libgomp/

	* target.c: New file.
	* Makefile.am (libgomp_la_SOURCES): Add target.c.
	* Makefile.in: Regenerated.
	* libgomp_g.h (GOMP_task): Add depend argument.
	(GOMP_barrier_cancel, GOMP_loop_end_cancel,
	GOMP_sections_end_cancel, GOMP_target, GOMP_target_data,
	GOMP_target_end_data, GOMP_target_update, GOMP_teams,
	GOMP_parallel_loop_static, GOMP_parallel_loop_dynamic,
	GOMP_parallel_loop_guided, GOMP_parallel_loop_runtime,
	GOMP_parallel, GOMP_cancel, GOMP_cancellation_point,
	GOMP_taskgroup_start, GOMP_taskgroup_end,
	GOMP_parallel_sections): New prototypes.
	* fortran.c (omp_is_initial_device): Add ialias_redirect.
	(omp_is_initial_device_): New function.
	(ULP, STR1, STR2, ialias_redirect): Removed.
	(omp_get_cancellation_, omp_get_proc_bind_, omp_set_default_device_,
	omp_set_default_device_8_, omp_get_default_device_,
	omp_get_num_devices_, omp_get_num_teams_, omp_get_team_num_): New
	functions.
	* libgomp.map (GOMP_barrier_cancel, GOMP_loop_end_cancel,
	GOMP_sections_end_cancel, GOMP_target, GOMP_target_data,
	GOMP_target_end_data, GOMP_target_update, GOMP_teams): Export
	@@GOMP_4.0.
	(omp_is_initial_device, omp_is_initial_device_, omp_get_cancellation,
	omp_get_cancellation_, omp_get_proc_bind, omp_get_proc_bind_,
	omp_set_default_device, omp_set_default_device_,
	omp_set_default_device_8_, omp_get_default_device,
	omp_get_default_device_, omp_get_num_devices, omp_get_num_devices_,
	omp_get_num_teams, omp_get_num_teams_, omp_get_team_num,
	omp_get_team_num_): Export @@OMP_4.0.
	* team.c (struct gomp_thread_start_data): Add place field.
	(gomp_thread_start): Clear thr->thread_pool and
	thr->task before returning.  Use gomp_team_barrier_wait_final
	instead of gomp_team_barrier_wait.  Initialize thr->place.
	(gomp_new_team): Initialize work_shares_to_free, work_share_cancelled,
	team_cancelled and task_queued_count fields.
	(gomp_free_pool_helper): Clear thr->thread_pool and thr->task
	before calling pthread_exit.
	(gomp_free_thread): No longer static.  Use
	gomp_managed_threads_lock instead of gomp_remaining_threads_lock.
	(gomp_team_start): Add flags argument.  Set
	thr->thread_pool->threads_busy to nthreads immediately after creating
	new pool.  Use gomp_managed_threads_lock instead of
	gomp_remaining_threads_lock.  Handle OpenMP 4.0 affinity.
	(gomp_team_end): Use gomp_managed_threads_lock instead of
	gomp_remaining_threads_lock.  Use gomp_team_barrier_wait_final instead
	of gomp_team_barrier_wait.  If team->team_cancelled, call
	gomp_fini_worshare on ws chain starting at team->work_shares_to_free
	rather than thr->ts.work_share.
	(initialize_team): Don't call gomp_sem_init here.
	* sections.c (GOMP_parallel_sections_start): Adjust gomp_team_start
	caller.
	(GOMP_parallel_sections, GOMP_sections_end_cancel): New functions.
	* env.c (gomp_global_icv): Add default_device_var, target_data and
	bind_var initializers.
	(gomp_cpu_affinity, gomp_cpu_affinity_len): Remove.
	(gomp_bind_var_list, gomp_bind_var_list_len, gomp_places_list,
	gomp_places_list_len): New variables.
	(parse_bind_var, parse_one_place, parse_places_var): New functions.
	(parse_affinity): Rewritten to construct OMP_PLACES list with unit
	sized places.
	(gomp_cancel_var): New global variable.
	(parse_int): New function.
	(handle_omp_display_env): New function.
	(initialize_env): Use it.  Initialize default_device_var.
	Parse OMP_CANCELLATION env var.  Use parse_bind_var to parse
	OMP_PROC_BIND instead of parse_boolean.  Use parse_places_var for
	OMP_PLACES parsing.  Don't call parse_affinity if OMP_PLACES has
	been successfully parsed (and call gomp_init_affinity in that case).
	(omp_get_cancellation, omp_get_proc_bind, omp_set_default_device,
	omp_get_default_device, omp_get_num_devices, omp_get_num_teams,
	omp_get_team_num, omp_is_initial_device): New functions.
	* libgomp.h: Include stdlib.h.
	(ialias_ulp, ialias_str1, ialias_str2, ialias_redirect, ialias_call):
	Define.
	(struct target_mem_desc): Forward declare.
	(struct gomp_task_icv): Add default_device_var, target_data, bind_var
	and thread_limit_var fields.
	(gomp_get_num_devices): New prototype.
	(gomp_cancel_var): New extern decl.
	(struct gomp_team): Add work_shares_to_free, work_share_cancelled,
	team_cancelled and task_queued_count fields.  Add comments about
	task_{,queued_,running_}count.
	(gomp_cancel_kind): New enum.
	(gomp_work_share_end_cancel): New prototype.
	(struct gomp_task): Add next_taskgroup, prev_taskgroup, taskgroup,
	copy_ctors_done, dependers, depend_hash, depend_count, num_dependees
	and depend fields.
	(struct gomp_taskgroup): New type.
	(struct gomp_task_depend_entry,
	struct gomp_dependers_vec): New types.
	(gomp_finish_task): Free depend_hash if non-NULL.
	(struct gomp_team_state): Add place_partition_off
	and place_partition_len fields.
	(gomp_bind_var_list, gomp_bind_var_list_len, gomp_places_list,
	gomp_places_list_len): New extern decls.
	(struct gomp_thread): Add place field.
	(gomp_cpu_affinity, gomp_cpu_affinity_len): Remove.
	(gomp_init_thread_affinity): Add place argument.
	(gomp_affinity_alloc, gomp_affinity_init_place, gomp_affinity_add_cpus,
	gomp_affinity_remove_cpu, gomp_affinity_copy_place,
	gomp_affinity_same_place, gomp_affinity_finalize_place_list,
	gomp_affinity_init_level, gomp_affinity_print_place): New
	prototypes.
	(gomp_team_start): Add flags argument.
	(gomp_thread_limit_var, gomp_remaining_threads_count,
	gomp_remaining_threads_lock): Remove.
	(gomp_managed_threads_lock): New variable.
	(struct gomp_thread_pool): Add threads_busy field.
	(gomp_free_thread): New prototype.
	* task.c: Include hashtab.h.
	(hash_entry_type): New typedef.
	(htab_alloc, htab_free, htab_hash, htab_eq): New inlines.
	(gomp_init_task): Clear dependers, depend_hash, depend_count,
	copy_ctors_done and taskgroup fields.
	(GOMP_task): Add depend argument, handle depend clauses.  If
	gomp_team_barrier_cancelled or if it's taskgroup has been
	cancelled, don't queue or start new tasks.  Set copy_ctors_done
	field if needed.  Initialize taskgroup field.  If copy_ctors_done
	and already cancelled, don't discard the task.  If taskgroup is
	non-NULL, enqueue the task into taskgroup queue.  Increment
	num_children field in taskgroup.  Increment task_queued_count.
	(gomp_task_run_pre, gomp_task_run_post_remove_parent,
	gomp_task_run_post_remove_taskgroup): New inline functions.
	(gomp_task_run_post_handle_depend_hash,
	gomp_task_run_post_handle_dependers,
	gomp_task_run_post_handle_depend): New functions.
	(GOMP_taskwait): Use them.  If more than one new tasks
	have been queued, wake other threads if needed.
	(gomp_barrier_handle_tasks): Likewise.  If
	gomp_team_barrier_cancelled, don't start any new tasks, just free
	all tasks.
	(GOMP_taskgroup_start, GOMP_taskgroup_end): New functions.
	* omp_lib.f90.in
	(omp_proc_bind_kind, omp_proc_bind_false,
	omp_proc_bind_true, omp_proc_bind_master, omp_proc_bind_close,
	omp_proc_bind_spread): New params.
	(omp_get_cancellation, omp_get_proc_bind, omp_set_default_device,
	omp_get_default_device, omp_get_num_devices, omp_get_num_teams,
	omp_get_team_num, omp_is_initial_device): New interfaces.
	(omp_get_dynamic, omp_get_nested, omp_in_parallel,
	omp_get_max_threads, omp_get_num_procs, omp_get_num_threads,
	omp_get_thread_num, omp_get_thread_limit, omp_set_max_active_levels,
	omp_get_max_active_levels, omp_get_level, omp_get_ancestor_thread_num,
	omp_get_team_size, omp_get_active_level, omp_in_final): Remove
	useless use omp_lib_kinds.
	* omp.h.in (omp_proc_bind_t): New typedef.
	(omp_get_cancellation, omp_get_proc_bind, omp_set_default_device,
	omp_get_default_device, omp_get_num_devices, omp_get_num_teams,
	omp_get_team_num, omp_is_initial_device): New prototypes.
	* loop.c (gomp_parallel_loop_start): Add flags argument, pass it
	through to gomp_team_start.
	(GOMP_parallel_loop_static_start, GOMP_parallel_loop_dynamic_start,
	GOMP_parallel_loop_guided_start, GOMP_parallel_loop_runtime_start):
	Adjust gomp_parallel_loop_start callers.
