcuda" } */
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <unistd.h>
+#include <openacc.h>
+#include <cuda.h>
+#include "timer.h"
+
+int
+main (int argc, char **argv)
+{
+  CUdevice dev;
+  CUfunction delay;
+  CUmodule module;
+  CUresult r;
+  int N;
+  int i;
+  CUstream stream;
+  unsigned long *a, *d_a, dticks;
+  int nbytes;
+  float atime, dtime, hitime, lotime;
+  void *kargs[2];
+  int clkrate;
+  int devnum, nprocs;
+
+  devnum = 2;
+
+  acc_init (acc_device_nvidia);
+
+  devnum = acc_get_device_num (acc_device_nvidia);
+
+  r = cuDeviceGet (&dev, devnum);
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuDeviceGet failed: %d\n", r);
+      abort ();
+    }
+
+  r =
+    cuDeviceGetAttribute (&nprocs, CU_DEVICE_ATTRIBUTE_MULTIPROCESSOR_COUNT,
+			  dev);
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuDeviceGetAttribute failed: %d\n", r);
+      abort ();
+    }
+
+  r = cuDeviceGetAttribute (&clkrate, CU_DEVICE_ATTRIBUTE_CLOCK_RATE, dev);
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuDeviceGetAttribute failed: %d\n", r);
+      abort ();
+    }
+
+  r = cuModuleLoad (&module, "subr.ptx");
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuModuleLoad failed: %d\n", r);
+      abort ();
+    }
+
+  r = cuModuleGetFunction (&delay, module, "delay");
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuModuleGetFunction failed: %d\n", r);
+      abort ();
+    }
+
+  nbytes = nprocs * sizeof (unsigned long);
+
+  dtime = 200.0;
+
+  dticks = (unsigned long) (dtime * clkrate);
+
+  N = nprocs;
+
+  a = (unsigned long *) malloc (nbytes);
+  d_a = (unsigned long *) acc_malloc (nbytes);
+
+  acc_map_data (a, d_a, nbytes);
+
+  r = cuStreamCreate (&stream, CU_STREAM_DEFAULT);
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuStreamCreate failed: %d\n", r);
+      abort ();
+    }
+
+  if (!acc_set_cuda_stream (1, stream))
+    abort ();
+
+  stream = (CUstream) acc_get_cuda_stream (0);
+  if (stream != NULL)
+    abort ();
+
+  r = cuStreamCreate (&stream, CU_STREAM_DEFAULT);
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuStreamCreate failed: %d\n", r);
+      abort ();
+    }
+
+  if (!acc_set_cuda_stream (0, stream))
+    abort ();
+
+  init_timers (1);
+
+  kargs[0] = (void *) &d_a;
+  kargs[1] = (void *) &dticks;
+
+  start_timer (0);
+
+  for (i = 0; i < N; i++)
+    {
+      r = cuLaunchKernel (delay, 1, 1, 1, 1, 1, 1, 0, stream, kargs, 0);
+      if (r != CUDA_SUCCESS)
+	{
+	  fprintf (stderr, "cuLaunchKernel failed: %d\n", r);
+	  abort ();
+	}
+    }
+
+  acc_wait_async (0, 1);
+
+  if (acc_async_test (0) != 0)
+    abort ();
+
+  if (acc_async_test (1) != 0)
+    abort ();
+
+  acc_wait (1);
+
+  atime = stop_timer (0);
+
+  if (acc_async_test (0) != 1)
+    abort ();
+
+  if (acc_async_test (1) != 1)
+    abort ();
+
+  hitime = dtime * N;
+  hitime += hitime * 0.02;
+
+  lotime = dtime * N;
+  lotime -= lotime * 0.02;
+
+  if (atime > hitime || atime < lotime)
+    {
+      fprintf (stderr, "actual time < delay time\n");
+      abort ();
+    }
+
+  acc_unmap_data (a);
+
+  fini_timers ();
+
+  free (a);
+  acc_free (d_a);
+
+  acc_shutdown (acc_device_nvidia);
+
+  exit (0);
+}
+
+/* { dg-output "" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-80.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-80.c
new file mode 100644
index 00000000000..0b5ec24b28e
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-80.c
@@ -0,0 +1,132 @@
+/* { dg-do run { target openacc_nvidia_accel_selected } } */
+/* { dg-additional-options "-lcuda" } */
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <unistd.h>
+#include <openacc.h>
+#include <cuda.h>
+#include "timer.h"
+
+int
+main (int argc, char **argv)
+{
+  CUdevice dev;
+  CUfunction delay;
+  CUmodule module;
+  CUresult r;
+  CUstream stream;
+  int N;
+  int i;
+  unsigned long *a, *d_a, dticks;
+  int nbytes;
+  float atime, dtime;
+  void *kargs[2];
+  int clkrate;
+  int devnum, nprocs;
+
+  acc_init (acc_device_nvidia);
+
+  devnum = acc_get_device_num (acc_device_nvidia);
+
+  r = cuDeviceGet (&dev, devnum);
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuDeviceGet failed: %d\n", r);
+      abort ();
+    }
+
+  r =
+    cuDeviceGetAttribute (&nprocs, CU_DEVICE_ATTRIBUTE_MULTIPROCESSOR_COUNT,
+			  dev);
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuDeviceGetAttribute failed: %d\n", r);
+      abort ();
+    }
+
+  r = cuDeviceGetAttribute (&clkrate, CU_DEVICE_ATTRIBUTE_CLOCK_RATE, dev);
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuDeviceGetAttribute failed: %d\n", r);
+      abort ();
+    }
+
+  r = cuModuleLoad (&module, "subr.ptx");
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuModuleLoad failed: %d\n", r);
+      abort ();
+    }
+
+  r = cuModuleGetFunction (&delay, module, "delay");
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuModuleGetFunction failed: %d\n", r);
+      abort ();
+    }
+
+  nbytes = nprocs * sizeof (unsigned long);
+
+  dtime = 200.0;
+
+  dticks = (unsigned long) (dtime * clkrate);
+
+  N = nprocs;
+
+  a = (unsigned long *) malloc (nbytes);
+  d_a = (unsigned long *) acc_malloc (nbytes);
+
+  acc_map_data (a, d_a, nbytes);
+
+  r = cuStreamCreate (&stream, CU_STREAM_DEFAULT);
+  if (r != CUDA_SUCCESS)
+	{
+	  fprintf (stderr, "cuStreamCreate failed: %d\n", r);
+	  abort ();
+	}
+
+  acc_set_cuda_stream (1, stream);
+
+  init_timers (1);
+
+  kargs[0] = (void *) &d_a;
+  kargs[1] = (void *) &dticks;
+
+  start_timer (0);
+
+  for (i = 0; i < N; i++)
+    {
+      r = cuLaunchKernel (delay, 1, 1, 1, 1, 1, 1, 0, stream, kargs, 0);
+      if (r != CUDA_SUCCESS)
+	{
+	  fprintf (stderr, "cuLaunchKernel failed: %d\n", r);
+	  abort ();
+	}
+    }
+
+  acc_wait_async (1, 1);
+
+  acc_wait (1);
+
+  atime = stop_timer (0);
+
+  if (atime < dtime)
+    {
+      fprintf (stderr, "actual time < delay time\n");
+      abort ();
+    }
+
+  acc_unmap_data (a);
+
+  fini_timers ();
+
+  free (a);
+  acc_free (d_a);
+
+  acc_shutdown (acc_device_nvidia);
+
+  return 0;
+}
+
+/* { dg-shouldfail "libgomp: identical parameters" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-81.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-81.c
new file mode 100644
index 00000000000..d5f18f00319
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-81.c
@@ -0,0 +1,211 @@
+/* { dg-do run { target openacc_nvidia_accel_selected } } */
+/* { dg-additional-options "-lcuda" } */
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <unistd.h>
+#include <openacc.h>
+#include <cuda.h>
+#include "timer.h"
+
+int
+main (int argc, char **argv)
+{
+  CUdevice dev;
+  CUfunction delay;
+  CUmodule module;
+  CUresult r;
+  int N;
+  int i;
+  CUstream *streams, stream;
+  unsigned long *a, *d_a, dticks;
+  int nbytes;
+  float atime, dtime;
+  void *kargs[2];
+  int clkrate;
+  int devnum, nprocs;
+
+  acc_init (acc_device_nvidia);
+
+  devnum = acc_get_device_num (acc_device_nvidia);
+
+  r = cuDeviceGet (&dev, devnum);
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuDeviceGet failed: %d\n", r);
+      abort ();
+    }
+
+  r =
+    cuDeviceGetAttribute (&nprocs, CU_DEVICE_ATTRIBUTE_MULTIPROCESSOR_COUNT,
+			  dev);
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuDeviceGetAttribute failed: %d\n", r);
+      abort ();
+    }
+
+  r = cuDeviceGetAttribute (&clkrate, CU_DEVICE_ATTRIBUTE_CLOCK_RATE, dev);
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuDeviceGetAttribute failed: %d\n", r);
+      abort ();
+    }
+
+  r = cuModuleLoad (&module, "subr.ptx");
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuModuleLoad failed: %d\n", r);
+      abort ();
+    }
+
+  r = cuModuleGetFunction (&delay, module, "delay");
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuModuleGetFunction failed: %d\n", r);
+      abort ();
+    }
+
+  nbytes = nprocs * sizeof (unsigned long);
+
+  dtime = 500.0;
+
+  dticks = (unsigned long) (dtime * clkrate);
+
+  N = nprocs;
+
+  a = (unsigned long *) malloc (nbytes);
+  d_a = (unsigned long *) acc_malloc (nbytes);
+
+  acc_map_data (a, d_a, nbytes);
+
+  streams = (CUstream *) malloc (N * sizeof (void *));
+
+  for (i = 0; i < N; i++)
+    {
+      streams[i] = (CUstream) acc_get_cuda_stream (i);
+      if (streams[i] != NULL)
+	abort ();
+
+      r = cuStreamCreate (&streams[i], CU_STREAM_DEFAULT);
+      if (r != CUDA_SUCCESS)
+	{
+	  fprintf (stderr, "cuStreamCreate failed: %d\n", r);
+	  abort ();
+	}
+
+        if (!acc_set_cuda_stream (i, streams[i]))
+	  abort ();
+    }
+
+  init_timers (1);
+
+  kargs[0] = (void *) &d_a;
+  kargs[1] = (void *) &dticks;
+
+  stream = (CUstream) acc_get_cuda_stream (N);
+  if (stream != NULL)
+    abort ();
+
+  r = cuStreamCreate (&stream, CU_STREAM_DEFAULT);
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuStreamCreate failed: %d\n", r);
+      abort ();
+    }
+
+  if (!acc_set_cuda_stream (N, stream))
+    abort ();
+
+  start_timer (0);
+
+  for (i = 0; i < N; i++)
+    {
+      r = cuLaunchKernel (delay, 1, 1, 1, 1, 1, 1, 0, streams[i], kargs, 0);
+      if (r != CUDA_SUCCESS)
+	{
+	  fprintf (stderr, "cuLaunchKernel failed: %d\n", r);
+	  abort ();
+	}
+    }
+
+  acc_wait_all_async (N);
+
+  for (i = 0; i <= N; i++)
+    {
+      if (acc_async_test (i) != 0)
+	abort ();
+    }
+
+  acc_wait (N);
+
+  for (i = 0; i <= N; i++)
+    {
+      if (acc_async_test (i) != 1)
+	abort ();
+    }
+
+  atime = stop_timer (0);
+
+  if (atime < dtime)
+    {
+      fprintf (stderr, "actual time < delay time\n");
+      abort ();
+    }
+
+  start_timer (0);
+
+  stream = (CUstream) acc_get_cuda_stream (N + 1);
+  if (stream != NULL)
+    abort ();
+
+  r = cuStreamCreate (&stream, CU_STREAM_DEFAULT);
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuStreamCreate failed: %d\n", r);
+      abort ();
+    }
+
+  if (!acc_set_cuda_stream (N + 1, stream))
+    abort ();
+
+  acc_wait_all_async (N + 1);
+
+  acc_wait (N + 1);
+
+  atime = stop_timer (0);
+
+  if (0.10 < atime)
+    {
+      fprintf (stderr, "actual time too long\n");
+      abort ();
+    }
+
+  start_timer (0);
+
+  acc_wait_all_async (N);
+
+  acc_wait (N);
+
+  atime = stop_timer (0);
+
+  if (0.10 < atime)
+    {
+      fprintf (stderr, "actual time too long\n");
+      abort ();
+    }
+
+  acc_unmap_data (a);
+
+  fini_timers ();
+
+  free (streams);
+  free (a);
+  acc_free (d_a);
+
+  acc_shutdown (acc_device_nvidia);
+
+  exit (0);
+}
+
+/* { dg-output "" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-82.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-82.c
new file mode 100644
index 00000000000..be30a7f28ac
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-82.c
@@ -0,0 +1,144 @@
+/* { dg-do run { target openacc_nvidia_accel_selected } } */
+/* { dg-additional-options "-lcuda" } */
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <unistd.h>
+#include <openacc.h>
+#include <cuda.h>
+
+int
+main (int argc, char **argv)
+{
+  CUdevice dev;
+  CUfunction delay2;
+  CUmodule module;
+  CUresult r;
+  int N;
+  int i;
+  CUstream *streams;
+  unsigned long **a, **d_a, *tid, ticks;
+  int nbytes;
+  void *kargs[3];
+  int clkrate;
+  int devnum, nprocs;
+
+  acc_init (acc_device_nvidia);
+
+  devnum = acc_get_device_num (acc_device_nvidia);
+
+  r = cuDeviceGet (&dev, devnum);
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuDeviceGet failed: %d\n", r);
+      abort ();
+    }
+
+  r =
+    cuDeviceGetAttribute (&nprocs, CU_DEVICE_ATTRIBUTE_MULTIPROCESSOR_COUNT,
+			  dev);
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuDeviceGetAttribute failed: %d\n", r);
+      abort ();
+    }
+
+  r = cuDeviceGetAttribute (&clkrate, CU_DEVICE_ATTRIBUTE_CLOCK_RATE, dev);
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuDeviceGetAttribute failed: %d\n", r);
+      abort ();
+    }
+
+  r = cuModuleLoad (&module, "subr.ptx");
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuModuleLoad failed: %d\n", r);
+      abort ();
+    }
+
+  r = cuModuleGetFunction (&delay2, module, "delay2");
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuModuleGetFunction failed: %d\n", r);
+      abort ();
+    }
+
+  nbytes = sizeof (int);
+
+  ticks = (unsigned long) (200.0 * clkrate);
+
+  N = nprocs;
+
+  streams = (CUstream *) malloc (N * sizeof (void *));
+
+  a = (unsigned long **) malloc (N * sizeof (unsigned long *));
+  d_a = (unsigned long **) malloc (N * sizeof (unsigned long *));
+  tid = (unsigned long *) malloc (N * sizeof (unsigned long));
+
+  for (i = 0; i < N; i++)
+    {
+      a[i] = (unsigned long *) malloc (sizeof (unsigned long));
+      *a[i] = N;
+      d_a[i] = (unsigned long *) acc_malloc (nbytes);
+      tid[i] = i;
+
+      acc_map_data (a[i], d_a[i], nbytes);
+
+      streams[i] = (CUstream) acc_get_cuda_stream (i);
+      if (streams[i] != NULL)
+        abort ();
+
+      r = cuStreamCreate (&streams[i], CU_STREAM_DEFAULT);
+      if (r != CUDA_SUCCESS)
+        {
+          fprintf (stderr, "cuStreamCreate failed: %d\n", r);
+          abort ();
+        }
+
+       if (!acc_set_cuda_stream (i, streams[i]))
+        abort ();
+    }
+
+  for (i = 0; i < N; i++)
+    {
+      kargs[0] = (void *) &d_a[i];
+      kargs[1] = (void *) &ticks;
+      kargs[2] = (void *) &tid[i];
+
+      r = cuLaunchKernel (delay2, 1, 1, 1, 1, 1, 1, 0, streams[i], kargs, 0);
+      if (r != CUDA_SUCCESS)
+	{
+	  fprintf (stderr, "cuLaunchKernel failed: %d\n", r);
+	  abort ();
+	}
+
+      ticks = (unsigned long) (50.0 * clkrate);
+    }
+
+  acc_wait_all_async (0);
+
+  for (i = 0; i < N; i++)
+    {
+      acc_copyout (a[i], nbytes);
+      if (*a[i] != i)
+	abort ();
+    }
+
+  free (streams);
+
+  for (i = 0; i < N; i++)
+    {
+      free (a[i]);
+    }
+
+  free (a);
+  free (d_a);
+  free (tid);
+
+  acc_shutdown (acc_device_nvidia);
+
+  exit (0);
+}
+
+/* { dg-output "" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-83.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-83.c
new file mode 100644
index 00000000000..1c2e52b4c5f
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-83.c
@@ -0,0 +1,58 @@
+/* { dg-do run { target openacc_nvidia_accel_selected } } */
+/* { dg-additional-options "-lcuda" } */
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <unistd.h>
+#include <openacc.h>
+#include "timer.h"
+
+int
+main (int argc, char **argv)
+{
+  float atime;
+  CUstream stream;
+  CUresult r;
+
+  acc_init (acc_device_nvidia);
+
+  (void) acc_get_device_num (acc_device_nvidia);
+
+  init_timers (1);
+
+  stream = (CUstream) acc_get_cuda_stream (0);
+  if (stream != NULL)
+    abort ();
+
+  r = cuStreamCreate (&stream, CU_STREAM_DEFAULT);
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuStreamCreate failed: %d\n", r);
+      abort ();
+    }
+
+  if (!acc_set_cuda_stream (0, stream))
+    abort ();
+
+  start_timer (0);
+
+  acc_wait_all_async (0);
+
+  acc_wait (0);
+
+  atime = stop_timer (0);
+
+  if (0.010 < atime)
+    {
+      fprintf (stderr, "actual time too long\n");
+      abort ();
+    }
+
+  fini_timers ();
+
+  acc_shutdown (acc_device_nvidia);
+
+  exit (0);
+}
+
+/* { dg-output "" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-84.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-84.c
new file mode 100644
index 00000000000..786b908f755
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-84.c
@@ -0,0 +1,66 @@
+/* { dg-do run { target openacc_nvidia_accel_selected } } */
+/* { dg-additional-options "-lcuda" } */
+
+#include <stdlib.h>
+#include <unistd.h>
+#include <stdio.h>
+#include <openacc.h>
+#include <cuda.h>
+
+int
+main (int argc, char **argv)
+{
+  const int N = 100;
+  int i;
+  CUstream *streams;
+  CUstream s;
+  CUresult r;
+
+  acc_init (acc_device_nvidia);
+
+  (void) acc_get_device_num (acc_device_nvidia);
+
+  streams = (CUstream *) malloc (N * sizeof (void *));
+
+  for (i = 0; i < N; i++)
+    {
+      streams[i] = (CUstream) acc_get_cuda_stream (i);
+      if (streams[i] != NULL)
+	abort ();
+
+      r = cuStreamCreate (&streams[i], CU_STREAM_DEFAULT);
+      if (r != CUDA_SUCCESS)
+	{
+	  fprintf (stderr, "cuStreamCreate failed: %d\n", r);
+	  abort ();
+	}
+
+        if (!acc_set_cuda_stream (i, streams[i]))
+	  abort ();
+    }
+
+  for (i = 0; i < N; i++)
+    {
+      int j;
+      int cnt;
+
+      cnt = 0;
+
+      s = streams[i];
+
+      for (j = 0; j < N; j++)
+	{
+	  if (s == streams[j])
+	    cnt++;
+	}
+
+      if (cnt != 1)
+	abort ();
+    }
+
+  acc_shutdown (acc_device_nvidia);
+
+  exit (0);
+}
+
+/* { dg-output "" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-85.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-85.c
new file mode 100644
index 00000000000..cf925a7b002
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-85.c
@@ -0,0 +1,52 @@
+/* { dg-do run { target openacc_nvidia_accel_selected } } */
+/* { dg-additional-options "-lcuda" } */
+
+#include <stdlib.h>
+#include <unistd.h>
+#include <openacc.h>
+#include <stdio.h>
+#include <cuda.h>
+
+int
+main (int argc, char **argv)
+{
+  const int N = 100;
+  int i;
+  CUstream *streams;
+  CUstream s;
+  CUresult r;
+
+  acc_init (acc_device_nvidia);
+
+  (void) acc_get_device_num (acc_device_nvidia);
+
+  streams = (CUstream *) malloc (N * sizeof (void *));
+
+  for (i = 0; i < N; i++)
+    {
+      streams[i] = (CUstream) acc_get_cuda_stream (i);
+      if (streams[i] != NULL)
+	abort ();
+
+      r = cuStreamCreate (&streams[i], CU_STREAM_DEFAULT);
+      if (r != CUDA_SUCCESS)
+	{
+	  fprintf (stderr, "cuStreamCreate failed: %d\n", r);
+	  abort ();
+	}
+
+        if (!acc_set_cuda_stream (i, streams[i]))
+	  abort ();
+    }
+
+  s = NULL;
+
+  if (acc_set_cuda_stream (N + 1, s) != 0)
+    abort ();
+
+  acc_shutdown (acc_device_nvidia);
+
+  exit (0);
+}
+
+/* { dg-output "" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-86.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-86.c
new file mode 100644
index 00000000000..b8a8ee94a58
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-86.c
@@ -0,0 +1,42 @@
+/* { dg-do run } */
+
+#include <stdlib.h>
+#include <unistd.h>
+#include <openacc.h>
+
+int
+main (int argc, char **argv)
+{
+  if (acc_get_num_devices (acc_device_nvidia) == 0)
+    return 0;
+
+  if (acc_get_current_cuda_device () != 0)
+    abort ();
+
+  acc_init (acc_device_host);
+
+  if (acc_get_current_cuda_device () != 0)
+    abort ();
+
+  acc_shutdown (acc_device_host);
+
+  if (acc_get_num_devices (acc_device_nvidia) == 0)
+    return 0;
+
+  if (acc_get_current_cuda_device () != 0)
+    abort ();
+
+  acc_init (acc_device_nvidia);
+
+  if (acc_get_current_cuda_device () == 0)
+    abort ();
+
+  acc_shutdown (acc_device_nvidia);
+
+  if (acc_get_current_cuda_device () != 0)
+    abort ();
+
+  return 0;
+}
+
+/* { dg-output "" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-87.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-87.c
new file mode 100644
index 00000000000..147d443b54d
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-87.c
@@ -0,0 +1,42 @@
+/* { dg-do run } */
+
+#include <stdlib.h>
+#include <unistd.h>
+#include <openacc.h>
+
+int
+main (int argc, char **argv)
+{
+  if (acc_get_num_devices (acc_device_nvidia) == 0)
+    return 0;
+
+  if (acc_get_current_cuda_context () != 0)
+    abort ();
+
+  acc_init (acc_device_host);
+
+  if (acc_get_current_cuda_context () != 0)
+    abort ();
+
+  acc_shutdown (acc_device_host);
+
+  if (acc_get_num_devices (acc_device_nvidia) == 0)
+    return 0;
+
+  if (acc_get_current_cuda_context () != 0)
+    abort ();
+
+  acc_init (acc_device_nvidia);
+
+  if (acc_get_current_cuda_context () == 0)
+    abort ();
+
+  acc_shutdown (acc_device_nvidia);
+
+  if (acc_get_current_cuda_context () != 0)
+    abort ();
+
+  return 0;
+}
+
+/* { dg-output "" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-88.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-88.c
new file mode 100644
index 00000000000..10f4ad8664a
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-88.c
@@ -0,0 +1,111 @@
+/* { dg-do run } */
+
+#include <stdio.h>
+#include <pthread.h>
+#include <string.h>
+#include <stdlib.h>
+#include <ctype.h>
+#include <openacc.h>
+
+unsigned char *x;
+void *d_x;
+const int N = 256;
+
+static void *
+test (void *arg)
+{
+  int i;
+
+  if (acc_get_current_cuda_context () != NULL)
+    abort ();
+
+  if (acc_is_present (x, N) != 1)
+    abort ();
+
+  memset (x, 0, N);
+
+  acc_copyout (x, N);
+
+  for (i = 0; i < N; i++)
+    {
+      if (x[i] != i)
+	abort ();
+
+      x[i] = N - i - 1;
+    }
+
+  d_x = acc_copyin (x, N);
+
+  return 0;
+}
+
+int
+main (int argc, char **argv)
+{
+  const int nthreads = 1;
+  int i;
+  pthread_attr_t attr;
+  pthread_t *tid;
+
+  if (acc_get_num_devices (acc_device_nvidia) == 0)
+    return 0;
+
+  acc_init (acc_device_nvidia);
+
+  x = (unsigned char *) malloc (N);
+
+  for (i = 0; i < N; i++)
+    {
+      x[i] = i;
+    }
+
+  d_x = acc_copyin (x, N);
+
+  if (acc_is_present (x, N) != 1)
+    abort ();
+
+  if (pthread_attr_init (&attr) != 0)
+    perror ("pthread_attr_init failed");
+
+  tid = (pthread_t *) malloc (nthreads * sizeof (pthread_t));
+
+  for (i = 0; i < nthreads; i++)
+    {
+      if (pthread_create (&tid[i], &attr, &test, (void *) (unsigned long) (i))
+	  != 0)
+	perror ("pthread_create failed");
+    }
+
+  if (pthread_attr_destroy (&attr) != 0)
+    perror ("pthread_attr_destroy failed");
+
+  for (i = 0; i < nthreads; i++)
+    {
+      void *res;
+
+      if (pthread_join (tid[i], &res) != 0)
+	perror ("pthread join failed");
+    }
+
+  if (acc_is_present (x, N) != 1)
+    abort ();
+
+  memset (x, 0, N);
+
+  acc_copyout (x, N);
+
+  for (i = 0; i < N; i++)
+    {
+      if (x[i] != N - i - 1)
+	abort ();
+    }
+
+  if (acc_is_present (x, N) != 0)
+    abort ();
+
+  acc_shutdown (acc_device_nvidia);
+
+  return 0;
+}
+
+/* { dg-output "" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-89.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-89.c
new file mode 100644
index 00000000000..061c4099c2d
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-89.c
@@ -0,0 +1,118 @@
+/* { dg-do run } */
+
+#include <stdio.h>
+#include <pthread.h>
+#include <string.h>
+#include <stdlib.h>
+#include <errno.h>
+#include <ctype.h>
+#include <openacc.h>
+
+unsigned char **x;
+void **d_x;
+const int N = 16;
+const int NTHREADS = 32;
+
+static void *
+test (void *arg)
+{
+  int i;
+  int tid;
+  unsigned char *p;
+  int devnum;
+
+  tid = (int) (long) arg;
+
+  devnum = acc_get_device_num (acc_device_nvidia);
+  acc_set_device_num (devnum, acc_device_nvidia);