	(GOMP_parallel_loop_static, GOMP_parallel_loop_dynamic,
	GOMP_parallel_loop_guided, GOMP_parallel_loop_runtime,
	GOMP_loop_end_cancel): New functions.
	(GOMP_parallel_end): Add ialias_redirect.
	* hashtab.h: New file.
	* libgomp.texi (Environment Variables): Minor cleanup,
	update section refs to OpenMP 4.0rc2.
	(OMP_DISPLAY_ENV, GOMP_SPINCOUNT): Document these
	environment variables.
	* work.c (gomp_work_share_end, gomp_work_share_end_nowait): Set
	team->work_shares_to_free to thr->ts.work_share before calling
	free_work_share.
	(gomp_work_share_end_cancel): New function.
	* config/linux/proc.c: Include errno.h.
	(gomp_get_cpuset_size, gomp_cpuset_size, gomp_cpusetp): New variables.
	(gomp_cpuset_popcount): Add cpusetsize argument, use it instead of
	sizeof (cpu_set_t) to determine number of iterations.  Fix up check
	extern decl.  Use CPU_COUNT_S if available, or CPU_COUNT if
	gomp_cpuset_size is sizeof (cpu_set_t).
	(gomp_init_num_threads): Initialize gomp_cpuset_size,
	gomp_get_cpuset_size and gomp_cpusetp here, use gomp_cpusetp instead
	of &cpuset and pass gomp_cpuset_size instead of sizeof (cpu_set_t)
	to pthread_getaffinity_np.  Free and clear gomp_cpusetp if it didn't
	contain any logical CPUs.
	(get_num_procs): Don't call pthread_getaffinity_np if gomp_cpusetp
	is NULL.  Use gomp_cpusetp instead of &cpuset and pass
	gomp_get_cpuset_size instead of sizeof (cpu_set_t) to
	pthread_getaffinity_np.  Check gomp_places_list instead of
	gomp_cpu_affinity.  Adjust gomp_cpuset_popcount caller.
	* config/linux/bar.c (gomp_barrier_wait_end,
	gomp_barrier_wait_last): Use BAR_* defines.
	(gomp_team_barrier_wait_end): Likewise.  Clear BAR_CANCELLED
	from state where needed.  Set work_share_cancelled to 0 on last
	thread.
	(gomp_team_barrier_wait_final, gomp_team_barrier_wait_cancel_end,
	gomp_team_barrier_wait_cancel, gomp_team_barrier_cancel): New
	functions.
	* config/linux/proc.h (gomp_cpuset_popcount): Add attribute_hidden.
	Add cpusetsize argument.
	(gomp_cpuset_size, gomp_cpusetp): Declare.
	* config/linux/affinity.c: Include errno.h, stdio.h and string.h.
	(affinity_counter): Remove.
	(CPU_ISSET_S, CPU_ZERO_S, CPU_SET_S, CPU_CLR_S): Define
	if CPU_ALLOC_SIZE isn't defined.
	(gomp_init_affinity): Rewritten, if gomp_places_list is NULL, try
	silently create OMP_PLACES=threads, if it is non-NULL afterwards,
	bind current thread to the first place.
	(gomp_init_thread_affinity): Rewritten.  Add place argument, just
	pthread_setaffinity_np to gomp_places_list[place].
	(gomp_affinity_alloc, gomp_affinity_init_place, gomp_affinity_add_cpus,
	gomp_affinity_remove_cpu, gomp_affinity_copy_place,
	gomp_affinity_same_place, gomp_affinity_finalize_place_list,
	gomp_affinity_init_level, gomp_affinity_print_place): New functions.
	* config/linux/bar.h (BAR_TASK_PENDING, BAR_WAS_LAST,
	BAR_WAITING_FOR_TASK, BAR_INCR, BAR_CANCELLED): Define.
	(gomp_barrier_t): Add awaited_final field.
	(gomp_barrier_init): Initialize awaited_final field.
	(gomp_team_barrier_wait_final, gomp_team_barrier_wait_cancel,
	gomp_team_barrier_wait_cancel_end, gomp_team_barrier_cancel): New
	prototypes.
	(gomp_barrier_wait_start): Preserve BAR_CANCELLED bit.  Use BAR_*
	defines.
	(gomp_barrier_wait_cancel_start, gomp_team_barrier_wait_final_start,
	gomp_team_barrier_cancelled): New inline functions.
	(gomp_barrier_last_thread,
	gomp_team_barrier_set_task_pending,
	gomp_team_barrier_clear_task_pending,
	gomp_team_barrier_set_waiting_for_tasks,
	gomp_team_barrier_waiting_for_tasks,
	gomp_team_barrier_done): Use BAR_* defines.
	* config/posix/bar.c (gomp_barrier_init): Clear cancellable field.
	(gomp_barrier_wait_end): Use BAR_* defines.
	(gomp_team_barrier_wait_end): Clear BAR_CANCELLED from state.
	Set work_share_cancelled to 0 on last thread, use __atomic_load_n.
	Use BAR_* defines.
	(gomp_team_barrier_wait_cancel_end, gomp_team_barrier_wait_cancel,
	gomp_team_barrier_cancel): New functions.
	* config/posix/affinity.c (gomp_init_thread_affinity): Add place
	argument.
	(gomp_affinity_alloc, gomp_affinity_init_place, gomp_affinity_add_cpus,
	gomp_affinity_remove_cpu, gomp_affinity_copy_place,
	gomp_affinity_same_place, gomp_affinity_finalize_place_list,
	gomp_affinity_init_level, gomp_affinity_print_place): New stubs.
	* config/posix/bar.h (BAR_TASK_PENDING, BAR_WAS_LAST,
	BAR_WAITING_FOR_TASK, BAR_INCR, BAR_CANCELLED): Define.
	(gomp_barrier_t): Add cancellable field.
	(gomp_team_barrier_wait_cancel, gomp_team_barrier_wait_cancel_end,
	gomp_team_barrier_cancel): New prototypes.
	(gomp_barrier_wait_start): Preserve BAR_CANCELLED bit.
	(gomp_barrier_wait_cancel_start, gomp_team_barrier_wait_final,
	gomp_team_barrier_cancelled): New inline functions.
	(gomp_barrier_wait_start, gomp_barrier_last_thread,
	gomp_team_barrier_set_task_pending,
	gomp_team_barrier_clear_task_pending,
	gomp_team_barrier_set_waiting_for_tasks,
	gomp_team_barrier_waiting_for_tasks,
	gomp_team_barrier_done): Use BAR_* defines.
	* barrier.c (GOMP_barrier_cancel): New function.
	* omp_lib.h.in (omp_proc_bind_kind, omp_proc_bind_false,
	omp_proc_bind_true, omp_proc_bind_master, omp_proc_bind_close,
	omp_proc_bind_spread): New params.
	(omp_get_cancellation, omp_get_proc_bind, omp_set_default_device,
	omp_get_default_device, omp_get_num_devices, omp_get_num_teams,
	omp_get_team_num, omp_is_initial_device): New externals.
	* parallel.c (GOMP_parallel, GOMP_cancel, GOMP_cancellation_point):
	New functions.
	(gomp_resolve_num_threads): Adjust for thread_limit now being in
	icv->thread_limit_var.  Use UINT_MAX instead of ULONG_MAX as
	infinity.  If not nested, just return minimum of max_num_threads
	and icv->thread_limit_var and if thr->thread_pool, set threads_busy
	to the returned value.  Otherwise, don't update atomically
	gomp_remaining_threads_count, but instead thr->thread_pool->threads_busy.
	(GOMP_parallel_end): Adjust for thread_limit now being in
	icv->thread_limit_var.  Use UINT_MAX instead of ULONG_MAX as
	infinity.  Adjust threads_busy in the pool rather than
	gomp_remaining_threads_count.  Remember team->nthreads and call
	gomp_team_end before adjusting threads_busy, if not nested
	afterwards, just set it to 1 non-atomically.  Add ialias.
	(GOMP_parallel_start): Adjust gomp_team_start caller.
	* testsuite/libgomp.c/atomic-14.c: Add parens to make it valid.
	* testsuite/libgomp.c/affinity-1.c: New test.
	* testsuite/libgomp.c/atomic-15.c: New test.
	* testsuite/libgomp.c/atomic-16.c: New test.
	* testsuite/libgomp.c/atomic-17.c: New test.
	* testsuite/libgomp.c/cancel-for-1.c: New test.
	* testsuite/libgomp.c/cancel-for-2.c: New test.
	* testsuite/libgomp.c/cancel-parallel-1.c: New test.
	* testsuite/libgomp.c/cancel-parallel-2.c: New test.
	* testsuite/libgomp.c/cancel-parallel-3.c: New test.
	* testsuite/libgomp.c/cancel-sections-1.c: New test.
	* testsuite/libgomp.c/cancel-taskgroup-1.c: New test.
	* testsuite/libgomp.c/cancel-taskgroup-2.c: New test.
	* testsuite/libgomp.c/depend-1.c: New test.
	* testsuite/libgomp.c/depend-2.c: New test.
	* testsuite/libgomp.c/depend-3.c: New test.
	* testsuite/libgomp.c/depend-4.c: New test.
	* testsuite/libgomp.c/for-1.c: New test.
	* testsuite/libgomp.c/for-1.h: New file.
	* testsuite/libgomp.c/for-2.c: New test.
	* testsuite/libgomp.c/for-2.h: New file.