+
+  if (acc_get_current_cuda_context () == NULL)
+    abort ();
+
+  p = (unsigned char *) malloc (N);
+
+  for (i = 0; i < N; i++)
+    {
+      p[i] = tid;
+    }
+
+  x[tid] = p;
+
+  d_x[tid] = acc_copyin (p, N);
+
+  return 0;
+}
+
+int
+main (int argc, char **argv)
+{
+  int i;
+  pthread_attr_t attr;
+  pthread_t *tid;
+
+  if (acc_get_num_devices (acc_device_nvidia) == 0)
+    return 0;
+
+  acc_init (acc_device_nvidia);
+
+  x = (unsigned char **) malloc (NTHREADS * N);
+  d_x = (void **) malloc (NTHREADS * N);
+
+  if (pthread_attr_init (&attr) != 0)
+    perror ("pthread_attr_init failed");
+
+  tid = (pthread_t *) malloc (NTHREADS * sizeof (pthread_t));
+
+  for (i = 0; i < NTHREADS; i++)
+    {
+      if (pthread_create (&tid[i], &attr, &test, (void *) (unsigned long) (i))
+	  != 0)
+	perror ("pthread_create failed");
+    }
+
+  if (pthread_attr_destroy (&attr) != 0)
+    perror ("pthread_attr_destroy failed");
+
+  for (i = 0; i < NTHREADS; i++)
+    {
+      void *res;
+
+      if (pthread_join (tid[i], &res) != 0)
+	perror ("pthread join failed");
+    }
+
+  for (i = 0; i < NTHREADS; i++)
+    {
+      if (acc_is_present (x[i], N) != 1)
+	abort ();
+    }
+
+  for (i = 0; i < NTHREADS; i++)
+    {
+      memset (x[i], 0, N);
+      acc_copyout (x[i], N);
+    }
+
+  for (i = 0; i < NTHREADS; i++)
+    {
+      unsigned char *p;
+      int j;
+
+      p = x[i];
+
+      for (j = 0; j < N; j++)
+	{
+	  if (p[j] != i)
+	    abort ();
+	}
+
+      if (acc_is_present (x[i], N) != 0)
+	abort ();
+    }
+
+  acc_shutdown (acc_device_nvidia);
+
+  return 0;
+}
+
+/* { dg-output "" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-9.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-9.c
new file mode 100644
index 00000000000..84045dbe328
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-9.c
@@ -0,0 +1,70 @@
+/* { dg-do run } */
+
+#include <stdlib.h>
+#include <openacc.h>
+
+int
+main (int argc, char **argv)
+{
+  int i;
+  int num_devices;
+  int devnum;
+  acc_device_t devtype = acc_device_host;
+
+#if ACC_DEVICE_TYPE_nvidia
+  devtype = acc_device_nvidia;
+#endif
+
+  num_devices = acc_get_num_devices (devtype);
+  if (num_devices == 0)
+    return 0;
+
+  acc_init (devtype);
+
+  for (i = 0; i < num_devices; i++)
+    {
+      acc_set_device_num (i, devtype);
+      devnum = acc_get_device_num (devtype);
+      if (devnum != i)
+	abort ();
+    }
+
+  acc_shutdown (devtype);
+
+  num_devices = acc_get_num_devices (devtype);
+  if (num_devices == 0)
+    abort ();
+
+  for (i = 0; i < num_devices; i++)
+    {
+      acc_set_device_num (i, devtype);
+      devnum = acc_get_device_num (devtype);
+      if (devnum != i)
+	abort ();
+    }
+
+  acc_shutdown (devtype);
+
+  acc_init (devtype);
+
+  acc_set_device_num (0, devtype);
+
+  devnum = acc_get_device_num (devtype);
+  if (devnum != 0)
+    abort ();
+
+  if (num_devices > 1)
+    {
+      acc_set_device_num (1, (acc_device_t) 0);
+
+      devnum = acc_get_device_num (devtype);
+      if (devnum != 0)
+	abort ();
+  }
+
+  acc_shutdown (devtype);
+
+  return 0;
+}
+
+/* { dg-output "" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-90.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-90.c
new file mode 100644
index 00000000000..d17755bf3eb
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-90.c
@@ -0,0 +1,137 @@
+/* { dg-do run { target openacc_nvidia_accel_selected } } */
+/* { dg-additional-options "-lcuda" } */
+
+#include <pthread.h>
+#include <stdio.h>
+#include <string.h>
+#include <stdlib.h>
+#include <unistd.h>
+#include <errno.h>
+#include <ctype.h>
+#include <openacc.h>
+#include <cuda.h>
+
+unsigned char **x;
+void **d_x;
+const int N = 16;
+const int NTHREADS = 32;
+
+static void *
+test (void *arg)
+{
+  int i;
+  int tid;
+  unsigned char *p;
+  int devnum;
+
+  tid = (int) (long) arg;
+
+  devnum = acc_get_device_num (acc_device_nvidia);
+  acc_set_device_num (devnum, acc_device_nvidia);
+
+  if (acc_get_current_cuda_context () == NULL)
+    abort ();
+
+  p = (unsigned char *) malloc (N);
+
+  for (i = 0; i < N; i++)
+    {
+      p[i] = tid;
+    }
+
+  x[tid] = p;
+
+  d_x[tid] = acc_copyin (p, N);
+
+  acc_wait_all ();
+
+  return 0;
+}
+
+int
+main (int argc, char **argv)
+{
+  int i;
+  pthread_attr_t attr;
+  pthread_t *tid;
+  CUresult r;
+  CUstream s;
+
+  acc_init (acc_device_nvidia);
+
+  x = (unsigned char **) malloc (NTHREADS * N);
+  d_x = (void **) malloc (NTHREADS * N);
+
+  if (pthread_attr_init (&attr) != 0)
+    perror ("pthread_attr_init failed");
+
+  tid = (pthread_t *) malloc (NTHREADS * sizeof (pthread_t));
+
+  r = cuStreamCreate (&s, CU_STREAM_DEFAULT);
+  if (r != CUDA_SUCCESS)
+	{
+	  fprintf (stderr, "cuStreamCreate failed: %d\n", r);
+	  abort ();
+	}
+
+  if (!acc_set_cuda_stream (0, s))
+	  abort ();
+
+  for (i = 0; i < NTHREADS; i++)
+    {
+      if (pthread_create (&tid[i], &attr, &test, (void *) (unsigned long) (i))
+	  != 0)
+	perror ("pthread_create failed");
+    }
+
+  if (pthread_attr_destroy (&attr) != 0)
+    perror ("pthread_attr_destroy failed");
+
+  for (i = 0; i < NTHREADS; i++)
+    {
+      void *res;
+
+      if (pthread_join (tid[i], &res) != 0)
+	perror ("pthread join failed");
+    }
+
+
+  for (i = 0; i < NTHREADS; i++)
+    {
+      if (acc_is_present (x[i], N) != 1)
+	abort ();
+    }
+
+  acc_get_cuda_stream (1);
+
+  for (i = 0; i < NTHREADS; i++)
+    {
+      memset (x[i], 0, N);
+      acc_copyout (x[i], N);
+    }
+
+  acc_wait_all ();
+
+  for (i = 0; i < NTHREADS; i++)
+    {
+      unsigned char *p;
+      int j;
+
+      p = x[i];
+
+      for (j = 0; j < N; j++)
+	{
+	  if (p[j] != i)
+	    abort ();
+	}
+
+      if (acc_is_present (x[i], N) != 0)
+	abort ();
+    }
+
+  acc_shutdown (acc_device_nvidia);
+
+  return 0;
+}
+
+/* { dg-output "" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-91.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-91.c
new file mode 100644
index 00000000000..e00ef4f7206
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-91.c
@@ -0,0 +1,84 @@
+/* { dg-do run { target openacc_nvidia_accel_selected } } */
+/* { dg-additional-options "-lcuda" } */
+
+#include <stdlib.h>
+#include <unistd.h>
+#include <openacc.h>
+#include <sys/time.h>
+#include <stdio.h>
+#include <cuda.h>
+
+int
+main (int argc, char **argv)
+{
+  const int N = 1024 * 1024;
+  int i;
+  unsigned char *h;
+  void *d;
+  float async, sync;
+  struct timeval start, stop;
+  CUresult r;
+  CUstream s;
+
+  acc_init (acc_device_nvidia);
+
+  h = (unsigned char *) malloc (N);
+
+  for (i = 0; i < N; i++)
+    {
+      h[i] = i;
+    }
+
+  d = acc_malloc (N);
+
+  acc_map_data (h, d, N);
+
+  gettimeofday (&start, NULL);
+
+  for (i = 0; i < 100; i++)
+    {
+#pragma acc update device(h[0:N])
+    }
+
+  gettimeofday (&stop, NULL);
+
+  sync = (float) (stop.tv_sec - start.tv_sec);
+  sync += (float) ((stop.tv_usec - start.tv_usec) / 1000000.0);
+
+  gettimeofday (&start, NULL);
+
+  r = cuStreamCreate (&s, CU_STREAM_DEFAULT);
+  if (r != CUDA_SUCCESS)
+	{
+	  fprintf (stderr, "cuStreamCreate failed: %d\n", r);
+	  abort ();
+	}
+
+  if (!acc_set_cuda_stream (0, s))
+	  abort ();
+
+  for (i = 0; i < 100; i++)
+    {
+#pragma acc update device(h[0:N]) async(0)
+    }
+
+  acc_wait_all ();
+
+  gettimeofday (&stop, NULL);
+
+  async = (float) (stop.tv_sec - start.tv_sec);
+  async += (float) ((stop.tv_usec - start.tv_usec) / 1000000.0);
+
+  if (async > (sync * 1.5))
+    abort ();
+
+  acc_free (d);
+
+  free (h);
+
+  acc_shutdown (acc_device_nvidia);
+
+  return 0;
+}
+
+/* { dg-output "" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-92.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-92.c
new file mode 100644
index 00000000000..18193e0e8f9
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/lib-92.c
@@ -0,0 +1,112 @@
+/* { dg-do run } */
+
+#include <pthread.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <errno.h>
+#include <ctype.h>
+#include <openacc.h>
+
+unsigned char **x;
+void **d_x;
+const int N = 32;
+const int NTHREADS = 32;
+
+static void *
+test (void *arg)
+{
+  int i;
+  int tid;
+  unsigned char *p;
+  int devnum;
+
+  tid = (int) (long) arg;
+
+  devnum = acc_get_device_num (acc_device_nvidia);
+  acc_set_device_num (devnum, acc_device_nvidia);
+
+  if (acc_get_current_cuda_context () == NULL)
+    abort ();
+
+  acc_copyout (x[tid], N);
+
+  p = x[tid];
+
+  for (i = 0; i < N; i++)
+    {
+      if (p[i] != i)
+	abort ();
+    }
+
+  return 0;
+}
+
+int
+main (int argc, char **argv)
+{
+  int i;
+  pthread_attr_t attr;
+  pthread_t *tid;
+  unsigned char *p;
+
+  if (acc_get_num_devices (acc_device_nvidia) == 0)
+    return 0;
+
+  acc_init (acc_device_nvidia);
+
+  x = (unsigned char **) malloc (NTHREADS * N);
+  d_x = (void **) malloc (NTHREADS * N);
+
+  for (i = 0; i < N; i++)
+    {
+      int j;
+
+      p = (unsigned char *) malloc (N);
+
+      x[i] = p;
+
+      for (j = 0; j < N; j++)
+	{
+	  p[j] = j;
+	}
+
+      d_x[i] = acc_copyin (p, N);
+    }
+
+  if (pthread_attr_init (&attr) != 0)
+    perror ("pthread_attr_init failed");
+
+  tid = (pthread_t *) malloc (NTHREADS * sizeof (pthread_t));
+
+  acc_get_cuda_stream (1);
+
+  for (i = 0; i < NTHREADS; i++)
+    {
+      if (pthread_create (&tid[i], &attr, &test, (void *) (unsigned long) (i))
+	  != 0)
+	perror ("pthread_create failed");
+    }
+
+  if (pthread_attr_destroy (&attr) != 0)
+    perror ("pthread_attr_destroy failed");
+
+  for (i = 0; i < NTHREADS; i++)
+    {
+      void *res;
+
+      if (pthread_join (tid[i], &res) != 0)
+	perror ("pthread join failed");
+    }
+
+  for (i = 0; i < NTHREADS; i++)
+    {
+      if (acc_is_present (x[i], N) != 0)
+	abort ();
+    }
+
+  acc_shutdown (acc_device_nvidia);
+
+  return 0;
+}
+
+/* { dg-output "" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/nested-1.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/nested-1.c
new file mode 100644
index 00000000000..ededf2ba5b7
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/nested-1.c
@@ -0,0 +1,680 @@
+/* { dg-do run } */
+/* { dg-skip-if "" { *-*-* } { "*" } { "-DACC_MEM_SHARED=0" } } */
+
+#include <openacc.h>
+#include <string.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <stdbool.h>
+
+int
+main (int argc, char **argv)
+{
+    int N = 8;
+    float *a, *b, *c, *d;
+    int i;
+
+    a = (float *) malloc (N * sizeof (float));
+    b = (float *) malloc (N * sizeof (float));
+    c = (float *) malloc (N * sizeof (float));
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 3.0;
+        b[i] = 0.0;
+    }
+
+#pragma acc data copyin (a[0:N]) copyout (b[0:N])
+    {
+#pragma acc parallel
+        {
+            int ii;
+
+            for (ii = 0; ii < N; ii++)
+                b[ii] = a[ii];
+        }
+    }
+
+    for (i = 0; i < N; i++)
+    {
+        if (b[i] != 3.0)
+            abort ();
+    }
+
+    if (acc_is_present (&a[0], (N * sizeof (float))))
+      abort ();
+
+    if (acc_is_present (&b[0], (N * sizeof (float))))
+      abort ();
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 5.0;
+        b[i] = 1.0;
+    }
+
+#pragma acc data copyin (a[0:N]) copyout (b[0:N])
+    {
+#pragma acc parallel
+        {
+            int ii;
+
+            for (ii = 0; ii < N; ii++)
+                b[ii] = a[ii];
+        }
+    }
+
+    for (i = 0; i < N; i++)
+    {
+        if (b[i] != 5.0)
+            abort ();
+    }
+
+    if (acc_is_present (&a[0], (N * sizeof (float))))
+      abort ();
+
+    if (acc_is_present (&b[0], (N * sizeof (float))))
+      abort ();
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 6.0;
+        b[i] = 0.0;
+    }
+
+    d = (float *) acc_copyin (&a[0], N * sizeof (float));
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 9.0;
+    }
+
+#pragma acc data present_or_copyin (a[0:N]) copyout (b[0:N])
+    {
+#pragma acc parallel
+        {
+            int ii;
+
+            for (ii = 0; ii < N; ii++)
+                b[ii] = a[ii];
+        }
+    }
+
+    for (i = 0; i < N; i++)
+    {
+        if (b[i] != 6.0)
+            abort ();
+    }
+
+    if (!acc_is_present (&a[0], (N * sizeof (float))))
+      abort ();
+
+    if (acc_is_present (&b[0], (N * sizeof (float))))
+      abort ();
+
+    acc_free (d);
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 6.0;
+        b[i] = 0.0;
+    }
+
+#pragma acc data copyin (a[0:N]) present_or_copyout (b[0:N])
+    {
+#pragma acc parallel
+        {
+            int ii;
+
+            for (ii = 0; ii < N; ii++)
+                b[ii] = a[ii];
+        }
+    }
+
+    for (i = 0; i < N; i++)
+    {
+        if (b[i] != 6.0)
+            abort ();
+    }
+
+    if (acc_is_present (&a[0], (N * sizeof (float))))
+      abort ();
+
+    if (acc_is_present (&b[0], (N * sizeof (float))))
+      abort ();
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 5.0;
+        b[i] = 2.0;
+    }
+
+    d = (float *) acc_copyin (&b[0], N * sizeof (float));
+
+#pragma acc data copyin (a[0:N]) present_or_copyout (b[0:N])
+    {
+#pragma acc parallel
+        {
+            int ii;
+
+            for (ii = 0; ii < N; ii++)
+                b[ii] = a[ii];
+        }
+    }
+
+    for (i = 0; i < N; i++)
+    {
+        if (a[i] != 5.0)
+            abort ();
+
+        if (b[i] != 2.0)
+            abort ();
+    }
+
+    if (acc_is_present (&a[0], (N * sizeof (float))))
+      abort ();
+
+    if (!acc_is_present (&b[0], (N * sizeof (float))))
+      abort ();
+
+    acc_free (d);
+
+    if (acc_is_present (&b[0], (N * sizeof (float))))
+      abort ();
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 3.0;
+        b[i] = 4.0;
+    }
+
+#pragma acc data copy (a[0:N]) copyout (b[0:N])
+    {
+#pragma acc parallel
+        {
+            int ii;
+
+            for (ii = 0; ii < N; ii++)
+            {
+                a[ii] = a[ii] + 1;
+                b[ii] = a[ii] + 2;
+            }
+        }
+    }
+
+    for (i = 0; i < N; i++)
+    {
+        if (a[i] != 4.0)
+            abort ();
+
+        if (b[i] != 6.0)
+            abort ();
+    }
+
+    if (acc_is_present (&a[0], (N * sizeof (float))))
+      abort ();
+
+    if (acc_is_present (&b[0], (N * sizeof (float))))
+      abort ();
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 4.0;
+        b[i] = 7.0;
+    }
+
+#pragma acc data present_or_copy (a[0:N]) present_or_copy (b[0:N])
+    {
+#pragma acc parallel
+        {
+            int ii;
+
+            for (ii = 0; ii < N; ii++)
+            {
+                a[ii] = a[ii] + 1;
+                b[ii] = b[ii] + 2;
+            }
+        }
+    }
+
+    for (i = 0; i < N; i++)
+    {
+        if (a[i] != 5.0)
+            abort ();
+
+        if (b[i] != 9.0)
+            abort ();
+    }
+
+    if (acc_is_present (&a[0], (N * sizeof (float))))
+      abort ();
+
+    if (acc_is_present (&b[0], (N * sizeof (float))))
+      abort ();
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 3.0;
+        b[i] = 7.0;
+    }
+
+    d = (float *) acc_copyin (&a[0], N * sizeof (float));
+    d = (float *) acc_copyin (&b[0], N * sizeof (float));
+
+#pragma acc data present_or_copy (a[0:N]) present_or_copy (b[0:N])
+    {
+#pragma acc parallel
+        {
+            int ii;
+
+            for (ii = 0; ii < N; ii++)
+            {
+                a[ii] = a[ii] + 1;
+                b[ii] = b[ii] + 2;
+            }
+        }
+    }
+
+    for (i = 0; i < N; i++)
+    {
+        if (a[i] != 3.0)
+            abort ();
+
+        if (b[i] != 7.0)
+            abort ();
+    }
+
+    if (!acc_is_present (&a[0], (N * sizeof (float))))
+      abort ();
+
+    if (!acc_is_present (&b[0], (N * sizeof (float))))
+      abort ();
+
+    d = (float *) acc_deviceptr (&a[0]);
+    acc_unmap_data (&a[0]);
+    acc_free (d);
+
+    d = (float *) acc_deviceptr (&b[0]);
+    acc_unmap_data (&b[0]);
+    acc_free (d);
+
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 3.0;
+        b[i] = 7.0;
+    }
+
+#pragma acc data copyin (a[0:N]) create (c[0:N]) copyout (b[0:N])
+    {
+#pragma acc parallel
+        {
+            int ii;
+
+            for (ii = 0; ii < N; ii++)
+            {
+                c[ii] = a[ii];
+                b[ii] = c[ii];
+            }
+        }
+    }
+
+    for (i = 0; i < N; i++)
+    {
+        if (a[i] != 3.0)
+            abort ();
+
+        if (b[i] != 3.0)
+            abort ();
+    }
+
+    if (acc_is_present (&a[0], (N * sizeof (float))))
+      abort ();
+
+    if (acc_is_present (&b[0], (N * sizeof (float))))
+      abort ();
+
+    if (acc_is_present (&c[0], (N * sizeof (float))))
+      abort ();
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 4.0;
+        b[i] = 8.0;
+    }
+
+#pragma acc data copyin (a[0:N]) present_or_create (c[0:N]) copyout (b[0:N])
+    {
+#pragma acc parallel
+        {
+            int ii;
+
+            for (ii = 0; ii < N; ii++)
+            {
+                c[ii] = a[ii];
+                b[ii] = c[ii];
+            }
+        }
+    }
+
+    for (i = 0; i < N; i++)
+    {
+        if (a[i] != 4.0)
+            abort ();
+
+        if (b[i] != 4.0)
+            abort ();
+    }
+
+    if (acc_is_present (&a[0], (N * sizeof (float))))
+      abort ();
+
+    if (acc_is_present (&b[0], (N * sizeof (float))))
+      abort ();
+
+    if (acc_is_present (&c[0], (N * sizeof (float))))
+      abort ();
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 2.0;
+        b[i] = 5.0;
+    }
+
+    d = (float *) acc_malloc (N * sizeof (float));
+    acc_map_data (c, d, N * sizeof (float));
+
+#pragma acc data copyin (a[0:N]) present_or_create (c[0:N]) copyout (b[0:N])
+    {
+#pragma acc parallel
+        {
+            int ii;
+
+            for (ii = 0; ii < N; ii++)
+            {
+                c[ii] = a[ii];
+                b[ii] = c[ii];
+            }
+        }
+    }
+
+    for (i = 0; i < N; i++)
+    {
+        if (a[i] != 2.0)
+            abort ();
+
+        if (b[i] != 2.0)
+            abort ();
+    }
+
+    if (acc_is_present (a, (N * sizeof (float))))
+      abort ();
+
+    if (acc_is_present (b, (N * sizeof (float))))
+      abort ();
+
+    if (!acc_is_present (c, (N * sizeof (float))))
+      abort ();
+
+    d = (float *) acc_deviceptr (c);
+
+    acc_unmap_data (c);
+
+    acc_free (d);
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 4.0;
+        b[i] = 8.0;
+    }
+
+    d = (float *) acc_malloc (N * sizeof (float));
+    acc_map_data (c, d, N * sizeof (float));
+
+#pragma acc data copyin (a[0:N]) present (c[0:N]) copyout (b[0:N])
+    {
+#pragma acc parallel
+        {
+            int ii;
+
+            for (ii = 0; ii < N; ii++)
+            {
+                c[ii] = a[ii];
+                b[ii] = c[ii];
+            }
+        }
+    }
+
+    for (i = 0; i < N; i++)
+    {
+        if (a[i] != 4.0)
+            abort ();
+
+        if (b[i] != 4.0)
+            abort ();
+    }
+
+    if (acc_is_present (a, (N * sizeof (float))))
+      abort ();
+
+    if (acc_is_present (b, (N * sizeof (float))))
+      abort ();
+
+    if (!acc_is_present (c, (N * sizeof (float))))
+      abort ();
+
+    acc_unmap_data (c);
+
+    if (acc_is_present (c, (N * sizeof (float))))
+      abort ();
+
+    acc_free (d);
+
+    d = (float *) acc_malloc (N * sizeof (float));
+    acc_map_data (c, d, N * sizeof (float));
+
+    if (!acc_is_present (c, (N * sizeof (float))))
+      abort ();
+
+    d = (float *) acc_malloc (N * sizeof (float));
+    acc_map_data (b, d, N * sizeof (float));
+
+    if (!acc_is_present (b, (N * sizeof (float))))
+      abort ();
+
+    d = (float *) acc_malloc (N * sizeof (float));
+    acc_map_data (a, d, N * sizeof (float));
+
+    if (!acc_is_present (a, (N * sizeof (float))))
+      abort ();
+
+#pragma acc data present (a[0:N]) present (c[0:N]) present (b[0:N])
+    {
+#pragma acc parallel
+        {
+            int ii;
+
+            for (ii = 0; ii < N; ii++)
+            {
+                a[ii] = 1.0;
+                c[ii] = 2.0;
+                b[ii] = 4.0;
+            }
+        }
+    }
+
+    if (!acc_is_present (a, (N * sizeof (float))))
+      abort ();
+
+    if (!acc_is_present (b, (N * sizeof (float))))
+      abort ();
+
+    if (!acc_is_present (c, (N * sizeof (float))))
+      abort ();
+
+    acc_copyout (b, N * sizeof (float));
+
+    for (i = 0; i < N; i++)
+    {
+        if (a[i] != 4.0)
+            abort ();
+
+        if (b[i] != 4.0)
+            abort ();
+    }
+
+    d = (float *) acc_deviceptr (a);
+
+    acc_unmap_data (a);
+
+    acc_free (d);
+
+    d = (float *) acc_deviceptr (c);
+
+    acc_unmap_data (c);
+
+    acc_free (d);
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 3.0;
+        b[i] = 6.0;
+    }
+
+    d = (float *) acc_malloc (N * sizeof (float));
+
+#pragma acc parallel copyin (a[0:N]) deviceptr (d) copyout (b[0:N])
+    {
+        int ii;
+
+        for (ii = 0; ii < N; ii++)
+        {
+            d[ii] = a[ii];
+            b[ii] = d[ii];
+        }
+    }
+
+    for (i = 0; i < N; i++)
+    {
+        if (a[i] != 3.0)
+            abort ();
+
+        if (b[i] != 3.0)
+            abort ();
+    }
+
+    if (acc_is_present (a, (N * sizeof (float))))
+      abort ();
+
+    if (acc_is_present (b, (N * sizeof (float))))
+      abort ();
+
+    acc_free (d);
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 6.0;
+        b[i] = 0.0;
+    }
+
+    d = (float *) acc_copyin (&a[0], N * sizeof (float));
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 9.0;
+    }
+
+#pragma acc data pcopyin (a[0:N]) copyout (b[0:N])
+    {
+#pragma acc parallel
+        {
+            int ii;
+
+            for (ii = 0; ii < N; ii++)
+                b[ii] = a[ii];
+        }
+    }
+
+    for (i = 0; i < N; i++)
+    {
+        if (b[i] != 6.0)
+            abort ();
+    }
+
+    if (!acc_is_present (&a[0], (N * sizeof (float))))
+      abort ();
+
+    if (acc_is_present (&b[0], (N * sizeof (float))))
+      abort ();
+
+    acc_free (d);
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 6.0;
+        b[i] = 0.0;
+    }
+
+#pragma acc data copyin (a[0:N]) pcopyout (b[0:N])
+    {
+#pragma acc parallel
+        {
+            int ii;
+
+            for (ii = 0; ii < N; ii++)
+                b[ii] = a[ii];
+        }
+    }
+
+    for (i = 0; i < N; i++)
+    {
+        if (b[i] != 6.0)
+            abort ();
+    }
+
+    if (acc_is_present (&a[0], (N * sizeof (float))))
+      abort ();
+
+    if (acc_is_present (&b[0], (N * sizeof (float))))
+      abort ();
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 5.0;
+        b[i] = 7.0;
+    }
+
+#pragma acc data copyin (a[0:N]) pcreate (c[0:N]) copyout (b[0:N])
+    {
+#pragma acc parallel
+        {
+            int ii;
+
+            for (ii = 0; ii < N; ii++)
+            {
+                c[ii] = a[ii];
+                b[ii] = c[ii];
+            }
+        }
+    }
+
+    for (i = 0; i < N; i++)
+    {
+        if (a[i] != 5.0)
+            abort ();
+
+        if (b[i] != 5.0)
+            abort ();
+    }
+
+    if (acc_is_present (&a[0], (N * sizeof (float))))
+      abort ();
+
+    if (acc_is_present (&b[0], (N * sizeof (float))))
+      abort ();
+
+    if (acc_is_present (&c[0], (N * sizeof (float))))
+      abort ();
+
+    return 0;
+}
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/nested-2.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/nested-2.c
new file mode 100644
index 00000000000..c16459826af
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/nested-2.c
@@ -0,0 +1,141 @@
+/* { dg-do run } */
+
+#include <stdlib.h>
+
+int
+main (int argc, char *argv[])
+{
+#define N 10
+    char a[N];
+    int i;
+
+    for (i = 0; i < N; ++i)
+       a[i] = 0;
+
+#pragma acc data copy (a)
+    {
+#pragma acc parallel present (a)
+        {
+            int j;
+
+            for (j = 0; j < N; ++j)
+                a[j] = j;
+        }
+    }
+
+    for (i = 0; i < N; ++i)
+    {
+        if (a[i] != i)
+            abort ();
+    }
+
+    for (i = 0; i < N; ++i)
+       a[i] = 0;
+
+#pragma acc data copy (a)
+    {
+#pragma acc kernels present (a)
+        {
+            int j;
+
+            for (j = 0; j < N; ++j)
+                a[j] = j;
+        }
+    }
+
+    for (i = 0; i < N; ++i)
+    {
+        if (a[i] != i)
+            abort ();
+    }
+
+    for (i = 0; i < N; ++i)
+        a[i] = 0;
+
+#pragma acc data copy (a)
+    {
+#pragma acc data present (a)
+        {
+#pragma acc parallel present (a)
+            {
+                int j;
+
+                for (j = 0; j < N; ++j)
+                    a[j] = j;
+            }
+        }
+    }
+
+    for (i = 0; i < N; ++i)
+    {
+        if (a[i] != i)
+            abort ();
+    }
+
+#pragma acc data copy (a)
+    {
+#pragma acc data present (a)
+        {
+#pragma acc kernels present (a)
+            {
+                int j;
+
+                for (j = 0; j < N; ++j)
+                    a[j] = j;
+            }
+        }
+    }
+
+    for (i = 0; i < N; ++i)
+    {
+        if (a[i] != i)
+            abort ();
+    }
+
+    for (i = 0; i < N; ++i)
+        a[i] = 0;
+
+#pragma acc enter data copyin (a)
+
+#pragma acc data present (a)
+    {
+#pragma acc parallel present (a)
+        {
+            int j;
+
+            for (j = 0; j < N; ++j)
+                a[j] = j;
+        }
+    }
+
+#pragma acc exit data copyout (a)
+
+    for (i = 0; i < N; ++i)
+    {
+        if (a[i] != i)
+            abort ();
+    }
+
+#pragma acc enter data copyin (a)
+
+#pragma acc data present (a)
+    {
+#pragma acc kernels present (a)
+        {
+            int j;
+
+            for (j = 0; j < N; ++j)
+                a[j] = j;
+        }
+    }
+
+#pragma acc exit data copyout (a)
+
+    for (i = 0; i < N; ++i)
+    {
+        if (a[i] != i)
+            abort ();
+    }
+
+    return 0;
+}
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/offset-1.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/offset-1.c
new file mode 100644
index 00000000000..0bae23a46f7
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/offset-1.c
@@ -0,0 +1,97 @@
+/* { dg-do run } */
+/* { dg-skip-if "" { *-*-* } { "*" } { "-DACC_MEM_SHARED=0" } } */
+
+#include <openacc.h>
+#include <string.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <stdbool.h>
+
+int
+main(int argc, char **argv)
+{
+    int N = 8;
+    float *a, *b;
+    int i;
+
+    a = (float *) malloc(N * sizeof (float));
+    b = (float *) malloc(N * sizeof (float));
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 2.0;
+        b[i] = 5.0;
+    }
+
+#pragma acc parallel copyin(a[2:4]) copyout(b[2:4])
+    {
+        b[2] = a[2];
+        b[3] = a[3];
+    }
+
+    for (i = 2; i < 4; i++)
+    {
+        if (a[i] != 2.0)
+            abort();
+
+        if (b[i] != 2.0)
+            abort();
+    }
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 3.0;
+        b[i] = 1.0;
+    }
+
+#pragma acc parallel copyin(a[0:4]) copyout(b[0:4])
+    {
+        b[0] = a[0];
+        b[1] = a[1];
+        b[2] = a[2];
+        b[3] = a[3];
+    }
+
+    for (i = 0; i < 4; i++)
+    {
+        if (a[i] != 3.0)
+            abort();
+
+        if (b[i] != 3.0)
+            abort();
+    }
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 9.0;
+        b[i] = 6.0;
+    }
+
+#pragma acc parallel copyin(a[0:4]) copyout(b[4:4])
+    {
+        b[4] = a[0];
+        b[5] = a[1];
+        b[6] = a[2];
+        b[7] = a[3];
+    }
+
+    for (i = 0; i < 4; i++)
+    {
+        if (a[i] != 9.0)
+            abort();
+    }
+
+    for (i = 4; i < 8; i++)
+    {
+        if (b[i] != 9.0)
+            abort();
+    }
+
+    if (acc_is_present (a, (N * sizeof (float))))
+      abort();
+
+    if (acc_is_present (b, (N * sizeof (float))))
+      abort();
+
+    return 0;
+}
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/parallel-1.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/parallel-1.c
new file mode 100644
index 00000000000..fd9df33748f
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/parallel-1.c
@@ -0,0 +1,206 @@
+/* { dg-do run } */
+
+#include <stdlib.h>
+
+int i;
+
+int main(void)
+{
+  int j, v;
+
+  i = -1;
+  j = -2;
+  v = 0;
+#pragma acc parallel /* copyout */ present_or_copyout (v) copyin (i, j)
+  {
+    if (i != -1 || j != -2)
+      abort ();
+    i = 2;
+    j = 1;
+    if (i != 2 || j != 1)
+      abort ();
+    v = 1;
+  }
+#if ACC_MEM_SHARED
+  if (v != 1 || i != 2 || j != 1)
+    abort ();
+#else
+  if (v != 1 || i != -1 || j != -2)
+    abort ();
+#endif
+
+  i = -1;
+  j = -2;
+  v = 0;
+#pragma acc parallel /* copyout */ present_or_copyout (v) copyout (i, j)
+  {
+    i = 2;
+    j = 1;
+    if (i != 2 || j != 1)
+      abort ();
+    v = 1;
+  }
+  if (v != 1 || i != 2 || j != 1)
+    abort ();
+
+  i = -1;
+  j = -2;
+  v = 0;
+#pragma acc parallel /* copyout */ present_or_copyout (v) copy (i, j)
+  {
+    if (i != -1 || j != -2)
+      abort ();
+    i = 2;
+    j = 1;
+    if (i != 2 || j != 1)
+      abort ();
+    v = 1;
+  }
+  if (v != 1 || i != 2 || j != 1)
+    abort ();
+
+  i = -1;
+  j = -2;
+  v = 0;
+#pragma acc parallel /* copyout */ present_or_copyout (v) create (i, j)
+  {
+    i = 2;
+    j = 1;
+    if (i != 2 || j != 1)
+      abort ();
+    v = 1;
+  }
+#if ACC_MEM_SHARED
+  if (v != 1 || i != 2 || j != 1)
+    abort ();
+#else
+  if (v != 1 || i != -1 || j != -2)
+    abort ();
+#endif
+
+  i = -1;
+  j = -2;
+  v = 0;
+#pragma acc parallel /* copyout */ present_or_copyout (v) present_or_copyin (i, j)
+  {
+    if (i != -1 || j != -2)
+      abort ();
+    i = 2;
+    j = 1;
+    if (i != 2 || j != 1)
+      abort ();
+    v = 1;
+  }
+  if (v != 1)
+    abort ();
+#if ACC_MEM_SHARED
+  if (v != 1 || i != 2 || j != 1)
+    abort ();
+#else
+  if (v != 1 || i != -1 || j != -2)
+    abort ();
+#endif
+
+  i = -1;
+  j = -2;
+  v = 0;
+#pragma acc parallel /* copyout */ present_or_copyout (v) present_or_copyout (i, j)
+  {
+    i = 2;
+    j = 1;
+    if (i != 2 || j != 1)
+      abort ();
+    v = 1;
+  }
+  if (v != 1 || i != 2 || j != 1)
+    abort ();
+
+  i = -1;
+  j = -2;
+  v = 0;
+#pragma acc parallel /* copyout */ present_or_copyout (v) present_or_copy (i, j)
+  {
+    if (i != -1 || j != -2)
+      abort ();
+    i = 2;
+    j = 1;
+    if (i != 2 || j != 1)
+      abort ();
+    v = 1;
+  }
+  if (v != 1 || i != 2 || j != 1)
+    abort ();
+
+  i = -1;
+  j = -2;
+  v = 0;
+#pragma acc parallel /* copyout */ present_or_copyout (v) present_or_create (i, j)
+  {
+    i = 2;
+    j = 1;
+    if (i != 2 || j != 1)
+      abort ();
+    v = 1;
+  }
+  if (v != 1)
+    abort ();
+#if ACC_MEM_SHARED
+  if (v != 1 || i != 2 || j != 1)
+    abort ();
+#else
+  if (v != 1 || i != -1 || j != -2)
+    abort ();
+#endif
+
+  i = -1;
+  j = -2;
+  v = 0;
+
+#pragma acc data copyin (i, j)
+  {
+#pragma acc parallel /* copyout */ present_or_copyout (v) present (i, j)
+    {
+      if (i != -1 || j != -2)
+        abort ();
+      i = 2;
+      j = 1;
+      if (i != 2 || j != 1)
+        abort ();
+      v = 1;
+    }
+  }
+#if ACC_MEM_SHARED
+  if (v != 1 || i != 2 || j != 1)
+    abort ();
+#else
+  if (v != 1 || i != -1 || j != -2)
+    abort ();
+#endif
+
+  i = -1;
+  j = -2;
+  v = 0;
+
+#pragma acc data copyin(i, j)
+  {
+#pragma acc parallel /* copyout */ present_or_copyout (v)
+    {
+      if (i != -1 || j != -2)
+        abort ();
+      i = 2;
+      j = 1;
+      if (i != 2 || j != 1)
+        abort ();
+      v = 1;
+    }
+  }
+#if ACC_MEM_SHARED
+  if (v != 1 || i != 2 || j != 1)
+    abort ();
+#else
+  if (v != 1 || i != -1 || j != -2)
+    abort ();
+#endif
+
+  return 0;
+}
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/parallel-empty.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/parallel-empty.c
new file mode 100644
index 00000000000..8e3bb43f952
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/parallel-empty.c
@@ -0,0 +1,6 @@
+int
+main (void)
+{
+#pragma acc parallel
+  ;
+}
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/pointer-align-1.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/pointer-align-1.c
new file mode 100644
index 00000000000..f7d5b9bd6e8
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/pointer-align-1.c
@@ -0,0 +1,35 @@
+/* { dg-do run } */
+
+/* PR middle-end/63247 */
+
+#include <stdlib.h>
+
+int
+main(int argc, char **argv)
+{
+#define N 4
+    short a[N];
+
+    a[0] = 10;
+    a[1] = 10;
+    a[2] = 10;
+    a[3] = 10;
+
+#pragma acc parallel copy(a[1:N-1])
+    {
+      a[1] = 51;
+      a[2] = 52;
+      a[3] = 53;
+    }
+
+    if (a[0] != 10)
+      abort ();
+    if (a[1] != 51)
+      abort ();
+    if (a[2] != 52)
+      abort ();
+    if (a[3] != 53)
+      abort ();
+
+    return 0;
+}
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/present-1.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/present-1.c
new file mode 100644
index 00000000000..f331f1f6370
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/present-1.c
@@ -0,0 +1,48 @@
+/* { dg-do run } */
+/* { dg-skip-if "" { *-*-* } { "*" } { "-DACC_MEM_SHARED=0" } } */
+
+#include <openacc.h>
+#include <string.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <stdbool.h>
+
+int
+main (int argc, char **argv)
+{
+    int N = 8;
+    float *a, *b, *c, *d;
+    int i;
+
+    a = (float *) malloc (N * sizeof (float));
+    b = (float *) malloc (N * sizeof (float));
+    c = (float *) malloc (N * sizeof (float));
+
+    d = (float *) acc_malloc (N * sizeof (float));
+    acc_map_data (c, d, N * sizeof (float));
+
+#pragma acc data present (a[0:N]) present (c[0:N]) present (b[0:N])
+    {
+#pragma acc parallel
+        {
+            int ii;
+
+            for (ii = 0; ii < N; ii++)
+            {
+                c[ii] = a[ii];
+                b[ii] = c[ii];
+            }
+        }
+    }
+
+    d = (float *) acc_deviceptr (c);
+    acc_unmap_data (c);
+    acc_free (d);
+
+    free (a);
+    free (b);
+    free (c);
+
+    return 0;
+}
+/* { dg-shouldfail "libgomp: present clause: !acc_is_present" } */
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/present-2.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/present-2.c
new file mode 100644
index 00000000000..41efa70e4ec
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/present-2.c
@@ -0,0 +1,48 @@
+/* { dg-do run } */
+/* { dg-skip-if "" { *-*-* } { "*" } { "-DACC_MEM_SHARED=0" } } */
+
+#include <openacc.h>
+#include <stdlib.h>
+
+int
+main (int argc, char **argv)
+{
+  int N = 8;
+  float *a, *b;
+  int i;
+
+  a = (float *) malloc (N * sizeof (float));
+  b = (float *) malloc (N * sizeof (float));
+
+  for (i = 0; i < N; i++)
+    {
+      a[i] = 4.0;
+      b[i] = 0.0;
+    }
+
+#pragma acc data copyin(a[0:N]) copyout(b[0:N])
+  {
+
+#pragma acc parallel present(a[0:N])
+    {
+      int ii;
+
+      for (ii = 0; ii < N; ii++)
+	{
+	  b[ii] = a[ii];
+	}
+    }
+
+  }
+
+  for (i = 0; i < N; i++)
+    {
+      if (a[i] != 4.0)
+	abort ();
+
+      if (b[i] != 4.0)
+	abort ();
+    }
+
+  return 0;
+}
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/reduction-1.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/reduction-1.c
new file mode 100644
index 00000000000..acf95402ba1
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/reduction-1.c
@@ -0,0 +1,174 @@
+/* { dg-do run } */
+
+/* Integer reductions.  */
+
+#include <stdlib.h>
+#include <stdbool.h>
+
+#define vl 32
+
+int
+main(void)
+{
+  const int n = 1000;
+  int i;
+  int vresult, result, array[n];
+  bool lvresult, lresult;
+
+  for (i = 0; i < n; i++)
+    array[i] = i;
+
+  result = 0;
+  vresult = 0;
+
+  /* '+' reductions.  */
+#pragma acc parallel vector_length (vl)
+#pragma acc loop reduction (+:result)
+  for (i = 0; i < n; i++)
+    result += array[i];
+
+  /* Verify the reduction.  */
+  for (i = 0; i < n; i++)
+    vresult += array[i];
+
+  if (result != vresult)
+    abort ();
+
+  result = 0;
+  vresult = 0;
+
+  /* '*' reductions.  */
+#pragma acc parallel vector_length (vl)
+#pragma acc loop reduction (*:result)
+  for (i = 0; i < n; i++)
+    result *= array[i];
+
+  /* Verify the reduction.  */
+  for (i = 0; i < n; i++)
+    vresult *= array[i];
+
+  if (result != vresult)
+    abort ();
+
+//   result = 0;
+//   vresult = 0;
+// 
+//   /* 'max' reductions.  */
+// #pragma acc parallel vector_length (vl)
+// #pragma acc loop reduction (+:result)
+//   for (i = 0; i < n; i++)
+//       result = result > array[i] ? result : array[i];
+// 
+//   /* Verify the reduction.  */
+//   for (i = 0; i < n; i++)
+//       vresult = vresult > array[i] ? vresult : array[i];
+// 
+//   printf("%d != %d\n", result, vresult);
+//   if (result != vresult)
+//     abort ();
+// 
+//   result = 0;
+//   vresult = 0;
+// 
+//   /* 'min' reductions.  */
+// #pragma acc parallel vector_length (vl)
+// #pragma acc loop reduction (+:result)
+//   for (i = 0; i < n; i++)
+//       result = result < array[i] ? result : array[i];
+// 
+//   /* Verify the reduction.  */
+//   for (i = 0; i < n; i++)
+//       vresult = vresult < array[i] ? vresult : array[i];
+// 
+//   printf("%d != %d\n", result, vresult);
+//   if (result != vresult)
+//     abort ();
+
+  result = 0;
+  vresult = 0;
+
+  /* '&' reductions.  */
+#pragma acc parallel vector_length (vl)
+#pragma acc loop reduction (&:result)
+  for (i = 0; i < n; i++)
+    result &= array[i];
+
+  /* Verify the reduction.  */
+  for (i = 0; i < n; i++)
+    vresult &= array[i];
+
+  if (result != vresult)
+    abort ();
+
+  result = 0;
+  vresult = 0;
+
+  /* '|' reductions.  */
+#pragma acc parallel vector_length (vl)
+#pragma acc loop reduction (|:result)
+  for (i = 0; i < n; i++)
+    result |= array[i];
+
+  /* Verify the reduction.  */
+  for (i = 0; i < n; i++)
+    vresult |= array[i];
+
+  if (result != vresult)
+    abort ();
+
+  result = 0;
+  vresult = 0;
+
+  /* '^' reductions.  */
+#pragma acc parallel vector_length (vl)
+#pragma acc loop reduction (^:result)
+  for (i = 0; i < n; i++)
+    result ^= array[i];
+
+  /* Verify the reduction.  */
+  for (i = 0; i < n; i++)
+    vresult ^= array[i];
+
+  if (result != vresult)
+    abort ();
+
+  result = 5;
+  vresult = 5;
+
+  lresult = false;
+  lvresult = false;
+
+  /* '&&' reductions.  */
+#pragma acc parallel vector_length (vl)
+#pragma acc loop reduction (&&:lresult)
+  for (i = 0; i < n; i++)
+    lresult = lresult && (result > array[i]);
+
+  /* Verify the reduction.  */
+  for (i = 0; i < n; i++)
+    lvresult = lresult && (result > array[i]);
+
+  if (lresult != lvresult)
+    abort ();
+
+  result = 5;
+  vresult = 5;
+
+  lresult = false;
+  lvresult = false;
+
+  /* '||' reductions.  */
+#pragma acc parallel vector_length (vl)
+#pragma acc loop reduction (||:lresult)
+  for (i = 0; i < n; i++)
+    lresult = lresult || (result > array[i]);
+
+  /* Verify the reduction.  */
+  for (i = 0; i < n; i++)
+    lvresult = lresult || (result > array[i]);
+
+  if (lresult != lvresult)
+    abort ();
+
+  return 0;
+}
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/reduction-2.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/reduction-2.c
new file mode 100644
index 00000000000..c2ec110abd2
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/reduction-2.c
@@ -0,0 +1,126 @@
+/* { dg-do run } */
+
+/* float reductions.  */
+
+#include <stdlib.h>
+#include <stdbool.h>
+#include <math.h>
+
+#define vl 32
+
+int
+main(void)
+{
+  const int n = 1000;
+  int i;
+  float vresult, result, array[n];
+  bool lvresult, lresult;
+
+  for (i = 0; i < n; i++)
+    array[i] = i;
+
+  result = 0;
+  vresult = 0;
+
+  /* '+' reductions.  */
+#pragma acc parallel vector_length (vl)
+#pragma acc loop reduction (+:result)
+  for (i = 0; i < n; i++)
+    result += array[i];
+
+  /* Verify the reduction.  */
+  for (i = 0; i < n; i++)
+    vresult += array[i];
+
+  if (result != vresult)
+    abort ();
+
+  result = 0;
+  vresult = 0;
+
+  /* '*' reductions.  */
+#pragma acc parallel vector_length (vl)
+#pragma acc loop reduction (*:result)
+  for (i = 0; i < n; i++)
+    result *= array[i];
+
+  /* Verify the reduction.  */
+  for (i = 0; i < n; i++)
+    vresult *= array[i];
+
+  if (fabs(result - vresult) > .0001)
+    abort ();
+//   result = 0;
+//   vresult = 0;
+// 
+//   /* 'max' reductions.  */
+// #pragma acc parallel vector_length (vl)
+// #pragma acc loop reduction (+:result)
+//   for (i = 0; i < n; i++)
+//       result = result > array[i] ? result : array[i];
+// 
+//   /* Verify the reduction.  */
+//   for (i = 0; i < n; i++)
+//       vresult = vresult > array[i] ? vresult : array[i];
+// 
+//   printf("%d != %d\n", result, vresult);
+//   if (result != vresult)
+//     abort ();
+// 
+//   result = 0;
+//   vresult = 0;
+// 
+//   /* 'min' reductions.  */
+// #pragma acc parallel vector_length (vl)
+// #pragma acc loop reduction (+:result)
+//   for (i = 0; i < n; i++)
+//       result = result < array[i] ? result : array[i];
+// 
+//   /* Verify the reduction.  */
+//   for (i = 0; i < n; i++)
+//       vresult = vresult < array[i] ? vresult : array[i];
+// 
+//   printf("%d != %d\n", result, vresult);
+//   if (result != vresult)
+//     abort ();
+
+  result = 5;
+  vresult = 5;
+
+  lresult = false;
+  lvresult = false;
+
+  /* '&&' reductions.  */
+#pragma acc parallel vector_length (vl)
+#pragma acc loop reduction (&&:lresult)
+  for (i = 0; i < n; i++)
+    lresult = lresult && (result > array[i]);
+
+  /* Verify the reduction.  */
+  for (i = 0; i < n; i++)
+    lvresult = lresult && (result > array[i]);
+
+  if (lresult != lvresult)
+    abort ();
+
+  result = 5;
+  vresult = 5;
+
+  lresult = false;
+  lvresult = false;
+
+  /* '||' reductions.  */
+#pragma acc parallel vector_length (vl)
+#pragma acc loop reduction (||:lresult)
+  for (i = 0; i < n; i++)
+    lresult = lresult || (result > array[i]);
+
+  /* Verify the reduction.  */
+  for (i = 0; i < n; i++)
+    lvresult = lresult || (result > array[i]);
+
+  if (lresult != lvresult)
+    abort ();
+
+  return 0;
+}
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/reduction-3.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/reduction-3.c
new file mode 100644
index 00000000000..58b49ffa4be
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/reduction-3.c
@@ -0,0 +1,126 @@
+/* { dg-do run } */
+
+/* double reductions.  */
+
+#include <stdlib.h>
+#include <stdbool.h>
+#include <math.h>
+
+#define vl 32
+
+int
+main(void)
+{
+  const int n = 1000;
+  int i;
+  double vresult, result, array[n];
+  bool lvresult, lresult;
+
+  for (i = 0; i < n; i++)
+    array[i] = i;
+
+  result = 0;
+  vresult = 0;
+
+  /* '+' reductions.  */
+#pragma acc parallel vector_length (vl)
+#pragma acc loop reduction (+:result)
+  for (i = 0; i < n; i++)
+    result += array[i];
+
+  /* Verify the reduction.  */
+  for (i = 0; i < n; i++)
+    vresult += array[i];
+
+  if (result != vresult)
+    abort ();
+
+  result = 0;
+  vresult = 0;
+
+  /* '*' reductions.  */
+#pragma acc parallel vector_length (vl)
+#pragma acc loop reduction (*:result)
+  for (i = 0; i < n; i++)
+    result *= array[i];
+
+  /* Verify the reduction.  */
+  for (i = 0; i < n; i++)
+    vresult *= array[i];
+
+  if (fabs(result - vresult) > .0001)
+    abort ();
+//   result = 0;
+//   vresult = 0;
+// 
+//   /* 'max' reductions.  */
+// #pragma acc parallel vector_length (vl)
+// #pragma acc loop reduction (+:result)
+//   for (i = 0; i < n; i++)
+//       result = result > array[i] ? result : array[i];
+// 
+//   /* Verify the reduction.  */
+//   for (i = 0; i < n; i++)
+//       vresult = vresult > array[i] ? vresult : array[i];
+// 
+//   printf("%d != %d\n", result, vresult);
+//   if (result != vresult)
+//     abort ();
+// 
+//   result = 0;
+//   vresult = 0;
+// 
+//   /* 'min' reductions.  */
+// #pragma acc parallel vector_length (vl)
+// #pragma acc loop reduction (+:result)
+//   for (i = 0; i < n; i++)
+//       result = result < array[i] ? result : array[i];
+// 
+//   /* Verify the reduction.  */
+//   for (i = 0; i < n; i++)
+//       vresult = vresult < array[i] ? vresult : array[i];
+// 
+//   printf("%d != %d\n", result, vresult);
+//   if (result != vresult)
+//     abort ();
+
+  result = 5;
+  vresult = 5;
+
+  lresult = false;
+  lvresult = false;
+
+  /* '&&' reductions.  */
+#pragma acc parallel vector_length (vl)
+#pragma acc loop reduction (&&:lresult)
+  for (i = 0; i < n; i++)
+    lresult = lresult && (result > array[i]);
+
+  /* Verify the reduction.  */
+  for (i = 0; i < n; i++)
+    lvresult = lresult && (result > array[i]);
+
+  if (lresult != lvresult)
+    abort ();
+
+  result = 5;
+  vresult = 5;
+
+  lresult = false;
+  lvresult = false;
+
+  /* '||' reductions.  */
+#pragma acc parallel vector_length (vl)
+#pragma acc loop reduction (||:lresult)
+  for (i = 0; i < n; i++)
+    lresult = lresult || (result > array[i]);
+
+  /* Verify the reduction.  */
+  for (i = 0; i < n; i++)
+    lvresult = lresult || (result > array[i]);
+
+  if (lresult != lvresult)
+    abort ();
+
+  return 0;
+}