	* testsuite/libgomp.c/for-3.c: New test.
	* testsuite/libgomp.c/pr58392.c: New test.
	* testsuite/libgomp.c/simd-1.c: New test.
	* testsuite/libgomp.c/simd-2.c: New test.
	* testsuite/libgomp.c/simd-3.c: New test.
	* testsuite/libgomp.c/simd-4.c: New test.
	* testsuite/libgomp.c/simd-5.c: New test.
	* testsuite/libgomp.c/simd-6.c: New test.
	* testsuite/libgomp.c/target-1.c: New test.
	* testsuite/libgomp.c/target-2.c: New test.
	* testsuite/libgomp.c/target-3.c: New test.
	* testsuite/libgomp.c/target-4.c: New test.
	* testsuite/libgomp.c/target-5.c: New test.
	* testsuite/libgomp.c/target-6.c: New test.
	* testsuite/libgomp.c/target-7.c: New test.
	* testsuite/libgomp.c/taskgroup-1.c: New test.
	* testsuite/libgomp.c/thread-limit-1.c: New test.
	* testsuite/libgomp.c/thread-limit-2.c: New test.
	* testsuite/libgomp.c/thread-limit-3.c: New test.
	* testsuite/libgomp.c/udr-1.c: New test.
	* testsuite/libgomp.c/udr-2.c: New test.
	* testsuite/libgomp.c/udr-3.c: New test.
	* testsuite/libgomp.c++/affinity-1.C: New test.
	* testsuite/libgomp.c++/atomic-10.C: New test.
	* testsuite/libgomp.c++/atomic-11.C: New test.
	* testsuite/libgomp.c++/atomic-12.C: New test.
	* testsuite/libgomp.c++/atomic-13.C: New test.
	* testsuite/libgomp.c++/atomic-14.C: New test.
	* testsuite/libgomp.c++/atomic-15.C: New test.
	* testsuite/libgomp.c++/cancel-for-1.C: New test.
	* testsuite/libgomp.c++/cancel-for-2.C: New test.
	* testsuite/libgomp.c++/cancel-parallel-1.C: New test.
	* testsuite/libgomp.c++/cancel-parallel-2.C: New test.
	* testsuite/libgomp.c++/cancel-parallel-3.C: New test.
	* testsuite/libgomp.c++/cancel-sections-1.C: New test.
	* testsuite/libgomp.c++/cancel-taskgroup-1.C: New test.
	* testsuite/libgomp.c++/cancel-taskgroup-2.C: New test.
	* testsuite/libgomp.c++/cancel-taskgroup-3.C: New test.
	* testsuite/libgomp.c++/cancel-test.h: New file.
	* testsuite/libgomp.c++/for-9.C: New test.
	* testsuite/libgomp.c++/for-10.C: New test.
	* testsuite/libgomp.c++/for-11.C: New test.
	* testsuite/libgomp.c++/simd-1.C: New test.
	* testsuite/libgomp.c++/simd-2.C: New test.
	* testsuite/libgomp.c++/simd-3.C: New test.
	* testsuite/libgomp.c++/simd-4.C: New test.
	* testsuite/libgomp.c++/simd-5.C: New test.
	* testsuite/libgomp.c++/simd-6.C: New test.
	* testsuite/libgomp.c++/simd-7.C: New test.
	* testsuite/libgomp.c++/simd-8.C: New test.
	* testsuite/libgomp.c++/target-1.C: New test.
	* testsuite/libgomp.c++/target-2.C: New test.
	* testsuite/libgomp.c++/target-2-aux.cc: New file.
	* testsuite/libgomp.c++/target-3.C: New test.
	* testsuite/libgomp.c++/taskgroup-1.C: New test.
	* testsuite/libgomp.c++/udr-1.C: New test.
	* testsuite/libgomp.c++/udr-2.C: New test.
	* testsuite/libgomp.c++/udr-3.C: New test.
	* testsuite/libgomp.c++/udr-4.C: New test.
	* testsuite/libgomp.c++/udr-5.C: New test.
	* testsuite/libgomp.c++/udr-6.C: New test.
	* testsuite/libgomp.c++/udr-7.C: New test.
	* testsuite/libgomp.c++/udr-8.C: New test.
	* testsuite/libgomp.c++/udr-9.C: New test.
gcc/
	* tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE__LOOPTEMP_
	and new OpenMP 4.0 clauses, handle UDR OMP_CLAUSE_REDUCTION,
	formatting fixes, use pp_colon instead of pp_character (..., ':'),
	similarly pp_right_paren.
	(dump_generic_node): Handle OMP_DISTRIBUTE, OMP_TEAMS,
	OMP_TARGET_DATA, OMP_TARGET, OMP_TARGET_UPDATE, OMP_TASKGROUP,
	allow OMP_FOR_INIT to be NULL, handle OMP_ATOMIC_SEQ_CST.
	* tree.c (omp_clause_num_ops, omp_clause_code_name): Add OpenMP 4.0
	clauses.
	(omp_declare_simd_clauses_equal,
	omp_remove_redundant_declare_simd_attrs): New functions.
	(attribute_value_equal): Use omp_declare_simd_clauses_equal.
	(walk_tree_1): Handle new OpenMP 4.0 clauses.
	* tree.h (OMP_LOOP_CHECK): Define.
	(OMP_FOR_BODY, OMP_FOR_CLAUSES, OMP_FOR_INIT, OMP_FOR_COND,
	OMP_FOR_INCR, OMP_FOR_PRE_BODY): Use it.
	(OMP_TASKGROUP_BODY, OMP_TEAMS_BODY, OMP_TEAMS_CLAUSES,
	OMP_TARGET_DATA_BODY, OMP_TARGET_DATA_CLAUSES, OMP_TARGET_BODY,
	OMP_TARGET_CLAUSES, OMP_TARGET_UPDATE_CLAUSES, OMP_CLAUSE_SIZE,
	OMP_ATOMIC_SEQ_CST, OMP_CLAUSE_DEPEND_KIND, OMP_CLAUSE_MAP_KIND,
	OMP_CLAUSE_MAP_ZERO_BIAS_ARRAY_SECTION, OMP_CLAUSE_PROC_BIND_KIND,
	OMP_CLAUSE_REDUCTION_OMP_ORIG_REF, OMP_CLAUSE_ALIGNED_ALIGNMENT,
	OMP_CLAUSE_NUM_TEAMS_EXPR, OMP_CLAUSE_THREAD_LIMIT_EXPR,
	OMP_CLAUSE_DEVICE_ID, OMP_CLAUSE_DIST_SCHEDULE_CHUNK_EXPR,
	OMP_CLAUSE_SIMDLEN_EXPR): Define.
	(OMP_CLAUSE_DECL): Change range up to OMP_CLAUSE__LOOPTEMP_.
	(omp_remove_redundant_declare_simd_attrs): New prototype.
	* gimple.def (GIMPLE_OMP_TASKGROUP, GIMPLE_OMP_TARGET,
	GIMPLE_OMP_TEAMS): New codes.
	(GIMPLE_OMP_RETURN): Use GSS_OMP_ATOMIC_STORE instead of GSS_BASE.
	* omp-low.c (struct omp_context): Add cancel_label and cancellable
	fields.
	(target_nesting_level): New variable.
	(extract_omp_for_data): Handle GF_OMP_FOR_KIND_DISTRIBUTE and
	OMP_CLAUSE_DIST_SCHEDULE.  Don't fallback to library implementation
	for collapse > 1 static schedule unless ordered.
	(get_ws_args_for): Add par_stmt argument.  Handle combined loops.
	(determine_parallel_type): Adjust get_ws_args_for caller.
	(install_var_field): Handle mask & 4 for double indirection.
	(scan_sharing_clauses): Ignore shared clause on teams construct.
	Handle OMP_CLAUSE__LOOPTEMP_ and new OpenMP 4.0 clauses.
	(create_omp_child_function): If inside target or declare target
	constructs, set "omp declare target" attribute on the child
	function.
	(find_combined_for): New function.
	(scan_omp_parallel): Handle combined loops.
	(scan_omp_target, scan_omp_teams): New functions.
	(check_omp_nesting_restrictions): Check new OpenMP 4.0 nesting
	restrictions and set ctx->cancellable for cancellable constructs.
	(scan_omp_1_stmt): Call check_omp_nesting_restrictions also on
	selected builtin calls.  Handle GIMPLE_OMP_TASKGROUP,
	GIMPLE_OMP_TARGET, GIMPLE_OMP_TEAMS.
	(build_omp_barrier): Add lhs argument, return gimple rather than
	tree.
	(omp_clause_aligned_alignment): New function.
	(lower_rec_simd_input_clauses): Only call SET_DECL_VALUE_EXPR
	on decls.
	(lower_rec_input_clauses): Add FD argument.  Ignore shared clauses
	on teams constructs.  Handle user defined reductions and new
	OpenMP 4.0 clauses.
	(lower_reduction_clauses): Don't set placeholder to address of ref
	if it has already the right type.
	(lower_send_clauses): Handle OMP_CLAUSE__LOOPTEMP_.
	(expand_parallel_call): Use the new non-_start suffixed builtins,
	handle OMP_CLAUSE_PROC_BIND, don't call the outlined function
	and GOMP_parallel_end after the call.
	(expand_task_call): Handle OMP_CLAUSE_DEPEND.
	(expand_omp_for_init_counts): Handle combined loops.
	(expand_omp_for_init_vars): Add inner_stmt argument, handle combined
	loops.