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/reduction-4.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/reduction-4.c
new file mode 100644
index 00000000000..c8a9a6c0d36
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/reduction-4.c
@@ -0,0 +1,129 @@
+/* { dg-do run } */
+
+/* complex reductions.  */
+
+#include <stdlib.h>
+#include <stdbool.h>
+#include <math.h>
+#include <complex.h>
+
+#define vl 32
+
+int
+main(void)
+{
+  const int n = 1000;
+  int i;
+  double complex vresult, result, array[n];
+  bool lvresult, lresult;
+
+  for (i = 0; i < n; i++)
+    array[i] = i;
+
+  result = 0;
+  vresult = 0;
+
+  /* '+' reductions.  */
+#pragma acc parallel vector_length (vl)
+#pragma acc loop reduction (+:result)
+  for (i = 0; i < n; i++)
+    result += array[i];
+
+  /* Verify the reduction.  */
+  for (i = 0; i < n; i++)
+    vresult += array[i];
+
+  if (result != vresult)
+    abort ();
+
+  result = 0;
+  vresult = 0;
+
+  /* Needs support for complex multiplication.  */
+
+//   /* '*' reductions.  */
+// #pragma acc parallel vector_length (vl)
+// #pragma acc loop reduction (*:result)
+//   for (i = 0; i < n; i++)
+//     result *= array[i];
+// 
+//   /* Verify the reduction.  */
+//   for (i = 0; i < n; i++)
+//     vresult *= array[i];
+// 
+//   if (fabs(result - vresult) > .0001)
+//     abort ();
+//   result = 0;
+//   vresult = 0;
+
+//   /* 'max' reductions.  */
+// #pragma acc parallel vector_length (vl)
+// #pragma acc loop reduction (+:result)
+//   for (i = 0; i < n; i++)
+//       result = result > array[i] ? result : array[i];
+// 
+//   /* Verify the reduction.  */
+//   for (i = 0; i < n; i++)
+//       vresult = vresult > array[i] ? vresult : array[i];
+// 
+//   printf("%d != %d\n", result, vresult);
+//   if (result != vresult)
+//     abort ();
+// 
+//   result = 0;
+//   vresult = 0;
+// 
+//   /* 'min' reductions.  */
+// #pragma acc parallel vector_length (vl)
+// #pragma acc loop reduction (+:result)
+//   for (i = 0; i < n; i++)
+//       result = result < array[i] ? result : array[i];
+// 
+//   /* Verify the reduction.  */
+//   for (i = 0; i < n; i++)
+//       vresult = vresult < array[i] ? vresult : array[i];
+// 
+//   printf("%d != %d\n", result, vresult);
+//   if (result != vresult)
+//     abort ();
+
+  result = 5;
+  vresult = 5;
+
+  lresult = false;
+  lvresult = false;
+
+  /* '&&' reductions.  */
+#pragma acc parallel vector_length (vl)
+#pragma acc loop reduction (&&:lresult)
+  for (i = 0; i < n; i++)
+    lresult = lresult && (creal(result) > creal(array[i]));
+
+  /* Verify the reduction.  */
+  for (i = 0; i < n; i++)
+    lvresult = lresult && (creal(result) > creal(array[i]));
+
+  if (lresult != lvresult)
+    abort ();
+
+  result = 5;
+  vresult = 5;
+
+  lresult = false;
+  lvresult = false;
+
+  /* '||' reductions.  */
+#pragma acc parallel vector_length (vl)
+#pragma acc loop reduction (||:lresult)
+  for (i = 0; i < n; i++)
+    lresult = lresult || (creal(result) > creal(array[i]));
+
+  /* Verify the reduction.  */
+  for (i = 0; i < n; i++)
+    lvresult = lresult || (creal(result) > creal(array[i]));
+
+  if (lresult != lvresult)
+    abort ();
+
+  return 0;
+}
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/reduction-5.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/reduction-5.c
new file mode 100644
index 00000000000..757b8bec245
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/reduction-5.c
@@ -0,0 +1,32 @@
+#include <stdio.h>
+#include <stdlib.h>
+
+int
+main (void)
+{
+  int s1 = 2, s2 = 5, v1 = 2, v2 = 5;
+  int n = 100;
+  int i;
+
+#pragma acc parallel vector_length (1000)
+#pragma acc loop reduction (+:s1, s2)
+  for (i = 0; i < n; i++)
+    {
+      s1 = s1 + 3;
+      s2 = s2 + 2;
+    }
+
+  for (i = 0; i < n; i++)
+    {
+      v1 = v1 + 3;
+      v2 = v2 + 2;
+    }
+  
+  if (s1 != v1)
+    abort ();
+  
+  if (s2 != v2)
+    abort ();
+    
+  return 0;
+}
\ No newline at end of file
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/reduction-initial-1.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/reduction-initial-1.c
new file mode 100644
index 00000000000..81cf865e61b
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/reduction-initial-1.c
@@ -0,0 +1,25 @@
+/* { dg-do run } */
+
+int
+main(void)
+{
+#define I 5
+#define N 11
+#define A 8
+
+  int a = A;
+  int s = I;
+
+#pragma acc parallel vector_length(N)
+  {
+    int i;
+#pragma acc loop reduction(+:s)
+    for (i = 0; i < N; ++i)
+      s += a;
+  }
+
+  if (s != I + N * A)
+    __builtin_abort();
+
+  return 0;
+}
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/subr.h b/libgomp/testsuite/libgomp.oacc-c-c++-common/subr.h
new file mode 100644
index 00000000000..9db236c8362
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/subr.h
@@ -0,0 +1,46 @@
+
+#if ACC_DEVICE_TYPE_nvidia
+
+#pragma acc routine nohost
+static int clock (void)
+{
+  int thetime;
+
+  asm __volatile__ ("mov.u32 %0, %%clock;" : "=r"(thetime));
+
+  return thetime;
+}
+
+#endif
+
+void
+delay (unsigned long *d_o, unsigned long delay)
+{
+  int start, ticks;
+
+  start = clock ();
+
+  ticks = 0;
+
+  while (ticks < delay)
+    ticks = clock () - start;
+
+  return;
+}
+
+void
+delay2 (unsigned long *d_o, unsigned long delay, unsigned long tid)
+{
+  int start, ticks;
+
+  start = clock ();
+
+  ticks = 0;
+
+  while (ticks < delay)
+    ticks = clock () - start;
+
+  d_o[0] = tid;
+
+  return;
+}
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/subr.ptx b/libgomp/testsuite/libgomp.oacc-c-c++-common/subr.ptx
new file mode 100644
index 00000000000..6f748fcaf9e
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/subr.ptx
@@ -0,0 +1,148 @@
+// BEGIN PREAMBLE
+	.version	3.1
+	.target	sm_30
+	.address_size 64
+// END PREAMBLE
+
+// BEGIN FUNCTION DEF: clock
+.func (.param.u32 %out_retval)clock
+{
+.reg.u32 %retval;
+	.reg.u64 %hr10;
+	.reg.u32 %r22;
+	.reg.u32 %r23;
+	.reg.u32 %r24;
+	.local.align 8 .b8 %frame[8];
+	// #APP 
+// 7 "subr.c" 1
+	mov.u32 %r24, %clock;
+// 0 "" 2
+	// #NO_APP 
+		st.local.u32	[%frame], %r24;
+		ld.local.u32	%r22, [%frame];
+		mov.u32	%r23, %r22;
+		mov.u32	%retval, %r23;
+	st.param.u32	[%out_retval], %retval;
+	ret;
+	}
+// END FUNCTION DEF
+// BEGIN GLOBAL FUNCTION DEF: delay
+.visible .entry delay(.param.u64 %in_ar1, .param.u64 %in_ar2)
+{
+	.reg.u64 %ar1;
+	.reg.u64 %ar2;
+	.reg.u64 %hr10;
+	.reg.u64 %r22;
+	.reg.u32 %r23;
+	.reg.u64 %r24;
+	.reg.u64 %r25;
+	.reg.u32 %r26;
+	.reg.u32 %r27;
+	.reg.u32 %r28;
+	.reg.u32 %r29;
+	.reg.u32 %r30;
+	.reg.u64 %r31;
+	.reg.pred %r32;
+	.local.align 8 .b8 %frame[24];
+	ld.param.u64 %ar1, [%in_ar1];
+	ld.param.u64 %ar2, [%in_ar2];
+		mov.u64	%r24, %ar1;
+		st.u64	[%frame+8], %r24;
+		mov.u64	%r25, %ar2;
+		st.local.u64	[%frame+16], %r25;
+	{
+		.param.u32 %retval_in;
+	{
+		call (%retval_in), clock;
+	}
+		ld.param.u32	%r26, [%retval_in];
+}
+		st.local.u32	[%frame+4], %r26;
+		mov.u32	%r27, 0;
+		st.local.u32	[%frame], %r27;
+		bra	$L4;
+$L5:
+	{
+		.param.u32 %retval_in;
+	{
+		call (%retval_in), clock;
+	}
+		ld.param.u32	%r28, [%retval_in];
+}
+		mov.u32	%r23, %r28;
+		ld.local.u32	%r30, [%frame+4];
+		sub.u32	%r29, %r23, %r30;
+		st.local.u32	[%frame], %r29;
+$L4:
+		ld.local.s32	%r22, [%frame];
+		ld.local.u64	%r31, [%frame+16];
+		setp.lo.u64 %r32,%r22,%r31;
+	@%r32	bra	$L5;
+	ret;
+	}
+// END FUNCTION DEF
+// BEGIN GLOBAL FUNCTION DEF: delay2
+.visible .entry delay2(.param.u64 %in_ar1, .param.u64 %in_ar2, .param.u64 %in_ar3)
+{
+	.reg.u64 %ar1;
+	.reg.u64 %ar2;
+	.reg.u64 %ar3;
+	.reg.u64 %hr10;
+	.reg.u64 %r22;
+	.reg.u32 %r23;
+	.reg.u64 %r24;
+	.reg.u64 %r25;
+	.reg.u64 %r26;
+	.reg.u32 %r27;
+	.reg.u32 %r28;
+	.reg.u32 %r29;
+	.reg.u32 %r30;
+	.reg.u32 %r31;
+	.reg.u64 %r32;
+	.reg.pred %r33;
+	.reg.u64 %r34;
+	.reg.u64 %r35;
+	.local.align 8 .b8 %frame[32];
+	ld.param.u64 %ar1, [%in_ar1];
+	ld.param.u64 %ar2, [%in_ar2];
+	ld.param.u64 %ar3, [%in_ar3];
+		mov.u64	%r24, %ar1;
+		st.local.u64	[%frame+8], %r24;
+		mov.u64	%r25, %ar2;
+		st.local.u64	[%frame+16], %r25;
+		mov.u64	%r26, %ar3;
+		st.local.u64	[%frame+24], %r26;
+	{
+		.param.u32 %retval_in;
+	{
+		call (%retval_in), clock;
+	}
+		ld.param.u32	%r27, [%retval_in];
+}
+		st.local.u32	[%frame+4], %r27;
+		mov.u32	%r28, 0;
+		st.local.u32	[%frame], %r28;
+		bra	$L8;
+$L9:
+	{
+		.param.u32 %retval_in;
+	{
+		call (%retval_in), clock;
+	}
+		ld.param.u32	%r29, [%retval_in];
+}
+		mov.u32	%r23, %r29;
+		ld.local.u32	%r31, [%frame+4];
+		sub.u32	%r30, %r23, %r31;
+		st.local.u32	[%frame], %r30;
+$L8:
+		ld.local.s32	%r22, [%frame];
+		ld.local.u64	%r32, [%frame+16];
+		setp.lo.u64 %r33,%r22,%r32;
+	@%r33	bra	$L9;
+		ld.local.u64	%r34, [%frame+8];
+		ld.local.u64	%r35, [%frame+24];
+		st.u64	[%r34], %r35;
+	ret;
+	}
+// END FUNCTION DEF
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/timer.h b/libgomp/testsuite/libgomp.oacc-c-c++-common/timer.h
new file mode 100644
index 00000000000..53749da5a0d
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/timer.h
@@ -0,0 +1,103 @@
+
+#include <stdio.h>
+#include <cuda.h>
+
+static int _Tnum_timers;
+static CUevent *_Tstart_events, *_Tstop_events;
+static CUstream _Tstream;
+
+void
+init_timers (int ntimers)
+{
+  int i;
+  CUresult r;
+
+  _Tnum_timers = ntimers;
+
+  _Tstart_events = (CUevent *) malloc (_Tnum_timers * sizeof (CUevent));
+  _Tstop_events = (CUevent *) malloc (_Tnum_timers * sizeof (CUevent));
+
+  r = cuStreamCreate (&_Tstream, CU_STREAM_DEFAULT);
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuStreamCreate failed: %d\n", r);
+      abort ();
+    }
+
+  for (i = 0; i < _Tnum_timers; i++)
+    {
+      r = cuEventCreate (&_Tstart_events[i], CU_EVENT_DEFAULT);
+      if (r != CUDA_SUCCESS)
+	{
+	  fprintf (stderr, "cuEventCreate failed: %d\n", r);
+	  abort ();
+	}
+
+      r = cuEventCreate (&_Tstop_events[i], CU_EVENT_DEFAULT);
+      if (r != CUDA_SUCCESS)
+	{
+	  fprintf (stderr, "cuEventCreate failed: %d\n", r);
+	  abort ();
+	}
+    }
+}
+
+void
+fini_timers (void)
+{
+  int i;
+
+  for (i = 0; i < _Tnum_timers; i++)
+    {
+      cuEventDestroy (_Tstart_events[i]);
+      cuEventDestroy (_Tstop_events[i]);
+    }
+
+  cuStreamDestroy (_Tstream);
+
+  free (_Tstart_events);
+  free (_Tstop_events);
+}
+
+void
+start_timer (int timer)
+{
+  CUresult r;
+
+  r = cuEventRecord (_Tstart_events[timer], _Tstream);
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuEventRecord failed: %d\n", r);
+      abort ();
+    }
+}
+
+float
+stop_timer (int timer)
+{
+  CUresult r;
+  float etime;
+
+  r = cuEventRecord (_Tstop_events[timer], _Tstream);
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuEventRecord failed: %d\n", r);
+      abort ();
+    }
+
+  r = cuEventSynchronize (_Tstop_events[timer]);
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuEventSynchronize failed: %d\n", r);
+      abort ();
+    }
+
+  r = cuEventElapsedTime (&etime, _Tstart_events[timer], _Tstop_events[timer]);
+  if (r != CUDA_SUCCESS)
+    {
+      fprintf (stderr, "cuEventElapsedTime failed: %d\n", r);
+      abort ();
+    }
+
+  return etime;
+}
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/update-1-2.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/update-1-2.c
new file mode 100644
index 00000000000..c7e7257a873
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/update-1-2.c
@@ -0,0 +1,282 @@
+/* Copy of update-1.c with self exchanged with host for #pragma acc update.  */
+
+/* { dg-do run } */
+/* { dg-skip-if "" { *-*-* } { "*" } { "-DACC_MEM_SHARED=0" } } */
+
+#include <openacc.h>
+#include <string.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <stdbool.h>
+
+int
+main (int argc, char **argv)
+{
+    int N = 8;
+    float *a, *b, *c;
+    float *d_a, *d_b, *d_c;
+    int i;
+
+    a = (float *) malloc (N * sizeof (float));
+    b = (float *) malloc (N * sizeof (float));
+    c = (float *) malloc (N * sizeof (float));
+
+    d_a = (float *) acc_malloc (N * sizeof (float));
+    d_b = (float *) acc_malloc (N * sizeof (float));
+    d_c = (float *) acc_malloc (N * sizeof (float));
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 3.0;
+        b[i] = 0.0;
+    }
+
+    acc_map_data (a, d_a, N * sizeof (float));
+    acc_map_data (b, d_b, N * sizeof (float));
+    acc_map_data (c, d_c, N * sizeof (float));
+
+#pragma acc update device (a[0:N], b[0:N])
+
+#pragma acc parallel present (a[0:N], b[0:N])
+    {
+        int ii;
+
+        for (ii = 0; ii < N; ii++)
+            b[ii] = a[ii];
+    }
+
+#pragma acc update self (a[0:N], b[0:N])
+
+    for (i = 0; i < N; i++)
+    {
+        if (a[i] != 3.0)
+            abort ();
+
+        if (b[i] != 3.0)
+            abort ();
+    }
+
+    if (!acc_is_present (&a[0], (N * sizeof (float))))
+      abort ();
+
+    if (!acc_is_present (&b[0], (N * sizeof (float))))
+      abort ();
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 5.0;
+        b[i] = 1.0;
+    }
+
+#pragma acc update device (a[0:N], b[0:N])
+
+#pragma acc parallel present (a[0:N], b[0:N])
+    {
+        int ii;
+
+        for (ii = 0; ii < N; ii++)
+            b[ii] = a[ii];
+    }
+
+#pragma acc update self (a[0:N], b[0:N])
+
+    for (i = 0; i < N; i++)
+    {
+        if (a[i] != 5.0)
+            abort ();
+
+        if (b[i] != 5.0)
+            abort ();
+    }
+
+    if (!acc_is_present (&a[0], (N * sizeof (float))))
+      abort ();
+
+    if (!acc_is_present (&b[0], (N * sizeof (float))))
+      abort ();
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 5.0;
+        b[i] = 1.0;
+    }
+
+#pragma acc update device (a[0:N], b[0:N])
+
+#pragma acc parallel present (a[0:N], b[0:N])
+    {
+        int ii;
+
+        for (ii = 0; ii < N; ii++)
+            b[ii] = a[ii];
+    }
+
+#pragma acc update host (a[0:N], b[0:N])
+
+    for (i = 0; i < N; i++)
+    {
+        if (a[i] != 5.0)
+            abort ();
+
+        if (b[i] != 5.0)
+            abort ();
+    }
+
+    if (!acc_is_present (&a[0], (N * sizeof (float))))
+      abort ();
+
+    if (!acc_is_present (&b[0], (N * sizeof (float))))
+      abort ();
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 6.0;
+        b[i] = 0.0;
+    }
+
+#pragma acc update device (a[0:N], b[0:N])
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 9.0;
+    }
+
+#pragma acc parallel present (a[0:N], b[0:N])
+    {
+        int ii;
+
+        for (ii = 0; ii < N; ii++)
+            b[ii] = a[ii];
+    }
+
+#pragma acc update self (a[0:N], b[0:N])
+
+    for (i = 0; i < N; i++)
+    {
+        if (a[i] != 6.0)
+            abort ();
+
+        if (b[i] != 6.0)
+            abort ();
+    }
+
+    if (!acc_is_present (&a[0], (N * sizeof (float))))
+      abort ();
+
+    if (!acc_is_present (&b[0], (N * sizeof (float))))
+      abort ();
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 7.0;
+        b[i] = 2.0;
+    }
+
+#pragma acc update device (a[0:N], b[0:N])
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 9.0;
+    }
+
+#pragma acc parallel present (a[0:N], b[0:N])
+    {
+        int ii;
+
+        for (ii = 0; ii < N; ii++)
+            b[ii] = a[ii];
+    }
+
+#pragma acc update self (a[0:N], b[0:N])
+
+    for (i = 0; i < N; i++)
+    {
+        if (a[i] != 7.0)
+            abort ();
+
+        if (b[i] != 7.0)
+            abort ();
+    }
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 9.0;
+    }
+
+#pragma acc update device (a[0:N])
+
+#pragma acc parallel present (a[0:N], b[0:N])
+    {
+        int ii;
+
+        for (ii = 0; ii < N; ii++)
+            b[ii] = a[ii];
+    }
+
+#pragma acc update self (a[0:N], b[0:N])
+
+    for (i = 0; i < N; i++)
+    {
+        if (a[i] != 9.0)
+            abort ();
+
+        if (b[i] != 9.0)
+            abort ();
+    }
+
+    if (!acc_is_present (&a[0], (N * sizeof (float))))
+      abort ();
+
+    if (!acc_is_present (&b[0], (N * sizeof (float))))
+      abort ();
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 5.0;
+    }
+
+#pragma acc update device (a[0:N])
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 6.0;
+    }
+
+#pragma acc update device (a[0:N >> 1])
+
+#pragma acc parallel present (a[0:N], b[0:N])
+    {
+        int ii;
+
+        for (ii = 0; ii < N; ii++)
+            b[ii] = a[ii];
+    }
+
+#pragma acc update self (a[0:N], b[0:N])
+
+    for (i = 0; i < (N >> 1); i++)
+    {
+        if (a[i] != 6.0)
+            abort ();
+
+        if (b[i] != 6.0)
+            abort ();
+    }
+
+    for (i = (N >> 1); i < N; i++)
+    {
+        if (a[i] != 5.0)
+            abort ();
+
+        if (b[i] != 5.0)
+            abort ();
+    }
+
+    if (!acc_is_present (&a[0], (N * sizeof (float))))
+      abort ();
+
+    if (!acc_is_present (&b[0], (N * sizeof (float))))
+      abort ();
+
+    return 0;
+}
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/update-1.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/update-1.c
new file mode 100644
index 00000000000..dff139f03cc
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/update-1.c
@@ -0,0 +1,280 @@
+/* { dg-do run } */
+/* { dg-skip-if "" { *-*-* } { "*" } { "-DACC_MEM_SHARED=0" } } */
+
+#include <openacc.h>
+#include <string.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <stdbool.h>
+
+int
+main (int argc, char **argv)
+{
+    int N = 8;
+    float *a, *b, *c;
+    float *d_a, *d_b, *d_c;
+    int i;
+
+    a = (float *) malloc (N * sizeof (float));
+    b = (float *) malloc (N * sizeof (float));
+    c = (float *) malloc (N * sizeof (float));
+
+    d_a = (float *) acc_malloc (N * sizeof (float));
+    d_b = (float *) acc_malloc (N * sizeof (float));
+    d_c = (float *) acc_malloc (N * sizeof (float));
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 3.0;
+        b[i] = 0.0;
+    }
+
+    acc_map_data (a, d_a, N * sizeof (float));
+    acc_map_data (b, d_b, N * sizeof (float));
+    acc_map_data (c, d_c, N * sizeof (float));
+
+#pragma acc update device (a[0:N], b[0:N])
+
+#pragma acc parallel present (a[0:N], b[0:N])
+    {
+        int ii;
+
+        for (ii = 0; ii < N; ii++)
+            b[ii] = a[ii];
+    }
+
+#pragma acc update host (a[0:N], b[0:N])
+
+    for (i = 0; i < N; i++)
+    {
+        if (a[i] != 3.0)
+            abort ();
+
+        if (b[i] != 3.0)
+            abort ();
+    }
+
+    if (!acc_is_present (&a[0], (N * sizeof (float))))
+      abort ();
+
+    if (!acc_is_present (&b[0], (N * sizeof (float))))
+      abort ();
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 5.0;
+        b[i] = 1.0;
+    }
+
+#pragma acc update device (a[0:N], b[0:N])
+
+#pragma acc parallel present (a[0:N], b[0:N])
+    {
+        int ii;
+
+        for (ii = 0; ii < N; ii++)
+            b[ii] = a[ii];
+    }
+
+#pragma acc update host (a[0:N], b[0:N])
+
+    for (i = 0; i < N; i++)
+    {
+        if (a[i] != 5.0)
+            abort ();
+
+        if (b[i] != 5.0)
+            abort ();
+    }
+
+    if (!acc_is_present (&a[0], (N * sizeof (float))))
+      abort ();
+
+    if (!acc_is_present (&b[0], (N * sizeof (float))))
+      abort ();
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 5.0;
+        b[i] = 1.0;
+    }
+
+#pragma acc update device (a[0:N], b[0:N])
+
+#pragma acc parallel present (a[0:N], b[0:N])
+    {
+        int ii;
+
+        for (ii = 0; ii < N; ii++)
+            b[ii] = a[ii];
+    }
+
+#pragma acc update self (a[0:N], b[0:N])
+
+    for (i = 0; i < N; i++)
+    {
+        if (a[i] != 5.0)
+            abort ();
+
+        if (b[i] != 5.0)
+            abort ();
+    }
+
+    if (!acc_is_present (&a[0], (N * sizeof (float))))
+      abort ();
+
+    if (!acc_is_present (&b[0], (N * sizeof (float))))
+      abort ();
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 6.0;
+        b[i] = 0.0;
+    }
+
+#pragma acc update device (a[0:N], b[0:N])
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 9.0;
+    }
+
+#pragma acc parallel present (a[0:N], b[0:N])
+    {
+        int ii;
+
+        for (ii = 0; ii < N; ii++)
+            b[ii] = a[ii];
+    }
+
+#pragma acc update host (a[0:N], b[0:N])
+
+    for (i = 0; i < N; i++)
+    {
+        if (a[i] != 6.0)
+            abort ();
+
+        if (b[i] != 6.0)
+            abort ();
+    }
+
+    if (!acc_is_present (&a[0], (N * sizeof (float))))
+      abort ();
+
+    if (!acc_is_present (&b[0], (N * sizeof (float))))
+      abort ();
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 7.0;
+        b[i] = 2.0;
+    }
+
+#pragma acc update device (a[0:N], b[0:N])
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 9.0;
+    }
+
+#pragma acc parallel present (a[0:N], b[0:N])
+    {
+        int ii;
+
+        for (ii = 0; ii < N; ii++)
+            b[ii] = a[ii];
+    }
+
+#pragma acc update host (a[0:N], b[0:N])
+
+    for (i = 0; i < N; i++)
+    {
+        if (a[i] != 7.0)
+            abort ();
+
+        if (b[i] != 7.0)
+            abort ();
+    }
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 9.0;
+    }
+
+#pragma acc update device (a[0:N])
+
+#pragma acc parallel present (a[0:N], b[0:N])
+    {
+        int ii;
+
+        for (ii = 0; ii < N; ii++)
+            b[ii] = a[ii];
+    }
+
+#pragma acc update host (a[0:N], b[0:N])
+
+    for (i = 0; i < N; i++)
+    {
+        if (a[i] != 9.0)
+            abort ();
+
+        if (b[i] != 9.0)
+            abort ();
+    }
+
+    if (!acc_is_present (&a[0], (N * sizeof (float))))
+      abort ();
+
+    if (!acc_is_present (&b[0], (N * sizeof (float))))
+      abort ();
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 5.0;
+    }
+
+#pragma acc update device (a[0:N])
+
+    for (i = 0; i < N; i++)
+    {
+        a[i] = 6.0;
+    }
+
+#pragma acc update device (a[0:N >> 1])
+
+#pragma acc parallel present (a[0:N], b[0:N])
+    {
+        int ii;
+
+        for (ii = 0; ii < N; ii++)
+            b[ii] = a[ii];
+    }
+
+#pragma acc update host (a[0:N], b[0:N])
+
+    for (i = 0; i < (N >> 1); i++)
+    {
+        if (a[i] != 6.0)
+            abort ();
+
+        if (b[i] != 6.0)
+            abort ();
+    }
+
+    for (i = (N >> 1); i < N; i++)
+    {
+        if (a[i] != 5.0)
+            abort ();
+
+        if (b[i] != 5.0)
+            abort ();
+    }
+
+    if (!acc_is_present (&a[0], (N * sizeof (float))))
+      abort ();
+
+    if (!acc_is_present (&b[0], (N * sizeof (float))))
+      abort ();
+
+    return 0;
+}
diff --git a/libgomp/testsuite/libgomp.oacc-c/c.exp b/libgomp/testsuite/libgomp.oacc-c/c.exp
new file mode 100644
index 00000000000..c0c70bbacad
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c/c.exp
@@ -0,0 +1,71 @@
+# This whole file adapted from libgomp.c/c.exp.
+
+if [info exists lang_library_path] then {
+    unset lang_library_path
+    unset lang_link_flags
+}
+if [info exists lang_test_file] then {
+    unset lang_test_file
+}
+if [info exists lang_include_flags] then {
+    unset lang_include_flags
+}
+
+load_lib libgomp-dg.exp
+load_gcc_lib gcc-dg.exp
+
+# If a testcase doesn't have special options, use these.
+if ![info exists DEFAULT_CFLAGS] then {
+    set DEFAULT_CFLAGS "-O2"
+}
+
+# Initialize dg.
+dg-init
+
+# Turn on OpenACC.
+lappend ALWAYS_CFLAGS "additional_flags=-fopenacc"
+
+# Gather a list of all tests.
+set tests [lsort [concat \
+		      [find $srcdir/$subdir *.c] \
+		      [find $srcdir/$subdir/../libgomp.oacc-c-c++-common *.c]]]
+
+set ld_library_path $always_ld_library_path
+append ld_library_path [gcc-set-multilib-library-path $GCC_UNDER_TEST]
+set_ld_library_path_env_vars
+
+# Test OpenACC with available accelerators.
+set SAVE_ALWAYS_CFLAGS "$ALWAYS_CFLAGS"
+foreach offload_target_openacc $offload_targets_s_openacc {
+    set ALWAYS_CFLAGS "$SAVE_ALWAYS_CFLAGS"
+    set tagopt "-DACC_DEVICE_TYPE_$offload_target_openacc=1"
+
+    switch $offload_target_openacc {
+	host {
+	    set acc_mem_shared 1
+	}
+	host_nonshm {
+	    set acc_mem_shared 0
+	}
+	nvidia {
+	    # Copy ptx file (TEMPORARY)
+	    remote_download host $srcdir/libgomp.oacc-c-c++-common/subr.ptx
+
+	    # Where timer.h lives
+	    lappend ALWAYS_CFLAGS "additional_flags=-I${srcdir}/libgomp.oacc-c-c++-common"
+
+	    set acc_mem_shared 0
+	}
+	default {
+	    set acc_mem_shared 0
+	}
+    }
+    set tagopt "$tagopt -DACC_MEM_SHARED=$acc_mem_shared"
+
+    setenv ACC_DEVICE_TYPE $offload_target_openacc
+
+    dg-runtest $tests "$tagopt" $DEFAULT_CFLAGS
+}
+
+# All done.
+dg-finish
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/abort-1.f90 b/libgomp/testsuite/libgomp.oacc-fortran/abort-1.f90
new file mode 100644
index 00000000000..52b030bb6ce
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/abort-1.f90
@@ -0,0 +1,10 @@
+! { dg-shouldfail "" { *-*-* } { "*" } { "" } }
+
+program main
+  implicit none
+
+  !$acc parallel
+  call abort
+  !$acc end parallel
+
+end program main
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/abort-2.f90 b/libgomp/testsuite/libgomp.oacc-fortran/abort-2.f90
new file mode 100644
index 00000000000..2ba2bcb8aa6
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/abort-2.f90
@@ -0,0 +1,13 @@
+program main
+  implicit none
+
+  integer :: argc
+  argc = command_argument_count ()
+
+  !$acc parallel copyin(argc)
+  if (argc .ne. 0) then
+     call abort
+  end if
+  !$acc end parallel
+
+end program main
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/acc_on_device-1-1.f90 b/libgomp/testsuite/libgomp.oacc-fortran/acc_on_device-1-1.f90
new file mode 100644
index 00000000000..448881837ac
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/acc_on_device-1-1.f90
@@ -0,0 +1,52 @@
+! { dg-additional-options "-cpp" }
+! TODO: Have to disable the acc_on_device builtin for we want to test the
+! libgomp library function?  The command line option
+! '-fno-builtin-acc_on_device' is valid for C/C++/ObjC/ObjC++ but not for
+! Fortran.
+
+use openacc
+implicit none
+
+! Host.
+
+if (.not. acc_on_device (acc_device_none)) call abort
+if (.not. acc_on_device (acc_device_host)) call abort
+if (acc_on_device (acc_device_host_nonshm)) call abort
+if (acc_on_device (acc_device_not_host)) call abort
+if (acc_on_device (acc_device_nvidia)) call abort
+
+
+! Host via offloading fallback mode.
+
+!$acc parallel if(.false.)
+if (.not. acc_on_device (acc_device_none)) call abort
+if (.not. acc_on_device (acc_device_host)) call abort
+if (acc_on_device (acc_device_host_nonshm)) call abort
+if (acc_on_device (acc_device_not_host)) call abort
+if (acc_on_device (acc_device_nvidia)) call abort
+!$acc end parallel
+
+
+#if !ACC_DEVICE_TYPE_host
+
+! Offloaded.
+
+!$acc parallel
+if (acc_on_device (acc_device_none)) call abort
+if (acc_on_device (acc_device_host)) call abort
+#if ACC_DEVICE_TYPE_host_nonshm
+if (.not. acc_on_device (acc_device_host_nonshm)) call abort
+#else
+if (acc_on_device (acc_device_host_nonshm)) call abort
+#endif
+if (.not. acc_on_device (acc_device_not_host)) call abort
+#if ACC_DEVICE_TYPE_nvidia
+if (.not. acc_on_device (acc_device_nvidia)) call abort
+#else
+if (acc_on_device (acc_device_nvidia)) call abort
+#endif
+!$acc end parallel
+
+#endif
+
+end
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/acc_on_device-1-2.f b/libgomp/testsuite/libgomp.oacc-fortran/acc_on_device-1-2.f
new file mode 100644
index 00000000000..0047a194f66
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/acc_on_device-1-2.f
@@ -0,0 +1,52 @@
+! { dg-additional-options "-cpp" }
+! TODO: Have to disable the acc_on_device builtin for we want to test
+! the libgomp library function?  The command line option
+! '-fno-builtin-acc_on_device' is valid for C/C++/ObjC/ObjC++ but not
+! for Fortran.
+
+      USE OPENACC
+      IMPLICIT NONE
+
+!Host.
+
+      IF (.NOT. ACC_ON_DEVICE (ACC_DEVICE_NONE)) CALL ABORT
+      IF (.NOT. ACC_ON_DEVICE (ACC_DEVICE_HOST)) CALL ABORT
+      IF (ACC_ON_DEVICE (ACC_DEVICE_HOST_NONSHM)) CALL ABORT
+      IF (ACC_ON_DEVICE (ACC_DEVICE_NOT_HOST)) CALL ABORT
+      IF (ACC_ON_DEVICE (ACC_DEVICE_NVIDIA)) CALL ABORT
+
+
+!Host via offloading fallback mode.
+
+!$ACC PARALLEL IF(.FALSE.)
+      IF (.NOT. ACC_ON_DEVICE (ACC_DEVICE_NONE)) CALL ABORT
+      IF (.NOT. ACC_ON_DEVICE (ACC_DEVICE_HOST)) CALL ABORT
+      IF (ACC_ON_DEVICE (ACC_DEVICE_HOST_NONSHM)) CALL ABORT
+      IF (ACC_ON_DEVICE (ACC_DEVICE_NOT_HOST)) CALL ABORT
+      IF (ACC_ON_DEVICE (ACC_DEVICE_NVIDIA)) CALL ABORT
+!$ACC END PARALLEL
+
+
+#if !ACC_DEVICE_TYPE_host
+
+! Offloaded.
+
+!$ACC PARALLEL
+      IF (ACC_ON_DEVICE (ACC_DEVICE_NONE)) CALL ABORT
+      IF (ACC_ON_DEVICE (ACC_DEVICE_HOST)) CALL ABORT
+#if ACC_DEVICE_TYPE_host_nonshm
+      IF (.NOT. ACC_ON_DEVICE (ACC_DEVICE_HOST_NONSHM)) CALL ABORT
+#else
+      IF (ACC_ON_DEVICE (ACC_DEVICE_HOST_NONSHM)) CALL ABORT
+#endif
+      IF (.NOT. ACC_ON_DEVICE (ACC_DEVICE_NOT_HOST)) CALL ABORT
+#if ACC_DEVICE_TYPE_nvidia
+      IF (.NOT. ACC_ON_DEVICE (ACC_DEVICE_NVIDIA)) CALL ABORT
+#else
+      IF (ACC_ON_DEVICE (ACC_DEVICE_NVIDIA)) CALL ABORT
+#endif
+!$ACC END PARALLEL
+
+#endif
+
+      END
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/acc_on_device-1-3.f b/libgomp/testsuite/libgomp.oacc-fortran/acc_on_device-1-3.f
new file mode 100644
index 00000000000..49d7a720fe3
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/acc_on_device-1-3.f
@@ -0,0 +1,52 @@
+! { dg-additional-options "-cpp" }
+! TODO: Have to disable the acc_on_device builtin for we want to test
+! the libgomp library function?  The command line option
+! '-fno-builtin-acc_on_device' is valid for C/C++/ObjC/ObjC++ but not
+! for Fortran.
+
+      IMPLICIT NONE
+      INCLUDE "openacc_lib.h"
+
+!Host.
+
+      IF (.NOT. ACC_ON_DEVICE (ACC_DEVICE_NONE)) CALL ABORT
+      IF (.NOT. ACC_ON_DEVICE (ACC_DEVICE_HOST)) CALL ABORT
+      IF (ACC_ON_DEVICE (ACC_DEVICE_HOST_NONSHM)) CALL ABORT
+      IF (ACC_ON_DEVICE (ACC_DEVICE_NOT_HOST)) CALL ABORT
+      IF (ACC_ON_DEVICE (ACC_DEVICE_NVIDIA)) CALL ABORT
+
+
+!Host via offloading fallback mode.
+
+!$ACC PARALLEL IF(.FALSE.)
+      IF (.NOT. ACC_ON_DEVICE (ACC_DEVICE_NONE)) CALL ABORT
+      IF (.NOT. ACC_ON_DEVICE (ACC_DEVICE_HOST)) CALL ABORT
+      IF (ACC_ON_DEVICE (ACC_DEVICE_HOST_NONSHM)) CALL ABORT
+      IF (ACC_ON_DEVICE (ACC_DEVICE_NOT_HOST)) CALL ABORT
+      IF (ACC_ON_DEVICE (ACC_DEVICE_NVIDIA)) CALL ABORT
+!$ACC END PARALLEL
+
+
+#if !ACC_DEVICE_TYPE_host
+
+! Offloaded.
+
+!$ACC PARALLEL
+      IF (ACC_ON_DEVICE (ACC_DEVICE_NONE)) CALL ABORT
+      IF (ACC_ON_DEVICE (ACC_DEVICE_HOST)) CALL ABORT
+#if ACC_DEVICE_TYPE_host_nonshm
+      IF (.NOT. ACC_ON_DEVICE (ACC_DEVICE_HOST_NONSHM)) CALL ABORT
+#else
+      IF (ACC_ON_DEVICE (ACC_DEVICE_HOST_NONSHM)) CALL ABORT
+#endif
+      IF (.NOT. ACC_ON_DEVICE (ACC_DEVICE_NOT_HOST)) CALL ABORT
+#if ACC_DEVICE_TYPE_nvidia
+      IF (.NOT. ACC_ON_DEVICE (ACC_DEVICE_NVIDIA)) CALL ABORT
+#else
+      IF (ACC_ON_DEVICE (ACC_DEVICE_NVIDIA)) CALL ABORT
+#endif
+!$ACC END PARALLEL
+
+#endif
+
+      END
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/asyncwait-1.f90 b/libgomp/testsuite/libgomp.oacc-fortran/asyncwait-1.f90
new file mode 100644
index 00000000000..b6e637ba43f
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/asyncwait-1.f90
@@ -0,0 +1,135 @@
+! { dg-do run }
+
+program asyncwait
+  integer, parameter :: N = 64
+  real, allocatable :: a(:), b(:), c(:), d(:), e(:)
+  integer i
+
+  allocate (a(N))
+  allocate (b(N))
+  allocate (c(N))
+  allocate (d(N))
+  allocate (e(N))
+
+  a(:) = 3.0
+  b(:) = 0.0
+
+  !$acc data copy (a(1:N)) copy (b(1:N))
+
+  !$acc parallel async
+  !$acc loop
+  do i = 1, N
+     b(i) = a(i)
+  end do
+  !$acc end parallel
+
+  !$acc wait
+  !$acc end data
+
+  do i = 1, N
+     if (a(i) .ne. 3.0) call abort
+     if (b(i) .ne. 3.0) call abort
+  end do
+
+  a(:) = 2.0
+  b(:) = 0.0
+
+  !$acc data copy (a(1:N)) copy (b(1:N))
+
+  !$acc parallel async (1)
+  !$acc loop
+  do i = 1, N
+     b(i) = a(i)
+  end do
+  !$acc end parallel
+
+  !$acc wait (1)
+  !$acc end data
+
+  do i = 1, N
+     if (a(i) .ne. 2.0) call abort
+     if (b(i) .ne. 2.0) call abort
+  end do
+
+  a(:) = 3.0
+  b(:) = 0.0
+  c(:) = 0.0
+  d(:) = 0.0
+
+  !$acc data copy (a(1:N)) copy (b(1:N)) copy (c(1:N)) copy (d(1:N))
+
+  !$acc parallel async (1)
+  do i = 1, N
+     b(i) = (a(i) * a(i) * a(i)) / a(i)
+  end do
+  !$acc end parallel
+
+  !$acc parallel async (1)
+  do i = 1, N
+     c(i) = (a(i) * 4) / a(i)
+  end do
+  !$acc end parallel
+
+  !$acc parallel async (1)
+  !$acc loop
+  do i = 1, N
+     d(i) = ((a(i) * a(i) + a(i)) / a(i)) - a(i)
+  end do
+  !$acc end parallel
+
+  !$acc wait (1)
+  !$acc end data
+
+  do i = 1, N
+     if (a(i) .ne. 3.0) call abort
+     if (b(i) .ne. 9.0) call abort
+     if (c(i) .ne. 4.0) call abort
+     if (d(i) .ne. 1.0) call abort
+  end do
+
+  a(:) = 2.0
+  b(:) = 0.0
+  c(:) = 0.0
+  d(:) = 0.0
+  e(:) = 0.0
+
+  !$acc data copy (a(1:N), b(1:N), c(1:N), d(1:N), e(1:N))
+
+  !$acc parallel async (1)
+  do i = 1, N
+     b(i) = (a(i) * a(i) * a(i)) / a(i)
+  end do
+  !$acc end parallel
+
+  !$acc parallel async (1)
+  !$acc loop
+  do i = 1, N
+     c(i) = (a(i) * 4) / a(i)
+  end do
+  !$acc end parallel
+
+  !$acc parallel async (1)
+  !$acc loop
+  do i = 1, N
+     d(i) = ((a(i) * a(i) + a(i)) / a(i)) - a(i)
+  end do
+  !$acc end parallel
+
+  !$acc parallel wait (1) async (1)
+  !$acc loop
+  do i = 1, N
+     e(i) = a(i) + b(i) + c(i) + d(i)
+  end do
+  !$acc end parallel
+
+  !$acc wait (1)
+  !$acc end data
+
+  do i = 1, N
+     if (a(i) .ne. 2.0) call abort
+     if (b(i) .ne. 4.0) call abort
+     if (c(i) .ne. 4.0) call abort
+     if (d(i) .ne. 1.0) call abort
+     if (e(i) .ne. 11.0) call abort
+  end do
+end program asyncwait
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/asyncwait-2.f90 b/libgomp/testsuite/libgomp.oacc-fortran/asyncwait-2.f90
new file mode 100644
index 00000000000..bade52bdd7d
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/asyncwait-2.f90
@@ -0,0 +1,40 @@
+! { dg-do run }
+
+program parallel_wait
+  integer, parameter :: N = 64
+  real, allocatable :: a(:), b(:), c(:)
+  integer i
+
+  allocate (a(N))
+  allocate (b(N))
+  allocate (c(N))
+
+  !$acc parallel async (0)
+  !$acc loop
+  do i = 1, N
+    a(i) = 1
+  end do
+  !$acc end parallel
+
+  !$acc parallel async (1)
+  !$acc loop
+  do i = 1, N
+    b(i) = 1
+  end do
+  !$acc end parallel
+
+  !$acc parallel wait (0, 1)
+  !$acc loop
+  do i = 1, N
+    c(i) = a(i) + b(i)
+  end do
+  !$acc end parallel
+
+  do i = 1, N
+    if (c(i) .ne. 2.0) call abort
+  end do
+  
+  deallocate (a)
+  deallocate (b)
+  deallocate (c)
+end program parallel_wait
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/asyncwait-3.f90 b/libgomp/testsuite/libgomp.oacc-fortran/asyncwait-3.f90
new file mode 100644
index 00000000000..d48dc11bfd6
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/asyncwait-3.f90
@@ -0,0 +1,42 @@
+! { dg-do run }
+
+program parallel_wait
+  integer, parameter :: N = 64
+  real, allocatable :: a(:), b(:), c(:)
+  integer i
+
+  allocate (a(N))
+  allocate (b(N))
+  allocate (c(N))
+
+  !$acc parallel async (0)
+  !$acc loop
+  do i = 1, N
+    a(i) = 1
+  end do
+  !$acc end parallel
+
+  !$acc parallel async (1)
+  !$acc loop
+  do i = 1, N
+    b(i) = 1
+  end do
+  !$acc end parallel
+
+  !$acc wait (0, 1)
+
+  !$acc parallel
+  !$acc loop
+  do i = 1, N
+    c(i) = a(i) + b(i)
+  end do
+  !$acc end parallel
+
+  do i = 1, N
+    if (c(i) .ne. 2.0) call abort
+  end do
+  
+  deallocate (a)
+  deallocate (b)
+  deallocate (c)
+end program parallel_wait
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/collapse-1.f90 b/libgomp/testsuite/libgomp.oacc-fortran/collapse-1.f90
new file mode 100644
index 00000000000..4c07bc2108e
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/collapse-1.f90
@@ -0,0 +1,27 @@
+! { dg-do run }
+
+program collapse1
+  integer :: i, j, k, a(1:3, 4:6, 5:7)
+  logical :: l
+  l = .false.
+  a(:, :, :) = 0
+  !$acc parallel
+  !$acc loop collapse(4 - 1)
+    do i = 1, 3
+      do j = 4, 6
+        do k = 5, 7
+          a(i, j, k) = i + j + k
+        end do
+      end do
+    end do
+  !$acc loop collapse(2) reduction(.or.:l)
+    do i = 1, 3
+      do j = 4, 6
+        do k = 5, 7
+          if (a(i, j, k) .ne. (i + j + k)) l = .true.
+        end do
+      end do
+    end do
+  !$acc end parallel
+  if (l) call abort
+end program collapse1
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/collapse-2.f90 b/libgomp/testsuite/libgomp.oacc-fortran/collapse-2.f90
new file mode 100644
index 00000000000..ca3b638a461
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/collapse-2.f90
@@ -0,0 +1,25 @@
+! { dg-do run }
+
+program collapse2
+  integer :: i, j, k, a(1:3, 4:6, 5:7)
+  logical :: l
+  l = .false.
+  a(:, :, :) = 0
+  !$acc parallel
+  !$acc loop collapse(4 - 1)
+    do 164 i = 1, 3
+      do 164 j = 4, 6
+        do 164 k = 5, 7
+          a(i, j, k) = i + j + k
+164      end do
+  !$acc loop collapse(2) reduction(.or.:l)
+firstdo: do i = 1, 3
+      do j = 4, 6
+        do k = 5, 7
+          if (a(i, j, k) .ne. (i + j + k)) l = .true.
+        end do
+      end do
+    end do firstdo
+  !$acc end parallel
+  if (l) call abort
+end program collapse2
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/collapse-3.f90 b/libgomp/testsuite/libgomp.oacc-fortran/collapse-3.f90
new file mode 100644
index 00000000000..50e6100a1a0
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/collapse-3.f90
@@ -0,0 +1,28 @@
+! { dg-do run }
+
+program collapse3
+  integer :: a(3,3,3), k, kk, kkk, l, ll, lll
+  !$acc parallel
+  !$acc loop collapse(3)
+    do 115 k=1,3
+dokk: do kk=1,3
+        do kkk=1,3
+          a(k,kk,kkk) = 1
+        enddo
+      enddo dokk
+115   continue
+  !$acc end parallel
+  if (any(a(1:3,1:3,1:3).ne.1)) call abort
+
+  !$acc parallel
+  !$acc loop collapse(3)
+dol: do 120 l=1,3
+doll: do ll=1,3
+        do lll=1,3
+          a(l,ll,lll) = 2
+        enddo
+      enddo doll
+120 end do dol
+  !$acc end parallel
+  if (any(a(1:3,1:3,1:3).ne.2)) call abort
+end program collapse3
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/collapse-4.f90 b/libgomp/testsuite/libgomp.oacc-fortran/collapse-4.f90
new file mode 100644
index 00000000000..41b66db43ec
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/collapse-4.f90
@@ -0,0 +1,40 @@
+! { dg-do run }
+
+! collapse3.f90:test1
+program collapse4
+  integer :: i, j, k, a(1:7, -3:5, 12:19), b(1:7, -3:5, 12:19)
+  logical :: l, r
+  l = .false.
+  r = .false.
+  a(:, :, :) = 0
+  b(:, :, :) = 0
+  !$acc parallel
+  !$acc loop collapse (3) reduction (.or.:l)
+    do i = 2, 6
+      do j = -2, 4
+        do k = 13, 18
+          l = l.or.i.lt.2.or.i.gt.6.or.j.lt.-2.or.j.gt.4
+          l = l.or.k.lt.13.or.k.gt.18
+          if (.not.l) a(i, j, k) = a(i, j, k) + 1
+        end do
+      end do
+    end do
+  !$acc end parallel
+  do i = 2, 6
+    do j = -2, 4
+      do k = 13, 18
+        r = r.or.i.lt.2.or.i.gt.6.or.j.lt.-2.or.j.gt.4
+        r = r.or.k.lt.13.or.k.gt.18
+        if (.not.l) b(i, j, k) = b(i, j, k) + 1
+      end do
+    end do
+  end do
+  if (l .neqv. r) call abort
+  do i = 2, 6
+    do j = -2, 4
+      do k = 13, 18
+         if (a(i, j, k) .ne. b(i, j, k)) call abort
+      end do
+    end do
+  end do
+end program collapse4
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/collapse-5.f90 b/libgomp/testsuite/libgomp.oacc-fortran/collapse-5.f90
new file mode 100644
index 00000000000..8c20f043f4c
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/collapse-5.f90
@@ -0,0 +1,48 @@
+! { dg-do run }
+
+! collapse3.f90:test2
+program collapse5
+  integer :: i, j, k, a(1:7, -3:5, 12:19), b(1:7, -3:5, 12:19)
+  integer :: v1, v2, v3, v4, v5, v6
+  logical :: l, r
+  l = .false.
+  r = .false.
+  a(:, :, :) = 0
+  b(:, :, :) = 0
+  v1 = 3
+  v2 = 6
+  v3 = -2
+  v4 = 4
+  v5 = 13
+  v6 = 18
+  !$acc parallel
+  !$acc loop collapse (3) reduction (.or.:l)
+    do i = v1, v2
+      do j = v3, v4
+        do k = v5, v6
+          l = l.or.i.lt.2.or.i.gt.6.or.j.lt.-2.or.j.gt.4
+          l = l.or.k.lt.13.or.k.gt.18
+          if (.not.l) a(i, j, k) = a(i, j, k) + 1
+          m = i * 100 + j * 10 + k
+        end do
+      end do
+    end do
+  !$acc end parallel
+  do i = v1, v2
+    do j = v3, v4
+      do k = v5, v6
+        r = r.or.i.lt.2.or.i.gt.6.or.j.lt.-2.or.j.gt.4
+        r = r.or.k.lt.13.or.k.gt.18
+        if (.not.l) b(i, j, k) = b(i, j, k) + 1
+      end do
+    end do
+  end do
+  if (l .neqv. r) call abort
+  do i = v1, v2
+    do j = v3, v4
+      do k = v5, v6
+         if (a(i, j, k) .ne. b(i, j, k)) call abort
+      end do
+    end do
+  end do
+end program collapse5
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/collapse-6.f90 b/libgomp/testsuite/libgomp.oacc-fortran/collapse-6.f90
new file mode 100644
index 00000000000..7404b9182ac
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/collapse-6.f90
@@ -0,0 +1,50 @@
+! { dg-do run }
+
+! collapse3.f90:test3
+program collapse6
+  integer :: i, j, k, a(1:7, -3:5, 12:19), b(1:7, -3:5, 12:19)
+  integer :: v1, v2, v3, v4, v5, v6, v7, v8, v9
+  logical :: l, r
+  l = .false.
+  r = .false.