	(expand_omp_for_generic): Likewise.  Use GOMP_loop_end_cancel at the
	end of cancellable loops.
	(expand_omp_for_static_nochunk, expand_omp_for_static_chunk):
	Likewise.  Handle collapse > 1 loops.
	(expand_omp_simd): Handle combined loops.
	(expand_omp_for): Add inner_stmt argument, adjust callers of
	expand_omp_for* functions, use expand_omp_for_static*chunk even
	for collapse > 1 unless ordered.
	(expand_omp_sections): Use GOMP_sections_end_cancel at the end
	of cancellable sections.
	(expand_omp_single): Remove need_barrier variable, just rely on
	gimple_omp_return_nowait_p.  Adjust build_omp_barrier caller.
	(expand_omp_synch): Allow GIMPLE_OMP_TASKGROUP and GIMPLE_OMP_TEAMS.
	(expand_omp_atomic_load, expand_omp_atomic_store,
	expand_omp_atomic_fetch_op): Handle gimple_omp_atomic_seq_cst_p.
	(expand_omp_target): New function.
	(expand_omp): Handle combined loops.  Handle GIMPLE_OMP_TASKGROUP,
	GIMPLE_OMP_TEAMS, GIMPLE_OMP_TARGET.
	(build_omp_regions_1): Immediately close region for
	GF_OMP_TARGET_KIND_UPDATE.
	(maybe_add_implicit_barrier_cancel): New function.
	(lower_omp_sections): Adjust lower_rec_input_clauses caller.  Handle
	cancellation.
	(lower_omp_single): Likewise.  Add clobber after the barrier.
	(lower_omp_taskgroup): New function.
	(lower_omp_for): Handle combined loops.  Adjust
	lower_rec_input_clauses caller.  Handle cancellation.
	(lower_depend_clauses): New function.
	(lower_omp_taskreg): Lower depend clauses.  Adjust
	lower_rec_input_clauses caller.  Add clobber after the call.  Handle
	cancellation.
	(lower_omp_target, lower_omp_teams): New functions.
	(lower_omp_1): Handle cancellation.  Handle GIMPLE_OMP_TASKGROUP,
	GIMPLE_OMP_TARGET, GIMPLE_OMP_TEAMS and GOMP_barrier, GOMP_cancel
	and GOMP_cancellation_point calls.
	(lower_omp): Fold stmts inside of target region.
	(diagnose_sb_1, diagnose_sb_2): Handle GIMPLE_OMP_TASKGROUP,
	GIMPLE_OMP_TARGET and GIMPLE_OMP_TEAMS.
	* builtin-types.def (DEF_FUNCTION_TYPE_8): Document.
	(BT_FN_VOID_OMPFN_PTR_UINT,
	BT_FN_VOID_OMPFN_PTR_UINT_LONG_LONG_LONG,
	BT_FN_VOID_OMPFN_PTR_UINT_LONG_LONG_LONG_LONG,
	BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT): Remove.
	(BT_FN_VOID_OMPFN_PTR_UINT_UINT_UINT,
	BT_FN_VOID_OMPFN_PTR_UINT_LONG_LONG_LONG_UINT,
	BT_FN_VOID_OMPFN_PTR_UINT_LONG_LONG_LONG_LONG_UINT,
	BT_FN_BOOL_INT, BT_FN_BOOL_INT_BOOL, BT_FN_VOID_UINT_UINT,
	BT_FN_VOID_INT_PTR_SIZE_PTR_PTR_PTR,
	BT_FN_VOID_INT_OMPFN_PTR_SIZE_PTR_PTR_PTR,
	BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR): New.
	* tree-ssa-alias.c (ref_maybe_used_by_call_p_1,
	call_may_clobber_ref_p_1): Handle BUILT_IN_GOMP_BARRIER_CANCEL,
	BUILT_IN_GOMP_TASKGROUP_END, BUILT_IN_GOMP_LOOP_END_CANCEL,
	BUILT_IN_GOMP_SECTIONS_END_CANCEL.  Don't handle
	BUILT_IN_GOMP_PARALLEL_END.
	* gimple-low.c (lower_stmt): Handle GIMPLE_OMP_TASKGROUP,
	GIMPLE_OMP_TARGET and GIMPLE_OMP_TEAMS.
	* gimple-pretty-print.c (dump_gimple_omp_for): Handle
	GF_OMP_FOR_KIND_DISTRIBUTE.
	(dump_gimple_omp_target, dump_gimple_omp_teams): New functions.
	(dump_gimple_omp_block): Handle GIMPLE_OMP_TASKGROUP.
	(dump_gimple_omp_return): Print lhs if it has any.
	(dump_gimple_omp_atomic_load, dump_gimple_omp_atomic_store): Handle
	gimple_omp_atomic_seq_cst_p.
	(pp_gimple_stmt_1): Handle GIMPLE_OMP_TASKGROUP, GIMPLE_OMP_TARGET
	and GIMPLE_OMP_TEAMS.
	* langhooks.c (lhd_omp_mappable_type): New function.
	* tree-vectorizer.c (struct simd_array_to_simduid): Fix up comment.
	* langhooks.h (struct lang_hooks_for_types): Add omp_mappable_type
	hook.
	* gimplify.c (enum gimplify_omp_var_data): Add GOVD_MAP,
	GOVD_ALIGNED and GOVD_MAP_TO_ONLY.
	(enum omp_region_type): Add ORT_TEAMS, ORT_TARGET_DATA and
	ORT_TARGET.
	(struct gimplify_omp_ctx): Add combined_loop field.
	(gimplify_call_expr, gimplify_modify_expr): Don't call fold_stmt
	on stmts inside of target region.
	(is_gimple_stmt): Return true for OMP_DISTRIBUTE and OMP_TASKGROUP.
	(omp_firstprivatize_variable): Handle GOVD_MAP, GOVD_ALIGNED,
	ORT_TARGET and ORT_TARGET_DATA.
	(omp_add_variable): Avoid checks on readding var for GOVD_ALIGNED.
	Handle GOVD_MAP.
	(omp_notice_threadprivate_variable): Complain about threadprivate
	variables in target region.
	(omp_notice_variable): Complain about vars with non-mappable type
	in target region.  Handle ORT_TEAMS, ORT_TARGET and ORT_TARGET_DATA.
	(omp_check_private): Ignore ORT_TARGET* regions.
	(gimplify_scan_omp_clauses, gimplify_adjust_omp_clauses_1,
	gimplify_adjust_omp_clauses): Handle new OpenMP 4.0 clauses.
	(find_combined_omp_for): New function.
	(gimplify_omp_for): Handle gimplification of combined loops.
	(gimplify_omp_workshare): Gimplify also OMP_TARGET, OMP_TARGET_DATA,
	OMP_TEAMS.
	(gimplify_omp_target_update): New function.
	(gimplify_omp_atomic): Handle OMP_ATOMIC_SEQ_CST.
	(gimplify_expr): Handle OMP_DISTRIBUTE, OMP_TARGET, OMP_TARGET_DATA,
	OMP_TARGET_UPDATE, OMP_TEAMS, OMP_TASKGROUP.
	(gimplify_body): If fndecl has "omp declare target" attribute, add
	implicit ORT_TARGET context around it.
	* tree.def (OMP_DISTRIBUTE, OMP_TEAMS, OMP_TARGET_DATA, OMP_TARGET,
	OMP_TASKGROUP, OMP_TARGET_UPDATE): New tree codes.
	* tree-nested.c (convert_nonlocal_reference_stmt,
	convert_local_reference_stmt, convert_gimple_call): Handle
	GIMPLE_OMP_TARGET, GIMPLE_OMP_TEAMS and GIMPLE_OMP_TASKGROUP.
	* omp-builtins.def (BUILT_IN_GOMP_TASK): Use
	BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR
	instead of BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT.
	(BUILT_IN_GOMP_TARGET, BUILT_IN_GOMP_TARGET_DATA,
	BUILT_IN_GOMP_TARGET_END_DATA, BUILT_IN_GOMP_TARGET_UPDATE,
	BUILT_IN_GOMP_TEAMS, BUILT_IN_BARRIER_CANCEL,
	BUILT_IN_GOMP_LOOP_END_CANCEL,
	BUILT_IN_GOMP_SECTIONS_END_CANCEL, BUILT_IN_OMP_GET_TEAM_NUM,
	BUILT_IN_OMP_GET_NUM_TEAMS, BUILT_IN_GOMP_TASKGROUP_START,
	BUILT_IN_GOMP_TASKGROUP_END, BUILT_IN_GOMP_PARALLEL_LOOP_STATIC,
	BUILT_IN_GOMP_PARALLEL_LOOP_DYNAMIC,
	BUILT_IN_GOMP_PARALLEL_LOOP_GUIDED,
	BUILT_IN_GOMP_PARALLEL_LOOP_RUNTIME, BUILT_IN_GOMP_PARALLEL,
	BUILT_IN_GOMP_PARALLEL_SECTIONS, BUILT_IN_GOMP_CANCEL,
	BUILT_IN_GOMP_CANCELLATION_POINT): New built-ins.
	(BUILT_IN_GOMP_PARALLEL_LOOP_STATIC_START,
	BUILT_IN_GOMP_PARALLEL_LOOP_DYNAMIC_START,
	BUILT_IN_GOMP_PARALLEL_LOOP_GUIDED_START,
	BUILT_IN_GOMP_PARALLEL_LOOP_RUNTIME_START,
	BUILT_IN_GOMP_PARALLEL_START, BUILT_IN_GOMP_PARALLEL_END,
	BUILT_IN_GOMP_PARALLEL_SECTIONS_START): Remove.