+  a(:, :, :) = 0
+  b(:, :, :) = 0
+  v1 = 3
+  v2 = 6
+  v3 = -2
+  v4 = 4
+  v5 = 13
+  v6 = 18
+  v7 = 1
+  v8 = 1
+  v9 = 1
+  !$acc parallel
+  !$acc loop collapse (3) reduction (.or.:l)
+    do i = v1, v2, v7
+      do j = v3, v4, v8
+        do k = v5, v6, v9
+          l = l.or.i.lt.2.or.i.gt.6.or.j.lt.-2.or.j.gt.4
+          l = l.or.k.lt.13.or.k.gt.18
+          if (.not.l) a(i, j, k) = a(i, j, k) + 1
+        end do
+      end do
+    end do
+  !$acc end parallel
+  do i = v1, v2, v7
+    do j = v3, v4, v8
+      do k = v5, v6, v9
+        r = r.or.i.lt.2.or.i.gt.6.or.j.lt.-2.or.j.gt.4
+        r = r.or.k.lt.13.or.k.gt.18
+        if (.not.r) b(i, j, k) = b(i, j, k) + 1
+      end do
+    end do
+  end do
+  if (l .neqv. r) call abort
+  do i = v1, v2, v7
+    do j = v3, v4, v8
+       do k = v5, v6, v9
+         if (a(i, j, k) .ne. b(i, j, k)) call abort
+       end do
+    end do
+  end do
+end program collapse6
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/collapse-7.f90 b/libgomp/testsuite/libgomp.oacc-fortran/collapse-7.f90
new file mode 100644
index 00000000000..12efd8c0ab4
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/collapse-7.f90
@@ -0,0 +1,40 @@
+! { dg-do run }
+
+! collapse3.f90:test4
+program collapse7
+  integer :: i, j, k, a(1:7, -3:5, 12:19), b(1:7, -3:5, 12:19)
+  logical :: l, r
+  l = .false.
+  r = .false.
+  a(:, :, :) = 0
+  b(:, :, :) = 0
+  !$acc parallel
+  !$acc loop collapse (3) reduction (.or.:l)
+    do i = 2, 6
+      do j = -2, 4
+        do k = 13, 18
+          l = l.or.i.lt.2.or.i.gt.6.or.j.lt.-2.or.j.gt.4
+          l = l.or.k.lt.13.or.k.gt.18
+          if (.not.l) a(i, j, k) = a(i, j, k) + 1
+        end do
+      end do
+    end do
+  !$acc end parallel
+  do i = 2, 6
+    do j = -2, 4
+      do k = 13, 18
+        r = r.or.i.lt.2.or.i.gt.6.or.j.lt.-2.or.j.gt.4
+        r = r.or.k.lt.13.or.k.gt.18
+        if (.not.r) b(i, j, k) = b(i, j, k) + 1
+      end do
+    end do
+  end do
+  if (l .neqv. r) call abort
+  do i = 1, 7
+    do j = -3, 5
+      do k = 12, 19
+         if (a(i, j, k) .ne. b(i, j, k)) call abort
+      end do
+    end do
+  end do
+end program collapse7
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/collapse-8.f90 b/libgomp/testsuite/libgomp.oacc-fortran/collapse-8.f90
new file mode 100644
index 00000000000..04fbcfef49a
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/collapse-8.f90
@@ -0,0 +1,47 @@
+! { dg-do run }
+
+! collapse3.f90:test5
+program collapse8
+  integer :: i, j, k, a(1:7, -3:5, 12:19), b(1:7, -3:5, 12:19)
+  integer :: v1, v2, v3, v4, v5, v6
+  logical :: l, r
+  l = .false.
+  r = .false.
+  a(:, :, :) = 0
+  b(:, :, :) = 0
+  v1 = 3
+  v2 = 6
+  v3 = -2
+  v4 = 4
+  v5 = 13
+  v6 = 18
+  !$acc parallel
+  !$acc loop collapse (3) reduction (.or.:l)
+    do i = v1, v2
+      do j = v3, v4
+        do k = v5, v6
+          l = l.or.i.lt.2.or.i.gt.6.or.j.lt.-2.or.j.gt.4
+          l = l.or.k.lt.13.or.k.gt.18
+          if (.not.l) a(i, j, k) = a(i, j, k) + 1
+        end do
+      end do
+    end do
+  !$acc end parallel
+  do i = v1, v2
+    do j = v3, v4
+      do k = v5, v6
+        r = r.or.i.lt.2.or.i.gt.6.or.j.lt.-2.or.j.gt.4
+        r = r.or.k.lt.13.or.k.gt.18
+        if (.not.r) b(i, j, k) = b(i, j, k) + 1
+      end do
+    end do
+  end do
+  if (l .neqv. r) call abort
+  do i = v1, v2
+    do j = v3, v4
+      do k = v5, v6
+         if (a(i, j, k) .ne. b(i, j, k)) call abort
+      end do
+    end do
+  end do
+end program collapse8
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/data-1.f90 b/libgomp/testsuite/libgomp.oacc-fortran/data-1.f90
new file mode 100644
index 00000000000..5e94e2d7f47
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/data-1.f90
@@ -0,0 +1,45 @@
+! { dg-do run }
+
+program test
+  integer, parameter :: N = 8
+  real, allocatable :: a(:), b(:)
+
+  allocate (a(N))
+  allocate (b(N))
+
+  a(:) = 3.0
+  b(:) = 0.0
+
+  !$acc enter data copyin (a(1:N), b(1:N))
+
+  !$acc parallel
+  do i = 1, n
+    b(i) = a (i)
+  end do
+  !$acc end parallel
+
+  !$acc exit data copyout (a(1:N), b(1:N))
+
+  do i = 1, n
+    if (a(i) .ne. 3.0) call abort
+    if (b(i) .ne. 3.0) call abort
+  end do
+
+  a(:) = 5.0
+  b(:) = 1.0
+
+  !$acc enter data copyin (a(1:N), b(1:N))
+
+  !$acc parallel
+  do i = 1, n
+    b(i) = a (i)
+  end do
+  !$acc end parallel
+
+  !$acc exit data copyout (a(1:N), b(1:N))
+
+  do i = 1, n
+    if (a(i) .ne. 5.0) call abort
+    if (b(i) .ne. 5.0) call abort
+  end do
+end program test
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/data-2.f90 b/libgomp/testsuite/libgomp.oacc-fortran/data-2.f90
new file mode 100644
index 00000000000..8736c2a1f10
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/data-2.f90
@@ -0,0 +1,31 @@
+! { dg-do run }
+
+program test
+  integer, parameter :: N = 8
+  real, allocatable :: a(:,:), b(:,:)
+
+  allocate (a(N,N))
+  allocate (b(N,N))
+
+  a(:,:) = 3.0
+  b(:,:) = 0.0
+
+  !$acc enter data copyin (a(1:N,1:N), b(1:N,1:N))
+
+  !$acc parallel
+  do i = 1, n
+    do j = 1, n
+      b(j,i) = a (j,i)
+    end do
+  end do
+  !$acc end parallel
+
+  !$acc exit data copyout (a(1:N,1:N), b(1:N,1:N))
+
+  do i = 1, n
+    do j = 1, n
+      if (a(j,i) .ne. 3.0) call abort
+      if (b(j,i) .ne. 3.0) call abort
+    end do
+  end do
+end program test
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/data-3.f90 b/libgomp/testsuite/libgomp.oacc-fortran/data-3.f90
new file mode 100644
index 00000000000..9868cb0da9d
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/data-3.f90
@@ -0,0 +1,131 @@
+! { dg-do run }
+
+program asyncwait
+  real, allocatable :: a(:), b(:), c(:), d(:), e(:)
+  integer i, N
+
+  N = 64
+
+  allocate (a(N))
+  allocate (b(N))
+  allocate (c(N))
+  allocate (d(N))
+  allocate (e(N))
+
+  a(:) = 3.0
+  b(:) = 0.0
+
+  !$acc enter data copyin (a(1:N)) copyin (b(1:N)) copyin (N) async
+
+  !$acc parallel async wait
+  do i = 1, N
+     b(i) = a(i)
+  end do
+  !$acc end parallel
+
+  !$acc wait
+  !$acc exit data copyout (a(1:N)) copyout (b(1:N))
+
+  do i = 1, N
+     if (a(i) .ne. 3.0) call abort
+     if (b(i) .ne. 3.0) call abort
+  end do
+
+  a(:) = 2.0
+  b(:) = 0.0
+
+  !$acc enter data copyin (a(1:N)) copyin (b(1:N)) async (1)
+
+  !$acc parallel async (1) wait (1)
+  do i = 1, N
+     b(i) = a(i)
+  end do
+  !$acc end parallel
+
+  !$acc wait (1)
+  !$acc exit data copyout (a(1:N)) copyout (b(1:N))
+
+  do i = 1, N
+     if (a(i) .ne. 2.0) call abort
+     if (b(i) .ne. 2.0) call abort
+  end do
+
+  a(:) = 3.0
+  b(:) = 0.0
+  c(:) = 0.0
+  d(:) = 0.0
+
+  !$acc enter data copyin (a(1:N)) create (b(1:N)) create (c(1:N)) create (d(1:N))
+
+  !$acc parallel async (1)
+  do i = 1, N
+     b(i) = (a(i) * a(i) * a(i)) / a(i)
+  end do
+  !$acc end parallel
+
+  !$acc parallel async (1)
+  do i = 1, N
+     c(i) = (a(i) * 4) / a(i)
+  end do
+  !$acc end parallel
+
+  !$acc parallel async (1)
+  do i = 1, N
+     d(i) = ((a(i) * a(i)  + a(i)) / a(i)) - a(i)
+  end do
+  !$acc end parallel
+
+  !$acc wait (1)
+  !$acc exit data copyout (a(1:N)) copyout (b(1:N)) copyout (c(1:N)) copyout (d(1:N))
+
+  do i = 1, N
+     if (a(i) .ne. 3.0) call abort
+     if (b(i) .ne. 9.0) call abort
+     if (c(i) .ne. 4.0) call abort
+     if (d(i) .ne. 1.0) call abort
+  end do
+
+  a(:) = 2.0
+  b(:) = 0.0
+  c(:) = 0.0
+  d(:) = 0.0
+  e(:) = 0.0
+
+  !$acc enter data copyin (a(1:N)) create (b(1:N)) create (c(1:N)) create (d(1:N)) copyin (e(1:N))
+
+  !$acc parallel async (1)
+  do i = 1, N
+     b(i) = (a(i) * a(i) * a(i)) / a(i)
+  end do
+  !$acc end parallel
+
+  !$acc parallel async (1)
+  do i = 1, N
+     c(i) = (a(i) * 4) / a(i)
+  end do
+  !$acc end parallel
+
+  !$acc parallel async (1)
+  do i = 1, N
+     d(i) = ((a(i) * a(i) + a(i)) / a(i)) - a(i)
+  end do
+  !$acc end parallel
+
+  !$acc parallel wait (1) async (1)
+  do i = 1, N
+     e(i) = a(i) + b(i) + c(i) + d(i)
+  end do
+  !$acc end parallel
+
+  !$acc wait (1)
+  !$acc exit data copyout (a(1:N)) copyout (b(1:N)) copyout (c(1:N)) copyout (d(1:N)) copyout (e(1:N))
+  !$acc exit data delete (N)
+
+  do i = 1, N
+     if (a(i) .ne. 2.0) call abort
+     if (b(i) .ne. 4.0) call abort
+     if (c(i) .ne. 4.0) call abort
+     if (d(i) .ne. 1.0) call abort
+     if (e(i) .ne. 11.0) call abort
+  end do
+end program asyncwait
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/data-4-2.f90 b/libgomp/testsuite/libgomp.oacc-fortran/data-4-2.f90
new file mode 100644
index 00000000000..16a85980b4d
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/data-4-2.f90
@@ -0,0 +1,138 @@
+! Copy of data-4.f90 with self exchanged with host for !acc update.
+
+! { dg-do run }
+
+program asyncwait
+  real, allocatable :: a(:), b(:), c(:), d(:), e(:)
+  integer i, N
+
+  N = 64
+
+  allocate (a(N))
+  allocate (b(N))
+  allocate (c(N))
+  allocate (d(N))
+  allocate (e(N))
+
+  a(:) = 3.0
+  b(:) = 0.0
+
+  !$acc enter data copyin (a(1:N)) copyin (b(1:N)) copyin (N) async
+
+  !$acc parallel async wait
+  !$acc loop
+  do i = 1, N
+     b(i) = a(i)
+  end do
+  !$acc end parallel
+
+  !$acc update self (a(1:N), b(1:N)) async wait
+  !$acc wait
+
+  do i = 1, N
+     if (a(i) .ne. 3.0) call abort
+     if (b(i) .ne. 3.0) call abort
+  end do
+
+  a(:) = 2.0
+  b(:) = 0.0
+
+  !$acc update device (a(1:N), b(1:N)) async (1)
+
+  !$acc parallel async (1) wait (1)
+  !$acc loop
+  do i = 1, N
+     b(i) = a(i)
+  end do
+  !$acc end parallel
+
+  !$acc update host (a(1:N), b(1:N)) async (1) wait (1)
+  !$acc wait (1)
+
+  do i = 1, N
+     if (a(i) .ne. 2.0) call abort
+     if (b(i) .ne. 2.0) call abort
+  end do
+
+  a(:) = 3.0
+  b(:) = 0.0
+  c(:) = 0.0
+  d(:) = 0.0
+
+  !$acc enter data copyin (c(1:N), d(1:N)) async (1)
+  !$acc update device (a(1:N), b(1:N)) async (1)
+
+  !$acc parallel async (1)
+  do i = 1, N
+     b(i) = (a(i) * a(i) * a(i)) / a(i)
+  end do
+  !$acc end parallel
+
+  !$acc parallel async (1)
+  do i = 1, N
+     c(i) = (a(i) * 4) / a(i)
+  end do
+  !$acc end parallel
+
+  !$acc parallel async (1)
+  do i = 1, N
+     d(i) = ((a(i) * a(i)  + a(i)) / a(i)) - a(i)
+  end do
+  !$acc end parallel
+
+  !$acc update self (a(1:N), b(1:N), c(1:N), d(1:N)) async (1) wait (1)
+
+  !$acc wait (1)
+
+  do i = 1, N
+     if (a(i) .ne. 3.0) call abort
+     if (b(i) .ne. 9.0) call abort
+     if (c(i) .ne. 4.0) call abort
+     if (d(i) .ne. 1.0) call abort
+  end do
+
+  a(:) = 2.0
+  b(:) = 0.0
+  c(:) = 0.0
+  d(:) = 0.0
+  e(:) = 0.0
+
+  !$acc enter data copyin (e(1:N)) async (1)
+  !$acc update device (a(1:N), b(1:N), c(1:N), d(1:N)) async (1)
+
+  !$acc parallel async (1)
+  do i = 1, N
+     b(i) = (a(i) * a(i) * a(i)) / a(i)
+  end do
+  !$acc end parallel
+
+  !$acc parallel async (1)
+  do i = 1, N
+     c(i) = (a(i) * 4) / a(i)
+  end do
+  !$acc end parallel
+
+  !$acc parallel async (1)
+  do i = 1, N
+     d(i) = ((a(i) * a(i) + a(i)) / a(i)) - a(i)
+  end do
+  !$acc end parallel
+
+  !$acc parallel wait (1) async (1)
+  do i = 1, N
+     e(i) = a(i) + b(i) + c(i) + d(i)
+  end do
+  !$acc end parallel
+
+  !$acc update self (a(1:N), b(1:N), c(1:N), d(1:N), e(1:N)) async (1) wait (1)
+  !$acc wait (1)
+  !$acc exit data delete (N, a(1:N), b(1:N), c(1:N), d(1:N), e(1:N))
+
+  do i = 1, N
+     if (a(i) .ne. 2.0) call abort
+     if (b(i) .ne. 4.0) call abort
+     if (c(i) .ne. 4.0) call abort
+     if (d(i) .ne. 1.0) call abort
+     if (e(i) .ne. 11.0) call abort
+  end do
+end program asyncwait
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/data-4.f90 b/libgomp/testsuite/libgomp.oacc-fortran/data-4.f90
new file mode 100644
index 00000000000..f6886b0e1e3
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/data-4.f90
@@ -0,0 +1,136 @@
+! { dg-do run }
+
+program asyncwait
+  real, allocatable :: a(:), b(:), c(:), d(:), e(:)
+  integer i, N
+
+  N = 64
+
+  allocate (a(N))
+  allocate (b(N))
+  allocate (c(N))
+  allocate (d(N))
+  allocate (e(N))
+
+  a(:) = 3.0
+  b(:) = 0.0
+
+  !$acc enter data copyin (a(1:N)) copyin (b(1:N)) copyin (N) async
+
+  !$acc parallel async wait
+  !$acc loop
+  do i = 1, N
+     b(i) = a(i)
+  end do
+  !$acc end parallel
+
+  !$acc update host (a(1:N), b(1:N)) async wait
+  !$acc wait
+
+  do i = 1, N
+     if (a(i) .ne. 3.0) call abort
+     if (b(i) .ne. 3.0) call abort
+  end do
+
+  a(:) = 2.0
+  b(:) = 0.0
+
+  !$acc update device (a(1:N), b(1:N)) async (1)
+
+  !$acc parallel async (1) wait (1)
+  !$acc loop
+  do i = 1, N
+     b(i) = a(i)
+  end do
+  !$acc end parallel
+
+  !$acc update self (a(1:N), b(1:N)) async (1) wait (1)
+  !$acc wait (1)
+
+  do i = 1, N
+     if (a(i) .ne. 2.0) call abort
+     if (b(i) .ne. 2.0) call abort
+  end do
+
+  a(:) = 3.0
+  b(:) = 0.0
+  c(:) = 0.0
+  d(:) = 0.0
+
+  !$acc enter data copyin (c(1:N), d(1:N)) async (1)
+  !$acc update device (a(1:N), b(1:N)) async (1)
+
+  !$acc parallel async (1)
+  do i = 1, N
+     b(i) = (a(i) * a(i) * a(i)) / a(i)
+  end do
+  !$acc end parallel
+
+  !$acc parallel async (1)
+  do i = 1, N
+     c(i) = (a(i) * 4) / a(i)
+  end do
+  !$acc end parallel
+
+  !$acc parallel async (1)
+  do i = 1, N
+     d(i) = ((a(i) * a(i)  + a(i)) / a(i)) - a(i)
+  end do
+  !$acc end parallel
+
+  !$acc update host (a(1:N), b(1:N), c(1:N), d(1:N)) async (1) wait (1)
+
+  !$acc wait (1)
+
+  do i = 1, N
+     if (a(i) .ne. 3.0) call abort
+     if (b(i) .ne. 9.0) call abort
+     if (c(i) .ne. 4.0) call abort
+     if (d(i) .ne. 1.0) call abort
+  end do
+
+  a(:) = 2.0
+  b(:) = 0.0
+  c(:) = 0.0
+  d(:) = 0.0
+  e(:) = 0.0
+
+  !$acc enter data copyin (e(1:N)) async (1)
+  !$acc update device (a(1:N), b(1:N), c(1:N), d(1:N)) async (1)
+
+  !$acc parallel async (1)
+  do i = 1, N
+     b(i) = (a(i) * a(i) * a(i)) / a(i)
+  end do
+  !$acc end parallel
+
+  !$acc parallel async (1)
+  do i = 1, N
+     c(i) = (a(i) * 4) / a(i)
+  end do
+  !$acc end parallel
+
+  !$acc parallel async (1)
+  do i = 1, N
+     d(i) = ((a(i) * a(i) + a(i)) / a(i)) - a(i)
+  end do
+  !$acc end parallel
+
+  !$acc parallel wait (1) async (1)
+  do i = 1, N
+     e(i) = a(i) + b(i) + c(i) + d(i)
+  end do
+  !$acc end parallel
+
+  !$acc update host (a(1:N), b(1:N), c(1:N), d(1:N), e(1:N)) async (1) wait (1)
+  !$acc wait (1)
+  !$acc exit data delete (N, a(1:N), b(1:N), c(1:N), d(1:N), e(1:N))
+
+  do i = 1, N
+     if (a(i) .ne. 2.0) call abort
+     if (b(i) .ne. 4.0) call abort
+     if (c(i) .ne. 4.0) call abort
+     if (d(i) .ne. 1.0) call abort
+     if (e(i) .ne. 11.0) call abort
+  end do
+end program asyncwait
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/data-already-1.f b/libgomp/testsuite/libgomp.oacc-fortran/data-already-1.f
new file mode 100644
index 00000000000..ac220ab4c7e
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/data-already-1.f
@@ -0,0 +1,17 @@
+! { dg-skip-if "" { *-*-* } { "*" } { "-DACC_MEM_SHARED=0" } }
+
+      IMPLICIT NONE
+      INCLUDE "openacc_lib.h"
+
+      INTEGER I
+
+      CALL ACC_COPYIN (I)
+
+!$ACC DATA COPY (I)
+      I = 0
+!$ACC END DATA
+
+      END
+
+! { dg-shouldfail "" }
+! { dg-output "Trying to map into device .* object when .* is already mapped" }
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/data-already-2.f b/libgomp/testsuite/libgomp.oacc-fortran/data-already-2.f
new file mode 100644
index 00000000000..2c5254b8684
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/data-already-2.f
@@ -0,0 +1,16 @@
+! { dg-skip-if "" { *-*-* } { "*" } { "-DACC_MEM_SHARED=0" } }
+
+      IMPLICIT NONE
+
+      INTEGER I
+
+!$ACC DATA PRESENT_OR_COPY (I)
+!$ACC DATA COPYOUT (I)
+      I = 0
+!$ACC END DATA
+!$ACC END DATA
+
+      END
+
+! { dg-shouldfail "" }
+! { dg-output "Trying to map into device .* object when .* is already mapped" }
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/data-already-3.f b/libgomp/testsuite/libgomp.oacc-fortran/data-already-3.f
new file mode 100644
index 00000000000..c41de28f539
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/data-already-3.f
@@ -0,0 +1,15 @@
+! { dg-skip-if "" { *-*-* } { "*" } { "-DACC_MEM_SHARED=0" } }
+
+      IMPLICIT NONE
+      INCLUDE "openacc_lib.h"
+
+      INTEGER I
+
+!$ACC DATA PRESENT_OR_COPY (I)
+      CALL ACC_COPYIN (I)
+!$ACC END DATA
+
+      END
+
+! { dg-shouldfail "" }
+! { dg-output "already mapped to" }
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/data-already-4.f b/libgomp/testsuite/libgomp.oacc-fortran/data-already-4.f
new file mode 100644
index 00000000000..f54bf580ce9
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/data-already-4.f
@@ -0,0 +1,14 @@
+! { dg-skip-if "" { *-*-* } { "*" } { "-DACC_MEM_SHARED=0" } }
+
+      IMPLICIT NONE
+      INCLUDE "openacc_lib.h"
+
+      INTEGER I
+
+      CALL ACC_PRESENT_OR_COPYIN (I)
+      CALL ACC_COPYIN (I)
+
+      END
+
+! { dg-shouldfail "" }
+! { dg-output "already mapped to" }
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/data-already-5.f b/libgomp/testsuite/libgomp.oacc-fortran/data-already-5.f
new file mode 100644
index 00000000000..9a3e94fac7a
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/data-already-5.f
@@ -0,0 +1,14 @@
+! { dg-skip-if "" { *-*-* } { "*" } { "-DACC_MEM_SHARED=0" } }
+
+      IMPLICIT NONE
+      INCLUDE "openacc_lib.h"
+
+      INTEGER I
+
+!$ACC ENTER DATA CREATE (I)
+      CALL ACC_COPYIN (I)
+
+      END
+
+! { dg-shouldfail "" }
+! { dg-output "already mapped to" }
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/data-already-6.f b/libgomp/testsuite/libgomp.oacc-fortran/data-already-6.f
new file mode 100644
index 00000000000..eaf5d982492
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/data-already-6.f
@@ -0,0 +1,14 @@
+! { dg-skip-if "" { *-*-* } { "*" } { "-DACC_MEM_SHARED=0" } }
+
+      IMPLICIT NONE
+      INCLUDE "openacc_lib.h"
+
+      INTEGER I
+
+      CALL ACC_PRESENT_OR_COPYIN (I)
+!$ACC ENTER DATA CREATE (I)
+
+      END
+
+! { dg-shouldfail "" }
+! { dg-output "already mapped to" }
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/data-already-7.f b/libgomp/testsuite/libgomp.oacc-fortran/data-already-7.f
new file mode 100644
index 00000000000..d96bf0b2f84
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/data-already-7.f
@@ -0,0 +1,14 @@
+! { dg-skip-if "" { *-*-* } { "*" } { "-DACC_MEM_SHARED=0" } }
+
+      IMPLICIT NONE
+      INCLUDE "openacc_lib.h"
+
+      INTEGER I
+
+!$ACC ENTER DATA CREATE (I)
+      CALL ACC_CREATE (I)
+
+      END
+
+! { dg-shouldfail "" }
+! { dg-output "already mapped to" }
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/data-already-8.f b/libgomp/testsuite/libgomp.oacc-fortran/data-already-8.f
new file mode 100644
index 00000000000..16da048861a
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/data-already-8.f
@@ -0,0 +1,16 @@
+! { dg-skip-if "" { *-*-* } { "*" } { "-DACC_MEM_SHARED=0" } }
+
+      IMPLICIT NONE
+
+      INTEGER I
+
+!$ACC DATA CREATE (I)
+!$ACC PARALLEL COPYIN (I)
+      I = 0
+!$ACC END PARALLEL
+!$ACC END DATA
+
+      END
+
+! { dg-shouldfail "" }
+! { dg-output "Trying to map into device .* object when .* is already mapped" }
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/fortran.exp b/libgomp/testsuite/libgomp.oacc-fortran/fortran.exp
new file mode 100644
index 00000000000..a8f62e84d67
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/fortran.exp
@@ -0,0 +1,98 @@
+# This whole file adapted from libgomp.fortran/fortran.exp.
+
+load_lib libgomp-dg.exp
+load_gcc_lib gcc-dg.exp
+load_gcc_lib gfortran-dg.exp
+
+global shlib_ext
+global ALWAYS_CFLAGS
+
+set shlib_ext [get_shlib_extension]
+set lang_library_path	"../libgfortran/.libs"
+set lang_link_flags	"-lgfortran"
+if [info exists lang_include_flags] then {
+    unset lang_include_flags
+}
+set lang_test_file_found 0
+set quadmath_library_path "../libquadmath/.libs"
+
+
+# Initialize dg.
+dg-init
+
+# Turn on OpenACC.
+lappend ALWAYS_CFLAGS "additional_flags=-fopenacc"
+
+if { $blddir != "" } {
+    set lang_source_re {^.*\.[fF](|90|95|03|08)$}
+    set lang_include_flags "-fintrinsic-modules-path=${blddir}"
+    # Look for a static libgfortran first.
+    if [file exists "${blddir}/${lang_library_path}/libgfortran.a"] {
+        set lang_test_file "${lang_library_path}/libgfortran.a"
+        set lang_test_file_found 1
+	# We may have a shared only build, so look for a shared libgfortran.