	* tree-inline.c (remap_gimple_stmt, estimate_num_insns):
	Handle GIMPLE_OMP_TARGET, GIMPLE_OMP_TEAMS and GIMPLE_OMP_TASKGROUP.
	* gimple.c (gimple_build_omp_taskgroup, gimple_build_omp_target,
	gimple_build_omp_teams): New functions.
	(walk_gimple_op): Handle GIMPLE_OMP_TARGET, GIMPLE_OMP_TEAMS and
	GIMPLE_OMP_TASKGROUP.  Walk optional lhs on GIMPLE_OMP_RETURN.
	(walk_gimple_stmt, gimple_copy): Handle GIMPLE_OMP_TARGET,
	GIMPLE_OMP_TEAMS and GIMPLE_OMP_TASKGROUP.
	* gimple.h (enum gf_mask): GF_OMP_FOR_KIND_DISTRIBUTE,
	GF_OMP_FOR_COMBINED, GF_OMP_FOR_COMBINED_INTO,
	GF_OMP_TARGET_KIND_MASK, GF_OMP_TARGET_KIND_REGION,
	GF_OMP_TARGET_KIND_DATA, GF_OMP_TARGET_KIND_UPDATE,
	GF_OMP_ATOMIC_SEQ_CST): New.
	(gimple_build_omp_taskgroup, gimple_build_omp_target,
	gimple_build_omp_teams): New prototypes.
	(gimple_has_substatements): Handle GIMPLE_OMP_TARGET,
	GIMPLE_OMP_TEAMS and GIMPLE_OMP_TASKGROUP.
	(gimple_omp_subcode): Use GIMPLE_OMP_TEAMS instead of
	GIMPLE_OMP_SINGLE as end of range.
	(gimple_omp_return_set_lhs, gimple_omp_return_lhs,
	gimple_omp_return_lhs_ptr, gimple_omp_atomic_seq_cst_p,
	gimple_omp_atomic_set_seq_cst, gimple_omp_for_combined_p,
	gimple_omp_for_set_combined_p, gimple_omp_for_combined_into_p,
	gimple_omp_for_set_combined_into_p, gimple_omp_target_clauses,
	gimple_omp_target_clauses_ptr, gimple_omp_target_set_clauses,
	gimple_omp_target_kind, gimple_omp_target_set_kind,
	gimple_omp_target_child_fn, gimple_omp_target_child_fn_ptr,
	gimple_omp_target_set_child_fn, gimple_omp_target_data_arg,
	gimple_omp_target_data_arg_ptr, gimple_omp_target_set_data_arg,
	gimple_omp_teams_clauses, gimple_omp_teams_clauses_ptr,
	gimple_omp_teams_set_clauses): New inlines.
	(CASE_GIMPLE_OMP): Add GIMPLE_OMP_TARGET, GIMPLE_OMP_TEAMS
	and GIMPLE_OMP_TASKGROUP.
	* tree-core.h (enum omp_clause_code): Add new OpenMP 4.0 clause
	codes.
	(enum omp_clause_depend_kind, enum omp_clause_map_kind,
	enum omp_clause_proc_bind_kind): New.
	(union omp_clause_subcode): Add depend_kind, map_kind and
	proc_bind_kind fields.
	* tree-cfg.c (make_edges): Handle GIMPLE_OMP_TARGET,
	GIMPLE_OMP_TEAMS and GIMPLE_OMP_TASKGROUP.
	* langhooks-def.h (lhd_omp_mappable_type): New prototype.
	(LANG_HOOKS_OMP_MAPPABLE_TYPE): Define.
	(LANG_HOOKS_FOR_TYPES_INITIALIZER): Add it.
gcc/c-family/
	* c-cppbuiltin.c (c_cpp_builtins): Predefine _OPENMP to
	201307 instead of 201107.
	* c-common.c (DEF_FUNCTION_TYPE_8): Define.
	(c_common_attribute_table): Add "omp declare target" and
	"omp declare simd" attributes.
	(handle_omp_declare_target_attribute,
	handle_omp_declare_simd_attribute): New functions.
	* c-omp.c: Include c-pragma.h.
	(c_finish_omp_taskgroup): New function.
	(c_finish_omp_atomic): Add swapped argument, if true,
	build the operation first with rhs, lhs arguments and use NOP_EXPR
	build_modify_expr.
	(c_finish_omp_for): Add code argument, pass it down to make_code.
	(c_omp_split_clauses): New function.
	(c_split_parallel_clauses): Removed.
	(c_omp_declare_simd_clause_cmp, c_omp_declare_simd_clauses_to_numbers,
	c_omp_declare_simd_clauses_to_decls): New functions.
	* c-common.h (omp_clause_mask): New type.
	(OMP_CLAUSE_MASK_1): Define.
	(omp_clause_mask::omp_clause_mask, omp_clause_mask::operator &=,
	omp_clause_mask::operator |=, omp_clause_mask::operator ~,
	omp_clause_mask::operator |, omp_clause_mask::operator &,
	omp_clause_mask::operator <<, omp_clause_mask::operator >>,
	omp_clause_mask::operator ==): New methods.
	(enum c_omp_clause_split): New.
	(c_finish_omp_taskgroup): New prototype.
	(c_finish_omp_atomic): Add swapped argument.
	(c_finish_omp_for): Add code argument.
	(c_omp_split_clauses): New prototype.
	(c_split_parallel_clauses): Removed.
	(c_omp_declare_simd_clauses_to_numbers,
	c_omp_declare_simd_clauses_to_decls): New prototypes.
	* c-pragma.c (omp_pragmas): Add new OpenMP 4.0 constructs.
	* c-pragma.h (enum pragma_kind): Add PRAGMA_OMP_CANCEL,
	PRAGMA_OMP_CANCELLATION_POINT, PRAGMA_OMP_DECLARE_REDUCTION,
	PRAGMA_OMP_DISTRIBUTE, PRAGMA_OMP_END_DECLARE_TARGET, PRAGMA_OMP_SIMD,
	PRAGMA_OMP_TARGET, PRAGMA_OMP_TASKGROUP and PRAGMA_OMP_TEAMS.
	Remove PRAGMA_OMP_PARALLEL_FOR and PRAGMA_OMP_PARALLEL_SECTIONS.
	(enum pragma_omp_clause): Add PRAGMA_OMP_CLAUSE_ALIGNED,
	PRAGMA_OMP_CLAUSE_DEPEND, PRAGMA_OMP_CLAUSE_DEVICE,
	PRAGMA_OMP_CLAUSE_DIST_SCHEDULE, PRAGMA_OMP_CLAUSE_FOR,
	PRAGMA_OMP_CLAUSE_FROM, PRAGMA_OMP_CLAUSE_INBRANCH,
	PRAGMA_OMP_CLAUSE_LINEAR, PRAGMA_OMP_CLAUSE_MAP,
	PRAGMA_OMP_CLAUSE_NOTINBRANCH, PRAGMA_OMP_CLAUSE_NUM_TEAMS,
	PRAGMA_OMP_CLAUSE_PARALLEL, PRAGMA_OMP_CLAUSE_PROC_BIND,
	PRAGMA_OMP_CLAUSE_SAFELEN, PRAGMA_OMP_CLAUSE_SECTIONS,
	PRAGMA_OMP_CLAUSE_SIMDLEN, PRAGMA_OMP_CLAUSE_TASKGROUP,
	PRAGMA_OMP_CLAUSE_THREAD_LIMIT, PRAGMA_OMP_CLAUSE_TO and
	PRAGMA_OMP_CLAUSE_UNIFORM.
gcc/ada/
	* gcc-interface/utils.c (DEF_FUNCTION_TYPE_8): Define.
gcc/fortran/
	* trans-openmp.c (gfc_omp_clause_default_ctor,
	gfc_omp_clause_dtor): Return NULL for OMP_CLAUSE_REDUCTION.
	* f95-lang.c (ATTR_NULL, DEF_FUNCTION_TYPE_8): Define.
	* types.def (DEF_FUNCTION_TYPE_8): Document.
	(BT_FN_VOID_OMPFN_PTR_UINT,
	BT_FN_VOID_OMPFN_PTR_UINT_LONG_LONG_LONG,
	BT_FN_VOID_OMPFN_PTR_UINT_LONG_LONG_LONG_LONG,
	BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT): Remove.
	(BT_FN_VOID_OMPFN_PTR_UINT_UINT_UINT,
	BT_FN_VOID_OMPFN_PTR_UINT_LONG_LONG_LONG_UINT,
	BT_FN_VOID_OMPFN_PTR_UINT_LONG_LONG_LONG_LONG_UINT,
	BT_FN_BOOL_INT, BT_FN_BOOL_INT_BOOL, BT_FN_VOID_UINT_UINT,
	BT_FN_VOID_INT_PTR_SIZE_PTR_PTR_PTR,
	BT_FN_VOID_INT_OMPFN_PTR_SIZE_PTR_PTR_PTR,
	BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR): New.
gcc/lto/
	* lto-lang.c (DEF_FUNCTION_TYPE_8): Define.
gcc/c/
	* c-lang.h (current_omp_declare_target_attribute): New extern
	decl.
	* c-parser.c: Include c-lang.h.
	(struct c_parser): Change tokens to c_token *.
	Add tokens_buf field.  Change tokens_avail type to unsigned int.
	(c_parser_consume_token): If parser->tokens isn't
	&parser->tokens_buf[0], increment parser->tokens.
	(c_parser_consume_pragma): Likewise.