+    } elseif [file exists "${blddir}/${lang_library_path}/libgfortran.${shlib_ext}"] {
+        set lang_test_file "${lang_library_path}/libgfortran.${shlib_ext}"
+        set lang_test_file_found 1
+    } else {
+        puts "No libgfortran library found, will not execute fortran tests"
+    }
+} elseif [info exists GFORTRAN_UNDER_TEST] {
+    set lang_test_file_found 1
+    # Needs to exist for libgomp.exp.
+    set lang_test_file ""
+} else {
+    puts "GFORTRAN_UNDER_TEST not defined, will not execute fortran tests"
+}
+
+if { $lang_test_file_found } {
+    # Gather a list of all tests.
+    set tests [lsort [find $srcdir/$subdir *.\[fF\]{,90,95,03,08}]]
+
+    if { $blddir != "" } {
+	if { [file exists "${blddir}/${quadmath_library_path}/libquadmath.a"]
+	     || [file exists "${blddir}/${quadmath_library_path}/libquadmath.${shlib_ext}"] } {
+	    lappend ALWAYS_CFLAGS "ldflags=-L${blddir}/${quadmath_library_path}/"
+	    # Allow for spec subsitution.
+	    lappend ALWAYS_CFLAGS "additional_flags=-B${blddir}/${quadmath_library_path}/"
+	    set ld_library_path "$always_ld_library_path:${blddir}/${lang_library_path}:${blddir}/${quadmath_library_path}"
+	} else {
+	    set ld_library_path "$always_ld_library_path:${blddir}/${lang_library_path}"
+	}
+    } else {
+        set ld_library_path "$always_ld_library_path"
+    }
+    append ld_library_path [gcc-set-multilib-library-path $GCC_UNDER_TEST]
+    set_ld_library_path_env_vars
+
+    # Test OpenACC with available accelerators.
+    foreach offload_target_openacc $offload_targets_s_openacc {
+	set tagopt "-DACC_DEVICE_TYPE_$offload_target_openacc=1"
+
+	switch $offload_target_openacc {
+	    host {
+		set acc_mem_shared 1
+	    }
+	    host_nonshm {
+		set acc_mem_shared 0
+	    }
+	    nvidia {
+		set acc_mem_shared 0
+	    }
+	    default {
+		set acc_mem_shared 0
+	    }
+	}
+	set tagopt "$tagopt -DACC_MEM_SHARED=$acc_mem_shared"
+
+	setenv ACC_DEVICE_TYPE $offload_target_openacc
+
+	# For Fortran we're doing torture testing, as Fortran has far more tests
+	# with arrays etc. that testing just -O0 or -O2 is insufficient, that is
+	# typically not the case for C/C++.
+	gfortran-dg-runtest $tests "$tagopt" ""
+    }
+}
+
+# All done.
+dg-finish
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/lib-1.f90 b/libgomp/testsuite/libgomp.oacc-fortran/lib-1.f90
new file mode 100644
index 00000000000..51dc452764b
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/lib-1.f90
@@ -0,0 +1,13 @@
+use openacc
+
+if (acc_get_num_devices (acc_device_host) .ne. 1) call abort
+call acc_set_device_type (acc_device_host)
+if (acc_get_device_type () .ne. acc_device_host) call abort
+call acc_set_device_num (0, acc_device_host)
+if (acc_get_device_num (acc_device_host) .ne. 0) call abort
+call acc_shutdown (acc_device_host)
+
+call acc_init (acc_device_host)
+call acc_shutdown (acc_device_host)
+
+end
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/lib-10.f90 b/libgomp/testsuite/libgomp.oacc-fortran/lib-10.f90
new file mode 100644
index 00000000000..a54d6a7d92e
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/lib-10.f90
@@ -0,0 +1,82 @@
+! { dg-do run }
+
+program main
+  implicit none
+  include "openacc_lib.h"
+
+  integer, target :: a_3d_i(10, 10, 10)
+  complex a_3d_c(10, 10, 10)
+  real a_3d_r(10, 10, 10)
+
+  integer i, j, k
+  complex c
+  real r
+  integer, parameter :: i_size = sizeof (i)
+  integer, parameter :: c_size = sizeof (c)
+  integer, parameter :: r_size = sizeof (r)
+
+  if (acc_get_num_devices (acc_device_nvidia) .eq. 0) call exit
+
+  call acc_init (acc_device_nvidia)
+
+  call set3d (.FALSE., a_3d_i, a_3d_c, a_3d_r)
+
+  call acc_copyin (a_3d_i)
+  call acc_copyin (a_3d_c)
+  call acc_copyin (a_3d_r)
+
+  if (acc_is_present (a_3d_i) .neqv. .TRUE.) call abort
+  if (acc_is_present (a_3d_c) .neqv. .TRUE.) call abort
+  if (acc_is_present (a_3d_r) .neqv. .TRUE.) call abort
+
+  do i = 1, 10
+    do j = 1, 10
+      do k = 1, 10
+        if (acc_is_present (a_3d_i(i, j, k), i_size) .neqv. .TRUE.) call abort
+        if (acc_is_present (a_3d_c(i, j, k), i_size) .neqv. .TRUE.) call abort
+        if (acc_is_present (a_3d_r(i, j, k), i_size) .neqv. .TRUE.) call abort
+      end do
+    end do
+  end do
+
+  call acc_shutdown (acc_device_nvidia)
+
+contains
+
+  subroutine set3d (clear, a_i, a_c, a_r)
+  logical clear
+  integer, dimension (:,:,:), intent (inout) :: a_i
+  complex, dimension (:,:,:), intent (inout) :: a_c
+  real, dimension (:,:,:), intent (inout) :: a_r
+
+  integer i, j, k
+  integer lb1, ub1, lb2, ub2, lb3, ub3
+
+  lb1 = lbound (a_i, 1)
+  ub1 = ubound (a_i, 1)
+
+  lb2 = lbound (a_i, 2)
+  ub2 = ubound (a_i, 2)
+
+  lb3 = lbound (a_i, 3)
+  ub3 = ubound (a_i, 3)
+
+  do i = lb1, ub1
+    do j = lb2, ub2
+      do k = lb3, ub3
+        if (clear) then
+          a_i(i, j, k) = 0
+          a_c(i, j, k) = cmplx (0.0, 0.0)
+          a_r(i, j, k) = 0.0
+        else
+          a_i(i, j, k) = i
+          a_c(i, j, k) = cmplx (i, j)
+          a_r(i, j, k) = i
+        end if
+      end do
+    end do
+  end do
+
+  end subroutine
+
+end program
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/lib-2.f b/libgomp/testsuite/libgomp.oacc-fortran/lib-2.f
new file mode 100644
index 00000000000..a9d70b2489b
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/lib-2.f
@@ -0,0 +1,13 @@
+      USE OPENACC
+
+      IF (ACC_GET_NUM_DEVICES (ACC_DEVICE_HOST) .NE. 1) CALL ABORT
+      CALL ACC_SET_DEVICE_TYPE (ACC_DEVICE_HOST)
+      IF (ACC_GET_DEVICE_TYPE () .NE. ACC_DEVICE_HOST) CALL ABORT
+      CALL ACC_SET_DEVICE_NUM (0, ACC_DEVICE_HOST)
+      IF (ACC_GET_DEVICE_NUM (ACC_DEVICE_HOST) .NE. 0) CALL ABORT
+      CALL ACC_SHUTDOWN (ACC_DEVICE_HOST)
+
+      CALL ACC_INIT (ACC_DEVICE_HOST)
+      CALL ACC_SHUTDOWN (ACC_DEVICE_HOST)
+
+      END
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/lib-3.f b/libgomp/testsuite/libgomp.oacc-fortran/lib-3.f
new file mode 100644
index 00000000000..56d2cd2b961
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/lib-3.f
@@ -0,0 +1,13 @@
+      INCLUDE "openacc_lib.h"
+
+      IF (ACC_GET_NUM_DEVICES (ACC_DEVICE_HOST) .NE. 1) CALL ABORT
+      CALL ACC_SET_DEVICE_TYPE (ACC_DEVICE_HOST)
+      IF (ACC_GET_DEVICE_TYPE () .NE. ACC_DEVICE_HOST) CALL ABORT
+      CALL ACC_SET_DEVICE_NUM (0, ACC_DEVICE_HOST)
+      IF (ACC_GET_DEVICE_NUM (ACC_DEVICE_HOST) .NE. 0) CALL ABORT
+      CALL ACC_SHUTDOWN (ACC_DEVICE_HOST)
+
+      CALL ACC_INIT (ACC_DEVICE_HOST)
+      CALL ACC_SHUTDOWN (ACC_DEVICE_HOST)
+
+      END
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/lib-4.f90 b/libgomp/testsuite/libgomp.oacc-fortran/lib-4.f90
new file mode 100644
index 00000000000..3a2b661b5b7
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/lib-4.f90
@@ -0,0 +1,35 @@
+! { dg-do run }
+
+program main
+  use openacc
+  implicit none
+
+  integer n
+
+  if (acc_get_num_devices (acc_device_host) .ne. 1) call abort
+
+  if (acc_get_num_devices (acc_device_none) .ne. 0) call abort
+
+  call acc_init (acc_device_host)
+
+  if (acc_get_device_type () .ne. acc_device_host) call abort
+
+  call acc_set_device_type (acc_device_host)
+
+  if (acc_get_device_type () .ne. acc_device_host) call abort
+
+  n = 0
+
+  call acc_set_device_num (n, acc_device_host)
+
+  if (acc_get_device_num (acc_device_host) .ne. 0) call abort
+
+  if (.NOT. acc_async_test (n) ) call abort
+
+  call acc_wait (n)
+
+  call acc_wait_all ()
+
+  call acc_shutdown (acc_device_host)
+
+end program
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/lib-5.f90 b/libgomp/testsuite/libgomp.oacc-fortran/lib-5.f90
new file mode 100644
index 00000000000..e68eb890e63
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/lib-5.f90
@@ -0,0 +1,31 @@
+! { dg-do run }
+
+program main
+  use openacc
+  implicit none
+
+  integer n
+
+  if (acc_get_num_devices (acc_device_nvidia) .eq. 0) call exit
+
+  call acc_init (acc_device_nvidia)
+
+  n = 0
+
+  call acc_set_device_num (n, acc_device_nvidia)
+
+  if (acc_get_device_num (acc_device_nvidia) .ne. 0) call abort
+
+  if (acc_get_num_devices (acc_device_nvidia) .gt. 1) then
+
+    n = 1
+
+    call acc_set_device_num (n, acc_device_nvidia)
+
+    if (acc_get_device_num (acc_device_nvidia) .ne. 1) call abort
+
+  end if
+
+  call acc_shutdown (acc_device_nvidia)
+
+end program
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/lib-6.f90 b/libgomp/testsuite/libgomp.oacc-fortran/lib-6.f90
new file mode 100644
index 00000000000..401ad661ee8
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/lib-6.f90
@@ -0,0 +1,35 @@
+! { dg-do run }
+
+program main
+  implicit none
+  include "openacc_lib.h"
+
+  integer n
+
+  if (acc_get_num_devices (acc_device_host) .ne. 1) call abort
+
+  if (acc_get_num_devices (acc_device_none) .ne. 0) call abort
+
+  call acc_init (acc_device_host)
+
+  if (acc_get_device_type () .ne. acc_device_host) call abort
+
+  call acc_set_device_type (acc_device_host)
+
+  if (acc_get_device_type () .ne. acc_device_host) call abort
+
+  n = 0
+
+  call acc_set_device_num (n, acc_device_host)
+
+  if (acc_get_device_num (acc_device_host) .ne. 0) call abort
+
+  if (.NOT. acc_async_test (n) ) call abort
+
+  call acc_wait (n)
+
+  call acc_wait_all ()
+
+  call acc_shutdown (acc_device_host)
+
+end program
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/lib-7.f90 b/libgomp/testsuite/libgomp.oacc-fortran/lib-7.f90
new file mode 100644
index 00000000000..422df53bd3a
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/lib-7.f90
@@ -0,0 +1,31 @@
+! { dg-do run }
+
+program main
+  implicit none
+  include "openacc_lib.h"
+
+  integer n
+
+  if (acc_get_num_devices (acc_device_nvidia) .eq. 0) call exit
+
+  call acc_init (acc_device_nvidia)
+
+  n = 0
+
+  call acc_set_device_num (n, acc_device_nvidia)
+
+  if (acc_get_device_num (acc_device_nvidia) .ne. 0) call abort
+
+  if (acc_get_num_devices (acc_device_nvidia) .gt. 1) then
+
+    n = 1
+
+    call acc_set_device_num (n, acc_device_nvidia)
+
+    if (acc_get_device_num (acc_device_nvidia) .ne. 1) call abort
+
+  end if
+
+  call acc_shutdown (acc_device_nvidia)
+
+end program
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/lib-8.f90 b/libgomp/testsuite/libgomp.oacc-fortran/lib-8.f90
new file mode 100644
index 00000000000..ad758b2e9d5
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/lib-8.f90
@@ -0,0 +1,83 @@
+! { dg-do run }
+
+program main
+  use openacc
+  use iso_c_binding
+  implicit none
+
+  integer, target :: a_3d_i(10, 10, 10)
+  complex a_3d_c(10, 10, 10)
+  real a_3d_r(10, 10, 10)
+
+  integer i, j, k
+  complex c
+  real r
+  integer, parameter :: i_size = sizeof (i)
+  integer, parameter :: c_size = sizeof (c)
+  integer, parameter :: r_size = sizeof (r)
+
+  if (acc_get_num_devices (acc_device_nvidia) .eq. 0) call exit
+
+  call acc_init (acc_device_nvidia)
+
+  call set3d (.FALSE., a_3d_i, a_3d_c, a_3d_r)
+
+  call acc_copyin (a_3d_i)
+  call acc_copyin (a_3d_c)
+  call acc_copyin (a_3d_r)
+
+  if (acc_is_present (a_3d_i) .neqv. .TRUE.) call abort
+  if (acc_is_present (a_3d_c) .neqv. .TRUE.) call abort
+  if (acc_is_present (a_3d_r) .neqv. .TRUE.) call abort
+
+  do i = 1, 10
+    do j = 1, 10
+      do k = 1, 10
+        if (acc_is_present (a_3d_i(i, j, k), i_size) .neqv. .TRUE.) call abort
+        if (acc_is_present (a_3d_c(i, j, k), i_size) .neqv. .TRUE.) call abort
+        if (acc_is_present (a_3d_r(i, j, k), i_size) .neqv. .TRUE.) call abort
+      end do
+    end do
+  end do
+
+  call acc_shutdown (acc_device_nvidia)
+
+contains
+
+  subroutine set3d (clear, a_i, a_c, a_r)
+  logical clear
+  integer, dimension (:,:,:), intent (inout) :: a_i
+  complex, dimension (:,:,:), intent (inout) :: a_c
+  real, dimension (:,:,:), intent (inout) :: a_r
+
+  integer i, j, k
+  integer lb1, ub1, lb2, ub2, lb3, ub3
+
+  lb1 = lbound (a_i, 1)
+  ub1 = ubound (a_i, 1)
+
+  lb2 = lbound (a_i, 2)
+  ub2 = ubound (a_i, 2)
+
+  lb3 = lbound (a_i, 3)
+  ub3 = ubound (a_i, 3)
+
+  do i = lb1, ub1
+    do j = lb2, ub2
+      do k = lb3, ub3
+        if (clear) then
+          a_i(i, j, k) = 0
+          a_c(i, j, k) = cmplx (0.0, 0.0)
+          a_r(i, j, k) = 0.0
+        else
+          a_i(i, j, k) = i
+          a_c(i, j, k) = cmplx (i, j)
+          a_r(i, j, k) = i
+        end if
+      end do
+    end do
+  end do
+
+  end subroutine
+
+end program
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/map-1.f90 b/libgomp/testsuite/libgomp.oacc-fortran/map-1.f90
new file mode 100644
index 00000000000..082dd8a1d6e
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/map-1.f90
@@ -0,0 +1,97 @@
+program map
+  integer, parameter     :: n = 20, c = 10
+  integer                :: i, a(n), b(n)
+
+  a(:) = 0
+  b(:) = 0
+
+  ! COPY
+
+  !$acc parallel copy (a)
+  !$acc loop
+  do i = 1, n
+     a(i) = i
+  end do
+  !$acc end parallel
+
+  do i = 1, n
+     b(i) = i
+  end do
+
+  call check (a, b, n)
+
+  ! COPYOUT
+
+  a(:) = 0
+
+  !$acc parallel copyout (a)
+  !$acc loop
+  do i = 1, n
+     a(i) = i
+  end do
+  !$acc end parallel
+
+  do i = 1, n
+     if (a(i) .ne. b(i)) call abort
+  end do
+  call check (a, b, n)
+
+  ! COPYIN
+
+  a(:) = 0
+
+  !$acc parallel copyout (a) copyin (b)
+  !$acc loop
+  do i = 1, n
+     a(i) = i
+  end do
+  !$acc end parallel
+
+  call check (a, b, n)
+
+  ! PRESENT_OR_COPY
+
+  !$acc parallel pcopy (a)
+  !$acc loop
+  do i = 1, n
+     a(i) = i
+  end do
+  !$acc end parallel
+
+  call check (a, b, n)
+
+  ! PRESENT_OR_COPYOUT
+
+  a(:) = 0
+
+  !$acc parallel pcopyout (a)
+  !$acc loop
+  do i = 1, n
+     a(i) = i
+  end do
+  !$acc end parallel
+
+  call check (a, b, n)
+
+  ! PRESENT_OR_COPYIN
+
+  a(:) = 0
+
+  !$acc parallel pcopyout (a) pcopyin (b)
+  !$acc loop
+  do i = 1, n
+     a(i) = i
+  end do
+  !$acc end parallel
+
+  call check (a, b, n)
+end program map
+
+subroutine check (a, b, n)
+  integer :: n, a(n), b(n)
+  integer :: i
+
+  do i = 1, n
+     if (a(i) .ne. b(i)) call abort
+  end do
+end subroutine check
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/openacc_version-1.f b/libgomp/testsuite/libgomp.oacc-fortran/openacc_version-1.f
new file mode 100644
index 00000000000..db3c6b18540
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/openacc_version-1.f
@@ -0,0 +1,9 @@
+! { dg-do run }
+
+      program main
+      implicit none
+      include "openacc_lib.h"
+
+      if (openacc_version .ne. 201306) call abort;
+
+      end program main
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/openacc_version-2.f90 b/libgomp/testsuite/libgomp.oacc-fortran/openacc_version-2.f90
new file mode 100644
index 00000000000..a14ecdd5032
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/openacc_version-2.f90
@@ -0,0 +1,9 @@
+! { dg-do run }
+
+program main
+  use openacc
+  implicit none
+
+  if (openacc_version .ne. 201306) call abort;
+
+end program main
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/pointer-align-1.f90 b/libgomp/testsuite/libgomp.oacc-fortran/pointer-align-1.f90
new file mode 100644
index 00000000000..a5e1fcbace4
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/pointer-align-1.f90
@@ -0,0 +1,21 @@
+! PR middle-end/63247
+
+program test
+  implicit none
+
+  integer(kind=2) a(4)
+
+  a = 10;
+
+  !$acc parallel copy(a(2:4))
+  a(2) = 52
+  a(3) = 53
+  a(4) = 54
+  !$acc end parallel
+
+  if (a(1) .ne. 10) call abort
+  if (a(2) .ne. 52) call abort
+  if (a(3) .ne. 53) call abort
+  if (a(4) .ne. 54) call abort
+
+end program test
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/pset-1.f90 b/libgomp/testsuite/libgomp.oacc-fortran/pset-1.f90
new file mode 100644
index 00000000000..1a1d4c724f8
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/pset-1.f90
@@ -0,0 +1,229 @@
+! { dg-do run }
+
+program test
+  implicit none
+  integer, allocatable :: a1(:)
+  integer, allocatable :: b1(:)
+  integer, allocatable :: c1(:)
+  integer, allocatable :: b2(:,:)
+  integer, allocatable :: c3(:,:,:)
+
+  allocate (a1(5))
+  if (.not.allocated (a1)) call abort()
+
+  a1 = 10
+
+  !$acc parallel copy(a1(1:5))
+  a1(1) = 1
+  a1(2) = 2
+  a1(3) = 3
+  a1(4) = 4
+  a1(5) = 5
+  !$acc end parallel
+
+  if (a1(1) .ne. 1) call abort
+  if (a1(2) .ne. 2) call abort
+  if (a1(3) .ne. 3) call abort
+  if (a1(4) .ne. 4) call abort
+  if (a1(5) .ne. 5) call abort
+
+  deallocate(a1)
+
+  allocate (a1(0:4))
+  if (.not.allocated (a1)) call abort()
+
+  a1 = 10
+
+  !$acc parallel copy(a1(0:4))
+  a1(0) = 1
+  a1(1) = 2
+  a1(2) = 3
+  a1(3) = 4
+  a1(4) = 5
+  !$acc end parallel
+
+  if (a1(0) .ne. 1) call abort
+  if (a1(1) .ne. 2) call abort
+  if (a1(2) .ne. 3) call abort
+  if (a1(3) .ne. 4) call abort
+  if (a1(4) .ne. 5) call abort
+
+  deallocate(a1)
+
+  allocate (b2(5,5))
+  if (.not.allocated (b2)) call abort()
+
+  b2 = 11
+
+  !$acc parallel copy(b2(1:5,1:5))
+  b2(1,1) = 1
+  b2(2,2) = 2
+  b2(3,3) = 3
+  b2(4,4) = 4
+  b2(5,5) = 5
+  !$acc end parallel
+
+  if (b2(1,1) .ne. 1) call abort
+  if (b2(2,2) .ne. 2) call abort
+  if (b2(3,3) .ne. 3) call abort
+  if (b2(4,4) .ne. 4) call abort
+  if (b2(5,5) .ne. 5) call abort
+
+  deallocate(b2)
+
+  allocate (b2(0:4,0:4))
+  if (.not.allocated (b2)) call abort()
+
+  b2 = 11
+
+  !$acc parallel copy(b2(0:4,0:4))
+  b2(0,0) = 1
+  b2(1,1) = 2
+  b2(2,2) = 3
+  b2(3,3) = 4
+  b2(4,4) = 5
+  !$acc end parallel
+
+  if (b2(0,0) .ne. 1) call abort
+  if (b2(1,1) .ne. 2) call abort
+  if (b2(2,2) .ne. 3) call abort
+  if (b2(3,3) .ne. 4) call abort
+  if (b2(4,4) .ne. 5) call abort
+
+  deallocate(b2)
+
+  allocate (c3(5,5,5))
+  if (.not.allocated (c3)) call abort()
+
+  c3 = 12
+
+  !$acc parallel copy(c3(1:5,1:5,1:5))
+  c3(1,1,1) = 1
+  c3(2,2,2) = 2
+  c3(3,3,3) = 3
+  c3(4,4,4) = 4
+  c3(5,5,5) = 5
+  !$acc end parallel
+
+  if (c3(1,1,1) .ne. 1) call abort
+  if (c3(2,2,2) .ne. 2) call abort
+  if (c3(3,3,3) .ne. 3) call abort
+  if (c3(4,4,4) .ne. 4) call abort
+  if (c3(5,5,5) .ne. 5) call abort
+
+  deallocate(c3)
+
+  allocate (c3(0:4,0:4,0:4))
+  if (.not.allocated (c3)) call abort()
+
+  c3 = 12
+
+  !$acc parallel copy(c3(0:4,0:4,0:4))
+  c3(0,0,0) = 1
+  c3(1,1,1) = 2
+  c3(2,2,2) = 3
+  c3(3,3,3) = 4
+  c3(4,4,4) = 5
+  !$acc end parallel
+
+  if (c3(0,0,0) .ne. 1) call abort
+  if (c3(1,1,1) .ne. 2) call abort
+  if (c3(2,2,2) .ne. 3) call abort
+  if (c3(3,3,3) .ne. 4) call abort
+  if (c3(4,4,4) .ne. 5) call abort
+
+  deallocate(c3)
+
+  allocate (a1(5))
+  if (.not.allocated (a1)) call abort()
+
+  allocate (b1(5))
+  if (.not.allocated (b1)) call abort()
+
+  allocate (c1(5))
+  if (.not.allocated (c1)) call abort()
+
+  a1 = 10
+  b1 = 3
+  c1 = 7
+   
+  !$acc parallel copyin(a1(1:5)) create(c1(1:5)) copyout(b1(1:5))
+  c1(1) = a1(1)
+  c1(2) = a1(2)
+  c1(3) = a1(3)
+  c1(4) = a1(4)
+  c1(5) = a1(5)
+
+  b1(1) = c1(1)
+  b1(2) = c1(2)
+  b1(3) = c1(3)
+  b1(4) = c1(4)
+  b1(5) = c1(5)
+  !$acc end parallel
+
+  if (b1(1) .ne. 10) call abort
+  if (b1(2) .ne. 10) call abort
+  if (b1(3) .ne. 10) call abort
+  if (b1(4) .ne. 10) call abort
+  if (b1(5) .ne. 10) call abort
+
+  deallocate(a1)
+  deallocate(b1)
+  deallocate(c1)
+
+  allocate (a1(0:4))
+  if (.not.allocated (a1)) call abort()
+
+  allocate (b1(0:4))
+  if (.not.allocated (b1)) call abort()
+
+  allocate (c1(0:4))
+  if (.not.allocated (c1)) call abort()
+
+  a1 = 10
+  b1 = 3
+  c1 = 7
+   
+  !$acc parallel copyin(a1(0:4)) create(c1(0:4)) copyout(b1(0:4))
+  c1(0) = a1(0)
+  c1(1) = a1(1)
+  c1(2) = a1(2)
+  c1(3) = a1(3)
+  c1(4) = a1(4)
+
+  b1(0) = c1(0)
+  b1(1) = c1(1)
+  b1(2) = c1(2)
+  b1(3) = c1(3)
+  b1(4) = c1(4)
+  !$acc end parallel
+
+  if (b1(0) .ne. 10) call abort
+  if (b1(1) .ne. 10) call abort
+  if (b1(2) .ne. 10) call abort
+  if (b1(3) .ne. 10) call abort
+  if (b1(4) .ne. 10) call abort
+
+  deallocate(a1)
+  deallocate(b1)
+  deallocate(c1)
+
+  allocate (a1(5))
+  if (.not.allocated (a1)) call abort()
+
+  a1 = 10
+
+  !$acc parallel copy(a1(2:3))
+  a1(2) = 2
+  a1(3) = 3
+  !$acc end parallel
+
+  if (a1(1) .ne. 10) call abort
+  if (a1(2) .ne. 2) call abort
+  if (a1(3) .ne. 3) call abort
+  if (a1(4) .ne. 10) call abort
+  if (a1(5) .ne. 10) call abort
+
+  deallocate(a1)
+
+end program test
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/reduction-1.f90 b/libgomp/testsuite/libgomp.oacc-fortran/reduction-1.f90
new file mode 100644
index 00000000000..89e7fe7fe58
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/reduction-1.f90
@@ -0,0 +1,225 @@
+! { dg-do run }
+
+! Integer reductions
+
+program reduction_1
+  implicit none
+
+  integer, parameter    :: n = 10, vl = 2
+  integer               :: i, vresult, result
+  logical               :: lresult, lvresult
+  integer, dimension (n) :: array
+
+  do i = 1, n
+     array(i) = i
+  end do
+
+  result = 0
+  vresult = 0
+
+  ! '+' reductions
+
+  !$acc parallel vector_length(vl) num_gangs(1)
+  !$acc loop reduction(+:result)
+  do i = 1, n
+     result = result + array(i)
+  end do
+  !$acc end parallel
+
+  ! Verify the results
+  do i = 1, n
+     vresult = vresult + array(i)
+  end do
+
+  if (result.ne.vresult) call abort
+
+  result = 0
+  vresult = 0
+
+  ! '*' reductions
+
+  !$acc parallel vector_length(vl) num_gangs(1)
+  !$acc loop reduction(*:result)
+  do i = 1, n
+     result = result * array(i)
+  end do
+  !$acc end parallel
+
+  ! Verify the results
+  do i = 1, n
+     vresult = vresult * array(i)
+  end do
+
+  if (result.ne.vresult) call abort
+
+  result = 0
+  vresult = 0
+
+  ! 'max' reductions
+
+  !$acc parallel vector_length(vl) num_gangs(1)
+  !$acc loop reduction(max:result)
+  do i = 1, n
+     result = max (result, array(i))
+  end do
+  !$acc end parallel
+
+  ! Verify the results
+  do i = 1, n
+     vresult = max (vresult, array(i))
+  end do
+
+  if (result.ne.vresult) call abort
+
+  result = 1
+  vresult = 1
+
+  ! 'min' reductions
+
+  !$acc parallel vector_length(vl) num_gangs(1)
+  !$acc loop reduction(min:result)
+  do i = 1, n
+     result = min (result, array(i))
+  end do
+  !$acc end parallel
+
+  ! Verify the results
+  do i = 1, n
+     vresult = min (vresult, array(i))
+  end do
+
+  if (result.ne.vresult) call abort
+
+  result = 1
+  vresult = 1
+
+  ! 'iand' reductions
+
+  !$acc parallel vector_length(vl) num_gangs(1)
+  !$acc loop reduction(iand:result)
+  do i = 1, n
+     result = iand (result, array(i))
+  end do
+  !$acc end parallel
+
+  ! Verify the results
+  do i = 1, n
+     vresult = iand (vresult, array(i))
+  end do
+
+  if (result.ne.vresult) call abort
+
+  result = 1
+  vresult = 1
+
+  ! 'ior' reductions
+
+  !$acc parallel vector_length(vl) num_gangs(1)
+  !$acc loop reduction(ior:result)
+  do i = 1, n
+     result = ior (result, array(i))
+  end do
+  !$acc end parallel
+
+  ! Verify the results
+  do i = 1, n
+     vresult = ior (vresult, array(i))
+  end do
+
+  if (result.ne.vresult) call abort
+
+  result = 0
+  vresult = 0
+
+  ! 'ieor' reductions
+
+  !$acc parallel vector_length(vl) num_gangs(1)
+  !$acc loop reduction(ieor:result)
+  do i = 1, n
+     result = ieor (result, array(i))
+  end do
+  !$acc end parallel
+
+  ! Verify the results
+  do i = 1, n
+     vresult = ieor (vresult, array(i))
+  end do
+
+  if (result.ne.vresult) call abort
+
+  lresult = .false.