	(enum pragma_context): Add pragma_struct and pragma_param.
	(c_parser_external_declaration): Adjust
	c_parser_declaration_or_fndef caller.
	(c_parser_declaration_or_fndef): Add omp_declare_simd_clauses
	argument, if it is non-vNULL vector, call c_finish_omp_declare_simd.
	Adjust recursive call.
	(c_parser_struct_or_union_specifier): Use pragma_struct instead
	of pragma_external.
	(c_parser_parameter_declaration): Use pragma_param instead of
	pragma_external.
	(c_parser_compound_statement_nostart, c_parser_label,
	c_parser_for_statement): Adjust
	c_parser_declaration_or_fndef callers.
	(c_parser_expr_no_commas): Add omp_atomic_lhs argument, pass
	it through to c_parser_conditional_expression.
	(c_parser_conditional_expression): Add omp_atomic_lhs argument,
	pass it through to c_parser_binary_expression.  Adjust recursive
	call.
	(c_parser_binary_expression): Remove prec argument, add
	omp_atomic_lhs argument instead.  Always start from PREC_NONE, if
	omp_atomic_lhs is non-NULL and one of the arguments of toplevel
	binop matches it, use build2 instead of parser_build_binary_op.
	(c_parser_pragma): Handle PRAGMA_OMP_CANCEL,
	PRAGMA_OMP_CANCELLATION_POINT, PRAGMA_OMP_TARGET,
	PRAGMA_OMP_END_DECLARE_TARGET, PRAGMA_OMP_DECLARE_REDUCTION.
	Handle pragma_struct and pragma_param the same as pragma_external.
	(c_parser_omp_clause_name): Parse new OpenMP 4.0 clause names.
	(c_parser_omp_variable_list): Parse array sections for
	OMP_CLAUSE_{DEPEND,MAP,TO,FROM} clauses.
	(c_parser_omp_clause_collapse): Fully fold collapse expression.
	(c_parser_omp_clause_reduction): Handle user defined reductions.
	(c_parser_omp_clause_branch, c_parser_omp_clause_cancelkind,
	c_parser_omp_clause_num_teams, c_parser_omp_clause_thread_limit,
	c_parser_omp_clause_aligned, c_parser_omp_clause_linear,
	c_parser_omp_clause_safelen, c_parser_omp_clause_simdlen,
	c_parser_omp_clause_depend, c_parser_omp_clause_map,
	c_parser_omp_clause_device, c_parser_omp_clause_dist_schedule,
	c_parser_omp_clause_proc_bind, c_parser_omp_clause_to,
	c_parser_omp_clause_from, c_parser_omp_clause_uniform): New functions.
	(c_parser_omp_all_clauses): Add finish_p argument.  Don't call
	c_finish_omp_clauses if it is false.  Handle new OpenMP 4.0 clauses.
	(c_parser_omp_atomic): Parse seq_cst clause, pass true if it is
	present to c_finish_omp_atomic.  Handle OpenMP 4.0 atomic forms.
	(c_parser_omp_for_loop): Add CODE argument, pass it through
	to c_finish_omp_for.  Change last argument to cclauses,
	and adjust uses to grab parallel clauses from the array of all
	the split clauses.  Adjust c_parser_binary_expression,
	c_parser_declaration_or_fndef and c_finish_omp_for callers.
	(omp_split_clauses): New function.
	(c_parser_omp_simd): New function.
	(c_parser_omp_for): Add p_name, mask and cclauses arguments.
	Allow the function to be called also when parsing combined constructs,
	and call c_parser_omp_simd when parsing for simd.
	(c_parser_omp_sections_scope): If section-sequence doesn't start with
	#pragma omp section, require exactly one structured-block instead of
	sequence of statements.
	(c_parser_omp_sections): Add p_name, mask and cclauses arguments.
	Allow the function to be called also when parsing combined constructs.
	(c_parser_omp_parallel): Add p_name, mask and cclauses arguments.
	Allow the function to be called also when parsing combined
	constructs.
	(c_parser_omp_taskgroup, c_parser_omp_cancel,
	c_parser_omp_cancellation_point, c_parser_omp_distribute,
	c_parser_omp_teams, c_parser_omp_target_data,
	c_parser_omp_target_update, c_parser_omp_target,
	c_parser_omp_declare_simd, c_finish_omp_declare_simd,
	c_parser_omp_declare_target, c_parser_omp_end_declare_target,
	c_parser_omp_declare_reduction, c_parser_omp_declare): New functions.
	(c_parser_omp_construct): Add p_name and mask vars.  Handle
	PRAGMA_OMP_DISTRIBUTE, PRAGMA_OMP_SIMD, PRAGMA_OMP_TASKGROUP,
	PRAGMA_OMP_TEAMS.  Adjust c_parser_omp_for, c_parser_omp_parallel
	and c_parser_omp_sections callers.
	(c_parse_file): Initialize tparser.tokens and the_parser->tokens here.
	(OMP_FOR_CLAUSE_MASK, OMP_SECTIONS_CLAUSE_MASK,
	OMP_SINGLE_CLAUSE_MASK): Use OMP_CLAUSE_MASK_1 instead of 1.
	(OMP_PARALLEL_CLAUSE_MASK): Likewise.  Add OMP_CLAUSE_PROC_BIND.
	(OMP_TASK_CLAUSE_MASK): Use OMP_CLAUSE_MASK_1 instead of 1.  Add
	OMP_CLAUSE_DEPEND.
	(OMP_SIMD_CLAUSE_MASK, OMP_CANCEL_CLAUSE_MASK,
	OMP_CANCELLATION_POINT_CLAUSE_MASK, OMP_DISTRIBUTE_CLAUSE_MASK,
	OMP_TEAMS_CLAUSE_MASK, OMP_TARGET_DATA_CLAUSE_MASK,
	OMP_TARGET_UPDATE_CLAUSE_MASK, OMP_TARGET_CLAUSE_MASK,
	OMP_DECLARE_SIMD_CLAUSE_MASK): Define.
	* c-typeck.c: Include tree-inline.h.
	(c_finish_omp_cancel, c_finish_omp_cancellation_point,
	handle_omp_array_sections_1, handle_omp_array_sections,
	c_clone_omp_udr, c_find_omp_placeholder_r): New functions.
	(c_finish_omp_clauses): Handle new OpenMP 4.0 clauses and
	user defined reductions.
	(c_tree_equal): New function.
	* c-tree.h (temp_store_parm_decls, temp_pop_parm_decls,
	c_finish_omp_cancel, c_finish_omp_cancellation_point, c_tree_equal,
	c_omp_reduction_id, c_omp_reduction_decl, c_omp_reduction_lookup,
	c_check_omp_declare_reduction_r): New prototypes.
	* c-decl.c (current_omp_declare_target_attribute): New variable.
	(c_decl_attributes): New function.
	(start_decl, start_function): Use it instead of decl_attributes.
	(temp_store_parm_decls, temp_pop_parm_decls, c_omp_reduction_id,
	c_omp_reduction_decl, c_omp_reduction_lookup,
	c_check_omp_declare_reduction_r): New functions.
gcc/cp/
	* decl.c (duplicate_decls): Error out for redeclaration of UDRs.
	(declare_simd_adjust_this): New function.
	(grokfndecl): If "omp declare simd" attribute is present,
	call declare_simd_adjust_this if needed and
	c_omp_declare_simd_clauses_to_numbers.
	* cp-array-notation.c (expand_array_notation_exprs): Handle
	OMP_TASKGROUP.
	* cp-gimplify.c (cp_gimplify_expr): Handle OMP_SIMD and
	OMP_DISTRIBUTE.  Handle is_invisiref_parm decls in
	OMP_CLAUSE_REDUCTION.
	(cp_genericize_r): Handle OMP_SIMD and OMP_DISTRIBUTE like
	OMP_FOR.
	(cxx_omp_privatize_by_reference): Return true for
	is_invisiref_parm decls.
	(cxx_omp_finish_clause): Adjust cxx_omp_create_clause_info
	caller.
	* pt.c (apply_late_template_attributes): For "omp declare simd"
	attribute call tsubst_omp_clauses,
	c_omp_declare_simd_clauses_to_decls, finish_omp_clauses
	and c_omp_declare_simd_clauses_to_numbers.
	(instantiate_class_template_1): Call cp_check_omp_declare_reduction
	for UDRs.
	(tsubst_decl): Handle UDRs.
	(tsubst_omp_clauses): Add declare_simd argument, if true don't
	call finish_omp_clauses.  Handle new OpenMP 4.0 clauses.
	Handle non-NULL OMP_CLAUSE_REDUCTION_PLACEHOLDER on
	OMP_CLAUSE_REDUCTION.
	(tsubst_expr): For UDRs call pushdecl and
	cp_check_omp_declare_reduction.  Adjust tsubst_omp_clauses
	callers.  Handle OMP_SIMD, OMP_DISTRIBUTE, OMP_TEAMS,
	OMP_TARGET_DATA, OMP_TARGET_UPDATE, OMP_TARGET, OMP_TASKGROUP.
	Adjust finish_omp_atomic caller.
	(tsubst_omp_udr): New function.
	(instantiate_decl): For UDRs at block scope, don't call
	start_preparsed_function/finish_function.  Call tsubst_omp_udr.
	* semantics.c (cxx_omp_create_clause_info): Add need_dtor argument,
	use it instead of need_default_ctor || need_copy_ctor.
	(struct cp_check_omp_declare_reduction_data): New type.
	(handle_omp_array_sections_1, handle_omp_array_sections,
	omp_reduction_id, omp_reduction_lookup,
	cp_remove_omp_priv_cleanup_stmt, cp_check_omp_declare_reduction_r,
	cp_check_omp_declare_reduction, clone_omp_udr,
	find_omp_placeholder_r, finish_omp_reduction_clause): New functions.
	(finish_omp_clauses): Handle new OpenMP 4.0 clauses and user defined
	reductions.
	(finish_omp_for): Add CODE argument, use it instead of hardcoded
	OMP_FOR.  Adjust c_finish_omp_for caller.
	(finish_omp_atomic): Add seq_cst argument, adjust
	c_finish_omp_atomic callers, handle seq_cst and new OpenMP 4.0
	atomic variants.
	(finish_omp_cancel, finish_omp_cancellation_point): New functions.
	* decl2.c (mark_used): Force immediate instantiation of
	DECL_OMP_DECLARE_REDUCTION_P decls.
	(is_late_template_attribute): Return true for "omp declare simd"
	attribute.
	(cp_omp_mappable_type): New function.
	(cplus_decl_attributes): Add implicit "omp declare target" attribute
	if requested.
	* parser.c (cp_debug_parser): Print
	parser->colon_doesnt_start_class_def_p.
	(cp_ensure_no_omp_declare_simd, cp_finalize_omp_declare_simd): New
	functions.
	(enum pragma_context): Add pragma_member and pragma_objc_icode.
	(cp_parser_binary_expression): Handle no_toplevel_fold_p
	even for binary operations other than comparison.
	(cp_parser_linkage_specification): Call
	cp_ensure_no_omp_declare_simd if needed.
	(cp_parser_namespace_definition): Likewise.
	(cp_parser_init_declarator): Call cp_finalize_omp_declare_simd.
	(cp_parser_direct_declarator): Pass declarator to
	cp_parser_late_return_type_opt.
	(cp_parser_late_return_type_opt): Add declarator argument,
	call cp_parser_late_parsing_omp_declare_simd for declare simd.
	(cp_parser_class_specifier_1): Call cp_ensure_no_omp_declare_simd.
	Parse UDRs before all other methods.
	(cp_parser_member_specification_opt): Use pragma_member instead of
	pragma_external.
	(cp_parser_member_declaration): Call cp_finalize_omp_declare_simd.
	(cp_parser_function_definition_from_specifiers_and_declarator,
	cp_parser_save_member_function_body): Likewise.
	(cp_parser_late_parsing_for_member): Handle UDRs specially.
	(cp_parser_next_token_starts_class_definition_p): Don't allow
	CPP_COLON if colon_doesnt_start_class_def_p flag is true.
	(cp_parser_objc_interstitial_code): Use pragma_objc_icode
	instead of pragma_external.
	(cp_parser_omp_clause_name): Parse new OpenMP 4.0 clause names.
	(cp_parser_omp_var_list_no_open): Parse array sections for
	OMP_CLAUSE_{DEPEND,MAP,TO,FROM} clauses.  Add COLON argument,
	if non-NULL, allow parsing to end with a colon rather than close
	paren.
	(cp_parser_omp_var_list): Adjust cp_parser_omp_var_list_no_open
	caller.
	(cp_parser_omp_clause_reduction): Handle user defined reductions.
	(cp_parser_omp_clause_branch, cp_parser_omp_clause_cancelkind,
	cp_parser_omp_clause_num_teams, cp_parser_omp_clause_thread_limit,
	cp_parser_omp_clause_aligned, cp_parser_omp_clause_linear,
	cp_parser_omp_clause_safelen, cp_parser_omp_clause_simdlen,
	cp_parser_omp_clause_depend, cp_parser_omp_clause_map,
	cp_parser_omp_clause_device, cp_parser_omp_clause_dist_schedule,
	cp_parser_omp_clause_proc_bind, cp_parser_omp_clause_to,
	cp_parser_omp_clause_from, cp_parser_omp_clause_uniform): New
	functions.
	(cp_parser_omp_all_clauses): Add finish_p argument.  Don't call
	finish_omp_clauses if it is false.  Handle new OpenMP 4.0 clauses.
	(cp_parser_omp_atomic): Parse seq_cst clause, pass
	true if it is present to finish_omp_atomic.  Handle new OpenMP 4.0
	atomic forms.
	(cp_parser_omp_for_loop): Add CODE argument, pass it through
	to finish_omp_for.  Change last argument to cclauses,
	and adjust uses to grab parallel clauses from the array of all
	the split clauses.
	(cp_omp_split_clauses): New function.
	(cp_parser_omp_simd): New function.
	(cp_parser_omp_for): Add p_name, mask and cclauses arguments.
	Allow the function to be called also when parsing combined constructs,
	and call c_parser_omp_simd when parsing for simd.
	(cp_parser_omp_sections_scope): If section-sequence doesn't start with
	#pragma omp section, require exactly one structured-block instead of
	sequence of statements.
	(cp_parser_omp_sections): Add p_name, mask and cclauses arguments.
	Allow the function to be called also when parsing combined constructs.
	(cp_parser_omp_parallel): Add p_name, mask and cclauses arguments.
	Allow the function to be called also when parsing combined
	constructs.
	(cp_parser_omp_taskgroup, cp_parser_omp_cancel,
	cp_parser_omp_cancellation_point, cp_parser_omp_distribute,
	cp_parser_omp_teams, cp_parser_omp_target_data,
	cp_parser_omp_target_update, cp_parser_omp_target,
	cp_parser_omp_declare_simd, cp_parser_late_parsing_omp_declare_simd,
	cp_parser_omp_declare_target, cp_parser_omp_end_declare_target,
	cp_parser_omp_declare_reduction_exprs, cp_parser_omp_declare_reduction,
	cp_parser_omp_declare): New functions.
	(cp_parser_omp_construct): Add p_name and mask vars.  Handle
	PRAGMA_OMP_DISTRIBUTE, PRAGMA_OMP_SIMD, PRAGMA_OMP_TASKGROUP,
	PRAGMA_OMP_TEAMS.  Adjust cp_parser_omp_for, cp_parser_omp_parallel
	and cp_parser_omp_sections callers.
	(cp_parser_pragma): Handle PRAGMA_OMP_CANCEL,
	PRAGMA_OMP_CANCELLATION_POINT, PRAGMA_OMP_DECLARE_REDUCTION,
	PRAGMA_OMP_DISTRIBUTE, PRAGMA_OMP_SIMD, PRAGMA_OMP_TASKGROUP,
	PRAGMA_OMP_TEAMS, PRAGMA_OMP_TARGET, PRAGMA_OMP_END_DECLARE_TARGET.
	Handle pragma_member and pragma_objc_icode like pragma_external.
	(OMP_FOR_CLAUSE_MASK, OMP_SECTIONS_CLAUSE_MASK,
	OMP_SINGLE_CLAUSE_MASK): Use OMP_CLAUSE_MASK_1 instead of 1.
	(OMP_PARALLEL_CLAUSE_MASK): Likewise.  Add OMP_CLAUSE_PROC_BIND.
	(OMP_TASK_CLAUSE_MASK): Use OMP_CLAUSE_MASK_1 instead of 1.  Add
	OMP_CLAUSE_DEPEND.
	(OMP_SIMD_CLAUSE_MASK, OMP_CANCEL_CLAUSE_MASK,
	OMP_CANCELLATION_POINT_CLAUSE_MASK, OMP_DISTRIBUTE_CLAUSE_MASK,
	OMP_TEAMS_CLAUSE_MASK, OMP_TARGET_DATA_CLAUSE_MASK,
	OMP_TARGET_UPDATE_CLAUSE_MASK, OMP_TARGET_CLAUSE_MASK,
	OMP_DECLARE_SIMD_CLAUSE_MASK): Define.
	* parser.h (struct cp_omp_declare_simd_data): New type.
	(struct cp_parser): Add colon_doesnt_start_class_def_p and
	omp_declare_simd fields.
	* cp-objcp-common.h (LANG_HOOKS_OMP_MAPPABLE_TYPE): Define.
	* cp-tree.h (struct lang_decl_fn): Add omp_declare_reduction_p
	bit.
	(DECL_OMP_DECLARE_REDUCTION_P): Define.
	(OMP_FOR_GIMPLIFYING_P): Use OMP_LOOP_CHECK macro.
	(struct saved_scope): Add omp_declare_target_attribute field.
	(cp_omp_mappable_type, omp_reduction_id,
	cp_remove_omp_priv_cleanup_stmt, cp_check_omp_declare_reduction,
	finish_omp_cancel, finish_omp_cancellation_point): New prototypes.
	(finish_omp_for): Add CODE argument.
	(finish_omp_atomic): Add seq_cst argument.
	(cxx_omp_create_clause_info): Add need_dtor argument.
gcc/testsuite/
	* c-c++-common/gomp/atomic-15.c: Adjust for C diagnostics.
	Remove error test that is now valid in OpenMP 4.0.
	* c-c++-common/gomp/atomic-16.c: New test.
	* c-c++-common/gomp/cancel-1.c: New test.
	* c-c++-common/gomp/depend-1.c: New test.
	* c-c++-common/gomp/depend-2.c: New test.
	* c-c++-common/gomp/map-1.c: New test.
	* c-c++-common/gomp/pr58472.c: New test.
	* c-c++-common/gomp/sections1.c: New test.
	* c-c++-common/gomp/simd1.c: New test.
	* c-c++-common/gomp/simd2.c: New test.
	* c-c++-common/gomp/simd3.c: New test.