+  lvresult = .false.
+
+  ! '.and.' reductions
+
+  !$acc parallel vector_length(vl) num_gangs(1)
+  !$acc loop reduction(.and.:lresult)
+  do i = 1, n
+     lresult = lresult .and. (array(i) .ge. 5)
+  end do
+  !$acc end parallel
+
+  ! Verify the results
+  do i = 1, n
+     lvresult = lvresult .and. (array(i) .ge. 5)
+  end do
+
+  if (result.ne.vresult) call abort
+
+  lresult = .false.
+  lvresult = .false.
+
+  ! '.or.' reductions
+
+  !$acc parallel vector_length(vl) num_gangs(1)
+  !$acc loop reduction(.or.:lresult)
+  do i = 1, n
+     lresult = lresult .or. (array(i) .ge. 5)
+  end do
+  !$acc end parallel
+
+  ! Verify the results
+  do i = 1, n
+     lvresult = lvresult .or. (array(i) .ge. 5)
+  end do
+
+  if (result.ne.vresult) call abort
+
+  lresult = .false.
+  lvresult = .false.
+
+  ! '.eqv.' reductions
+
+  !$acc parallel vector_length(vl) num_gangs(1)
+  !$acc loop reduction(.eqv.:lresult)
+  do i = 1, n
+     lresult = lresult .eqv. (array(i) .ge. 5)
+  end do
+  !$acc end parallel
+
+  ! Verify the results
+  do i = 1, n
+     lvresult = lvresult .eqv. (array(i) .ge. 5)
+  end do
+
+  if (result.ne.vresult) call abort
+
+  lresult = .false.
+  lvresult = .false.
+
+  ! '.neqv.' reductions
+
+  !$acc parallel vector_length(vl) num_gangs(1)
+  !$acc loop reduction(.neqv.:lresult)
+  do i = 1, n
+     lresult = lresult .neqv. (array(i) .ge. 5)
+  end do
+  !$acc end parallel
+
+  ! Verify the results
+  do i = 1, n
+     lvresult = lvresult .neqv. (array(i) .ge. 5)
+  end do
+
+  if (result.ne.vresult) call abort
+end program reduction_1
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/reduction-2.f90 b/libgomp/testsuite/libgomp.oacc-fortran/reduction-2.f90
new file mode 100644
index 00000000000..d3659c9fc3e
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/reduction-2.f90
@@ -0,0 +1,170 @@
+! { dg-do run }
+
+! real reductions
+
+program reduction_2
+  implicit none
+
+  integer, parameter    :: n = 10, vl = 2
+  integer               :: i
+  real, parameter       :: e = .001
+  real                  :: vresult, result
+  logical               :: lresult, lvresult
+  real, dimension (n) :: array
+
+  do i = 1, n
+     array(i) = i
+  end do
+
+  result = 0
+  vresult = 0
+
+  ! '+' reductions
+
+  !$acc parallel vector_length(vl) num_gangs(1)
+  !$acc loop reduction(+:result)
+  do i = 1, n
+     result = result + array(i)
+  end do
+  !$acc end parallel
+
+  ! Verify the results
+  do i = 1, n
+     vresult = vresult + array(i)
+  end do
+
+  if (abs (result - vresult) .ge. e) call abort
+
+  result = 1
+  vresult = 1
+
+  ! '*' reductions
+
+  !$acc parallel vector_length(vl) num_gangs(1)
+  !$acc loop reduction(*:result)
+  do i = 1, n
+     result = result * array(i)
+  end do
+  !$acc end parallel
+
+  ! Verify the results
+  do i = 1, n
+     vresult = vresult * array(i)
+  end do
+
+  if (result.ne.vresult) call abort
+
+  result = 0
+  vresult = 0
+
+  ! 'max' reductions
+
+  !$acc parallel vector_length(vl) num_gangs(1)
+  !$acc loop reduction(max:result)
+  do i = 1, n
+     result = max (result, array(i))
+  end do
+  !$acc end parallel
+
+  ! Verify the results
+  do i = 1, n
+     vresult = max (vresult, array(i))
+  end do
+
+  if (result.ne.vresult) call abort
+
+  result = 1
+  vresult = 1
+
+  ! 'min' reductions
+
+  !$acc parallel vector_length(vl) num_gangs(1)
+  !$acc loop reduction(min:result)
+  do i = 1, n
+     result = min (result, array(i))
+  end do
+  !$acc end parallel
+
+  ! Verify the results
+  do i = 1, n
+     vresult = min (vresult, array(i))
+  end do
+
+  if (result.ne.vresult) call abort
+
+  result = 1
+  vresult = 1
+
+  ! '.and.' reductions
+
+  !$acc parallel vector_length(vl) num_gangs(1)
+  !$acc loop reduction(.and.:lresult)
+  do i = 1, n
+     lresult = lresult .and. (array(i) .ge. 5)
+  end do
+  !$acc end parallel
+
+  ! Verify the results
+  do i = 1, n
+     lvresult = lvresult .and. (array(i) .ge. 5)
+  end do
+
+  if (result.ne.vresult) call abort
+
+  lresult = .false.
+  lvresult = .false.
+
+  ! '.or.' reductions
+
+  !$acc parallel vector_length(vl) num_gangs(1)
+  !$acc loop reduction(.or.:lresult)
+  do i = 1, n
+     lresult = lresult .or. (array(i) .ge. 5)
+  end do
+  !$acc end parallel
+
+  ! Verify the results
+  do i = 1, n
+     lvresult = lvresult .or. (array(i) .ge. 5)
+  end do
+
+  if (result.ne.vresult) call abort
+
+  lresult = .false.
+  lvresult = .false.
+
+  ! '.eqv.' reductions
+
+  !$acc parallel vector_length(vl) num_gangs(1)
+  !$acc loop reduction(.eqv.:lresult)
+  do i = 1, n
+     lresult = lresult .eqv. (array(i) .ge. 5)
+  end do
+  !$acc end parallel
+
+  ! Verify the results
+  do i = 1, n
+     lvresult = lvresult .eqv. (array(i) .ge. 5)
+  end do
+
+  if (result.ne.vresult) call abort
+
+  lresult = .false.
+  lvresult = .false.
+
+  ! '.neqv.' reductions
+
+  !$acc parallel vector_length(vl) num_gangs(1)
+  !$acc loop reduction(.neqv.:lresult)
+  do i = 1, n
+     lresult = lresult .neqv. (array(i) .ge. 5)
+  end do
+  !$acc end parallel
+
+  ! Verify the results
+  do i = 1, n
+     lvresult = lvresult .neqv. (array(i) .ge. 5)
+  end do
+
+  if (result.ne.vresult) call abort
+end program reduction_2
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/reduction-3.f90 b/libgomp/testsuite/libgomp.oacc-fortran/reduction-3.f90
new file mode 100644
index 00000000000..2b8005dac15
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/reduction-3.f90
@@ -0,0 +1,170 @@
+! { dg-do run }
+
+! double precision reductions
+
+program reduction_3
+  implicit none
+
+  integer, parameter    :: n = 10, vl = 2
+  integer               :: i
+  double precision, parameter :: e = .001
+  double precision      :: vresult, result
+  logical               :: lresult, lvresult
+  double precision, dimension (n) :: array
+
+  do i = 1, n
+     array(i) = i
+  end do
+
+  result = 0
+  vresult = 0
+
+  ! '+' reductions
+
+  !$acc parallel vector_length(vl) num_gangs(1)
+  !$acc loop reduction(+:result)
+  do i = 1, n
+     result = result + array(i)
+  end do
+  !$acc end parallel
+
+  ! Verify the results
+  do i = 1, n
+     vresult = vresult + array(i)
+  end do
+
+  if (abs (result - vresult) .ge. e) call abort
+
+  result = 1
+  vresult = 1
+
+  ! '*' reductions
+
+  !$acc parallel vector_length(vl) num_gangs(1)
+  !$acc loop reduction(*:result)
+  do i = 1, n
+     result = result * array(i)
+  end do
+  !$acc end parallel
+
+  ! Verify the results
+  do i = 1, n
+     vresult = vresult * array(i)
+  end do
+
+  if (result.ne.vresult) call abort
+
+  result = 0
+  vresult = 0
+
+  ! 'max' reductions
+
+  !$acc parallel vector_length(vl) num_gangs(1)
+  !$acc loop reduction(max:result)
+  do i = 1, n
+     result = max (result, array(i))
+  end do
+  !$acc end parallel
+
+  ! Verify the results
+  do i = 1, n
+     vresult = max (vresult, array(i))
+  end do
+
+  if (result.ne.vresult) call abort
+
+  result = 1
+  vresult = 1
+
+  ! 'min' reductions
+
+  !$acc parallel vector_length(vl) num_gangs(1)
+  !$acc loop reduction(min:result)
+  do i = 1, n
+     result = min (result, array(i))
+  end do
+  !$acc end parallel
+
+  ! Verify the results
+  do i = 1, n
+     vresult = min (vresult, array(i))
+  end do
+
+  if (result.ne.vresult) call abort
+
+  result = 1
+  vresult = 1
+
+  ! '.and.' reductions
+
+  !$acc parallel vector_length(vl) num_gangs(1)
+  !$acc loop reduction(.and.:lresult)
+  do i = 1, n
+     lresult = lresult .and. (array(i) .ge. 5)
+  end do
+  !$acc end parallel
+
+  ! Verify the results
+  do i = 1, n
+     lvresult = lvresult .and. (array(i) .ge. 5)
+  end do
+
+  if (result.ne.vresult) call abort
+
+  lresult = .false.
+  lvresult = .false.
+
+  ! '.or.' reductions
+
+  !$acc parallel vector_length(vl) num_gangs(1)
+  !$acc loop reduction(.or.:lresult)
+  do i = 1, n
+     lresult = lresult .or. (array(i) .ge. 5)
+  end do
+  !$acc end parallel
+
+  ! Verify the results
+  do i = 1, n
+     lvresult = lvresult .or. (array(i) .ge. 5)
+  end do
+
+  if (result.ne.vresult) call abort
+
+  lresult = .false.
+  lvresult = .false.
+
+  ! '.eqv.' reductions
+
+  !$acc parallel vector_length(vl) num_gangs(1)
+  !$acc loop reduction(.eqv.:lresult)
+  do i = 1, n
+     lresult = lresult .eqv. (array(i) .ge. 5)
+  end do
+  !$acc end parallel
+
+  ! Verify the results
+  do i = 1, n
+     lvresult = lvresult .eqv. (array(i) .ge. 5)
+  end do
+
+  if (result.ne.vresult) call abort
+
+  lresult = .false.
+  lvresult = .false.
+
+  ! '.neqv.' reductions
+
+  !$acc parallel vector_length(vl) num_gangs(1)
+  !$acc loop reduction(.neqv.:lresult)
+  do i = 1, n
+     lresult = lresult .neqv. (array(i) .ge. 5)
+  end do
+  !$acc end parallel
+
+  ! Verify the results
+  do i = 1, n
+     lvresult = lvresult .neqv. (array(i) .ge. 5)
+  end do
+
+  if (result.ne.vresult) call abort
+end program reduction_3
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/reduction-4.f90 b/libgomp/testsuite/libgomp.oacc-fortran/reduction-4.f90
new file mode 100644
index 00000000000..12f7a33053c
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/reduction-4.f90
@@ -0,0 +1,54 @@
+! { dg-do run }
+
+! complex reductions
+
+program reduction_4
+  implicit none
+
+  integer, parameter    :: n = 10, vl = 32
+  integer               :: i
+  complex               :: vresult, result
+  complex, dimension (n) :: array
+
+  do i = 1, n
+     array(i) = i
+  end do
+
+  result = 0
+  vresult = 0
+
+  ! '+' reductions
+
+  !$acc parallel vector_length(vl) num_gangs(1)
+  !$acc loop reduction(+:result)
+  do i = 1, n
+     result = result + array(i)
+  end do
+  !$acc end parallel
+
+  ! Verify the results
+  do i = 1, n
+     vresult = vresult + array(i)
+  end do
+
+  if (result .ne. vresult) call abort
+
+  result = 1
+  vresult = 1
+
+!  ! '*' reductions
+!
+!  !$acc parallel vector_length(vl)
+!  !$acc loop reduction(*:result)
+!  do i = 1, n
+!     result = result * array(i)
+!  end do
+!  !$acc end parallel
+!
+!  ! Verify the results
+!  do i = 1, n
+!     vresult = vresult * array(i)
+!  end do
+!
+!  if (result.ne.vresult) call abort
+end program reduction_4
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/reduction-5.f90 b/libgomp/testsuite/libgomp.oacc-fortran/reduction-5.f90
new file mode 100644
index 00000000000..df44a7ae69d
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/reduction-5.f90
@@ -0,0 +1,32 @@
+! { dg-do run }
+
+! subroutine reduction
+
+program reduction
+  integer, parameter    :: n = 40, c = 10
+  integer               :: i, vsum, sum
+
+  call redsub (sum, n, c)
+
+  vsum = 0
+
+  ! Verify the results
+  do i = 1, n
+     vsum = vsum + c
+  end do
+
+  if (sum.ne.vsum) call abort ()
+end program reduction
+
+subroutine redsub(sum, n, c)
+  integer :: sum, n, c
+
+  sum = 0
+
+  !$acc parallel vector_length(n) copyin (n, c) num_gangs(1)
+  !$acc loop reduction(+:sum)
+  do i = 1, n
+     sum = sum + c
+  end do
+  !$acc end parallel
+end subroutine redsub
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/reduction-6.f90 b/libgomp/testsuite/libgomp.oacc-fortran/reduction-6.f90
new file mode 100644
index 00000000000..63254318954
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/reduction-6.f90
@@ -0,0 +1,30 @@
+! { dg-do run }
+
+program reduction
+  implicit none
+
+  integer, parameter    :: n = 100
+  integer               :: i, s1, s2, vs1, vs2
+
+  s1 = 0
+  s2 = 0
+  vs1 = 0
+  vs2 = 0
+
+  !$acc parallel vector_length (1000)
+  !$acc loop reduction(+:s1, s2)
+  do i = 1, n
+     s1 = s1 + 1
+     s2 = s2 + 2
+  end do
+  !$acc end parallel
+
+  ! Verify the results
+  do i = 1, n
+     vs1 = vs1 + 1
+     vs2 = vs2 + 2
+  end do
+
+  if (s1.ne.vs1) call abort ()
+  if (s2.ne.vs2) call abort ()
+end program reduction
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/routine-1.f90 b/libgomp/testsuite/libgomp.oacc-fortran/routine-1.f90
new file mode 100644
index 00000000000..33905157d49
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/routine-1.f90
@@ -0,0 +1,32 @@
+! { dg-do run }
+! { dg-options "-fno-inline" }
+
+  interface
+    recursive function fact (x)
+      !$acc routine
+      integer, intent(in) :: x
+      integer :: fact
+    end function fact
+  end interface
+  integer, parameter :: n = 10
+  integer :: a(n), i
+  !$acc parallel
+  !$acc loop
+  do i = 1, n
+     a(i) = fact (i)
+  end do
+  !$acc end parallel
+  do i = 1, n
+     if (a(i) .ne. fact(i)) call abort
+  end do
+end
+recursive function fact (x) result (res)
+  !$acc routine
+  integer, intent(in) :: x
+  integer :: res
+  if (x < 1) then
+     res = 1
+  else
+     res = x * fact (x - 1)
+  end if
+end function fact
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/routine-2.f90 b/libgomp/testsuite/libgomp.oacc-fortran/routine-2.f90
new file mode 100644
index 00000000000..3d418b660c5
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/routine-2.f90
@@ -0,0 +1,29 @@
+! { dg-do run }
+! { dg-options "-fno-inline" }
+
+  module m1
+    contains
+    recursive function fact (x) result (res)
+      !$acc routine
+      integer, intent(in) :: x
+      integer :: res
+      if (x < 1) then
+         res = 1
+      else
+         res = x * fact (x - 1)
+      end if
+    end function fact
+  end module m1
+  use m1
+  integer, parameter :: n = 10
+  integer :: a(n), i
+  !$acc parallel
+  !$acc loop
+  do i = 1, n
+     a(i) = fact (i)
+  end do
+  !$acc end parallel
+  do i = 1, n
+     if (a(i) .ne. fact(i)) call abort
+  end do
+end
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/routine-3.f90 b/libgomp/testsuite/libgomp.oacc-fortran/routine-3.f90
new file mode 100644
index 00000000000..d233a63ee5a
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/routine-3.f90
@@ -0,0 +1,27 @@
+! { dg-do run }
+! { dg-options "-fno-inline" }
+
+  integer, parameter :: n = 10
+  integer :: a(n), i
+  integer, external :: fact
+  !$acc routine (fact)
+  !$acc parallel
+  !$acc loop
+  do i = 1, n
+     a(i) = fact (i)
+  end do
+  !$acc end parallel
+  do i = 1, n
+     if (a(i) .ne. fact(i)) call abort
+  end do
+end
+recursive function fact (x) result (res)
+  !$acc routine
+  integer, intent(in) :: x
+  integer :: res
+  if (x < 1) then
+     res = 1
+  else
+     res = x * fact (x - 1)
+  end if
+end function fact
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/routine-4.f90 b/libgomp/testsuite/libgomp.oacc-fortran/routine-4.f90
new file mode 100644
index 00000000000..3e5fb09c45c
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/routine-4.f90
@@ -0,0 +1,23 @@
+! { dg-do run }
+! { dg-options "-fno-inline" }
+
+  integer, parameter :: n = 10
+  integer :: a(n), i
+  do i = 1, n
+     a(i) = i
+  end do
+  !$acc parallel
+  !$acc loop
+  do i = 1, n
+     call incr(a(i))
+  end do
+  !$acc end parallel
+  do i = 1, n
+     if (a(i) .ne. (i + 1)) call abort
+  end do
+end
+subroutine incr (x)
+  !$acc routine
+  integer, intent(inout) :: x
+  x = x + 1
+end subroutine incr
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/subarrays-1.f90 b/libgomp/testsuite/libgomp.oacc-fortran/subarrays-1.f90
new file mode 100644
index 00000000000..b39414fe3e5
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/subarrays-1.f90
@@ -0,0 +1,97 @@
+program subarrays
+  integer, parameter     :: n = 20, c = 10
+  integer                :: i, a(n), b(n)
+
+  a(:) = 0
+  b(:) = 0
+
+  ! COPY
+
+  !$acc parallel copy (a(1:n))
+  !$acc loop
+  do i = 1, n
+     a(i) = i
+  end do
+  !$acc end parallel
+
+  do i = 1, n
+     b(i) = i
+  end do
+
+  call check (a, b, n)
+
+  ! COPYOUT
+
+  a(:) = 0
+
+  !$acc parallel copyout (a(1:n))
+  !$acc loop
+  do i = 1, n
+     a(i) = i
+  end do
+  !$acc end parallel
+
+  do i = 1, n
+     if (a(i) .ne. b(i)) call abort
+  end do
+  call check (a, b, n)
+
+  ! COPYIN
+
+  a(:) = 0
+
+  !$acc parallel copyout (a(1:n)) copyin (b(1:n))
+  !$acc loop
+  do i = 1, n
+     a(i) = i
+  end do
+  !$acc end parallel
+
+  call check (a, b, n)
+
+  ! PRESENT_OR_COPY
+
+  !$acc parallel pcopy (a(1:n))
+  !$acc loop
+  do i = 1, n
+     a(i) = i
+  end do
+  !$acc end parallel
+
+  call check (a, b, n)
+
+  ! PRESENT_OR_COPYOUT
+
+  a(:) = 0
+
+  !$acc parallel pcopyout (a(1:n))
+  !$acc loop
+  do i = 1, n
+     a(i) = i
+  end do
+  !$acc end parallel
+
+  call check (a, b, n)
+
+  ! PRESENT_OR_COPYIN
+
+  a(:) = 0
+
+  !$acc parallel pcopyout (a(1:n)) pcopyin (b(1:n))
+  !$acc loop
+  do i = 1, n
+     a(i) = i
+  end do
+  !$acc end parallel
+
+  call check (a, b, n)
+end program subarrays
+
+subroutine check (a, b, n)
+  integer :: n, a(n), b(n)
+  integer :: i
+
+  do i = 1, n
+     if (a(i) .ne. b(i)) call abort
+  end do
+end subroutine check
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/subarrays-2.f90 b/libgomp/testsuite/libgomp.oacc-fortran/subarrays-2.f90
new file mode 100644
index 00000000000..81799f60733
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/subarrays-2.f90
@@ -0,0 +1,100 @@
+program subarrays
+  integer, parameter     :: n = 20, c = 10, low = 5, high = 10
+  integer                :: i, a(n), b(n)
+
+  a(:) = 0
+  b(:) = 0
+
+  ! COPY
+
+  !$acc parallel copy (a(low:high))
+  !$acc loop
+  do i = low, high
+     a(i) = i
+  end do
+  !$acc end parallel
+
+  do i = low, high
+     b(i) = i
+  end do
+
+  call check (a, b, n)
+
+  ! COPYOUT
+
+  a(:) = 0
+
+  !$acc parallel copyout (a(low:high))
+  !$acc loop
+  do i = low, high
+     a(i) = i
+  end do
+  !$acc end parallel
+
+  do i = low, high
+     if (a(i) .ne. b(i)) call abort
+  end do
+  call check (a, b, n)
+
+  ! COPYIN
+
+  a(:) = 0
+
+  !$acc parallel copyout (a(low:high)) copyin (b(low:high))
+  !$acc loop
+  do i = low, high
+     a(i) = b(i)
+  end do
+  !$acc end parallel
+
+  call check (a, b, n)
+
+  ! PRESENT_OR_COPY
+
+  a(:) = 0
+  
+  !$acc parallel pcopy (a(low:high))
+  !$acc loop
+  do i = low, high
+     a(i) = i
+  end do
+  !$acc end parallel
+
+  call check (a, b, n)
+
+  ! PRESENT_OR_COPYOUT
+
+  a(:) = 0
+
+  !$acc parallel pcopyout (a(low:high))
+  !$acc loop
+  do i = low, high
+     a(i) = i
+  end do
+  !$acc end parallel
+
+  call check (a, b, n)
+
+  ! PRESENT_OR_COPYIN
+
+  a(:) = 0
+
+  !$acc parallel pcopyout (a(low:high)) &
+  !$acc & pcopyin (b(low:high))
+  !$acc loop
+  do i = low, high
+     a(i) = b(i)
+  end do
+  !$acc end parallel
+
+  call check (a, b, n)
+end program subarrays
+
+subroutine check (a, b, n)
+  integer :: n, a(n), b(n)
+  integer :: i
+
+  do i = 1, n
+     if (a(i) .ne. b(i)) call abort
+  end do
+end subroutine check