	* c-c++-common/gomp/simd4.c: New test.
	* c-c++-common/gomp/simd5.c: New test.
	* c-c++-common/gomp/single1.c: New test.
	* g++.dg/gomp/block-0.C: Adjust for stricter #pragma omp sections
	parser.
	* g++.dg/gomp/block-3.C: Likewise.
	* g++.dg/gomp/clause-3.C: Adjust error messages.
	* g++.dg/gomp/declare-simd-1.C: New test.
	* g++.dg/gomp/declare-simd-2.C: New test.
	* g++.dg/gomp/depend-1.C: New test.
	* g++.dg/gomp/depend-2.C: New test.
	* g++.dg/gomp/target-1.C: New test.
	* g++.dg/gomp/target-2.C: New test.
	* g++.dg/gomp/taskgroup-1.C: New test.
	* g++.dg/gomp/teams-1.C: New test.
	* g++.dg/gomp/udr-1.C: New test.
	* g++.dg/gomp/udr-2.C: New test.
	* g++.dg/gomp/udr-3.C: New test.
	* g++.dg/gomp/udr-4.C: New test.
	* g++.dg/gomp/udr-5.C: New test.
	* g++.dg/gomp/udr-6.C: New test.
	* gcc.dg/autopar/outer-1.c: Expect 4 instead of 5 loopfn matches.
	* gcc.dg/autopar/outer-2.c: Likewise.
	* gcc.dg/autopar/outer-3.c: Likewise.
	* gcc.dg/autopar/outer-4.c: Likewise.
	* gcc.dg/autopar/outer-5.c: Likewise.
	* gcc.dg/autopar/outer-6.c: Likewise.
	* gcc.dg/autopar/parallelization-1.c: Likewise.
	* gcc.dg/gomp/block-3.c: Adjust for stricter #pragma omp sections
	parser.
	* gcc.dg/gomp/clause-1.c: Adjust error messages.
	* gcc.dg/gomp/combined-1.c: Look for GOMP_parallel_loop_runtime
	instead of GOMP_parallel_loop_runtime_start.
	* gcc.dg/gomp/declare-simd-1.c: New test.
	* gcc.dg/gomp/declare-simd-2.c: New test.
	* gcc.dg/gomp/nesting-1.c: Adjust for stricter #pragma omp sections
	parser.  Add further #pragma omp sections nesting tests.
	* gcc.dg/gomp/target-1.c: New test.
	* gcc.dg/gomp/target-2.c: New test.
	* gcc.dg/gomp/taskgroup-1.c: New test.
	* gcc.dg/gomp/teams-1.c: New test.
	* gcc.dg/gomp/udr-1.c: New test.
	* gcc.dg/gomp/udr-2.c: New test.
	* gcc.dg/gomp/udr-3.c: New test.
	* gcc.dg/gomp/udr-4.c: New test.
	* gfortran.dg/gomp/appendix-a/a.35.5.f90: Add dg-error.


git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@203408 138bc75d-0d04-0410-961f-82ee72b054a4

8 files changed, 778 insertions, 111 deletions

diff --git a/libgomp/config/linux/affinity.c b/libgomp/config/linux/affinity.c
index dc6c7e5ed3b..789cdce077d 100644
--- a/libgomp/config/linux/affinity.c
+++ b/libgomp/config/linux/affinity.c
@@ -29,90 +29,327 @@
 #endif
 #include "libgomp.h"
 #include "proc.h"
+#include <errno.h>
 #include <stdlib.h>
+#include <stdio.h>
+#include <string.h>
 #include <unistd.h>
 
 #ifdef HAVE_PTHREAD_AFFINITY_NP
 
-static unsigned int affinity_counter;
+#ifndef CPU_ALLOC_SIZE
+#define CPU_ISSET_S(idx, size, set) CPU_ISSET(idx, set)
+#define CPU_ZERO_S(size, set) CPU_ZERO(set)
+#define CPU_SET_S(idx, size, set) CPU_SET(idx, set)
+#define CPU_CLR_S(idx, size, set) CPU_CLR(idx, set)
+#endif
 
 void
 gomp_init_affinity (void)
 {
-  cpu_set_t cpuset, cpusetnew;
-  size_t idx, widx;
-  unsigned long cpus = 0;
+  if (gomp_places_list == NULL)
+    {
+      if (!gomp_affinity_init_level (1, ULONG_MAX, true))
+	return;
+    }
+
+  struct gomp_thread *thr = gomp_thread ();
+  pthread_setaffinity_np (pthread_self (), gomp_cpuset_size,
+			  (cpu_set_t *) gomp_places_list[0]);
+  thr->place = 1;
+  thr->ts.place_partition_off = 0;
+  thr->ts.place_partition_len = gomp_places_list_len;
+}
+
+void
+gomp_init_thread_affinity (pthread_attr_t *attr, unsigned int place)
+{
+  pthread_attr_setaffinity_np (attr, gomp_cpuset_size,
+			       (cpu_set_t *) gomp_places_list[place]);
+}
+
+void **
+gomp_affinity_alloc (unsigned long count, bool quiet)
+{
+  unsigned long i;
+  void **ret;
+  char *p;
+
+  if (gomp_cpusetp == NULL)
+    {
+      if (!quiet)
+	gomp_error ("Could not get CPU affinity set");
+      return NULL;
+    }
 
-  if (pthread_getaffinity_np (pthread_self (), sizeof (cpuset), &cpuset))
+  ret = malloc (count * sizeof (void *) + count * gomp_cpuset_size);
+  if (ret == NULL)
     {
-      gomp_error ("could not get CPU affinity set");
-      free (gomp_cpu_affinity);
-      gomp_cpu_affinity = NULL;
-      gomp_cpu_affinity_len = 0;
-      return;
+      if (!quiet)
+	gomp_error ("Out of memory trying to allocate places list");
+      return NULL;
     }
 
-  CPU_ZERO (&cpusetnew);
-  if (gomp_cpu_affinity_len == 0)
+  p = (char *) (ret + count);
+  for (i = 0; i < count; i++, p += gomp_cpuset_size)
+    ret[i] = p;
+  return ret;
+}
+
+void
+gomp_affinity_init_place (void *p)
+{
+  cpu_set_t *cpusetp = (cpu_set_t *) p;
+  CPU_ZERO_S (gomp_cpuset_size, cpusetp);
+}
+
+bool
+gomp_affinity_add_cpus (void *p, unsigned long num,
+			unsigned long len, long stride, bool quiet)
+{
+  cpu_set_t *cpusetp = (cpu_set_t *) p;
+  unsigned long max = 8 * gomp_cpuset_size;
+  for (;;)
     {
-      unsigned long count = gomp_cpuset_popcount (&cpuset);
-      if (count >= 65536)
-	count = 65536;
-      gomp_cpu_affinity = malloc (count * sizeof (unsigned short));
-      if (gomp_cpu_affinity == NULL)
+      if (num >= max)
+	{
+	  if (!quiet)
+	    gomp_error ("Logical CPU number %lu out of range", num);
+	  return false;
+	}
+      CPU_SET_S (num, gomp_cpuset_size, cpusetp);
+      if (--len == 0)
+	return true;
+      if ((stride < 0 && num + stride > num)
+	  || (stride > 0 && num + stride < num))
 	{
-	  gomp_error ("not enough memory to store CPU affinity list");
-	  return;
+	  if (!quiet)
+	    gomp_error ("Logical CPU number %lu+%ld out of range",
+			num, stride);
+	  return false;
 	}
-      for (widx = idx = 0; widx < count && idx < 65536; idx++)
-	if (CPU_ISSET (idx, &cpuset))
+      num += stride;
+    }
+}
+
+bool
+gomp_affinity_remove_cpu (void *p, unsigned long num)
+{
+  cpu_set_t *cpusetp = (cpu_set_t *) p;
+  if (num >= 8 * gomp_cpuset_size)
+    {
+      gomp_error ("Logical CPU number %lu out of range", num);
+      return false;
+    }
+  if (!CPU_ISSET_S (num, gomp_cpuset_size, cpusetp))
+    {
+      gomp_error ("Logical CPU %lu to be removed is not in the set", num);
+      return false;
+    }
+  CPU_CLR_S (num, gomp_cpuset_size, cpusetp);
+  return true;
+}
+
+bool
+gomp_affinity_copy_place (void *p, void *q, long stride)
+{
+  unsigned long i, max = 8 * gomp_cpuset_size;
+  cpu_set_t *destp = (cpu_set_t *) p;
+  cpu_set_t *srcp = (cpu_set_t *) q;
+
+  CPU_ZERO_S (gomp_cpuset_size, destp);
+  for (i = 0; i < max; i++)
+    if (CPU_ISSET_S (i, gomp_cpuset_size, srcp))
+      {
+	if ((stride < 0 && i + stride > i)
+	    || (stride > 0 && (i + stride < i || i + stride >= max)))
+	  {
+	    gomp_error ("Logical CPU number %lu+%ld out of range", i, stride);
+	    return false;
+	  }
+	CPU_SET_S (i + stride, gomp_cpuset_size, destp);
+      }
+  return true;
+}
+
+bool
+gomp_affinity_same_place (void *p, void *q)
+{
+#ifdef CPU_EQUAL_S
+  return CPU_EQUAL_S (gomp_cpuset_size, (cpu_set_t *) p, (cpu_set_t *) q);
+#else
+  return memcmp (p, q, gomp_cpuset_size) == 0;
+#endif
+}
+
+bool
+gomp_affinity_finalize_place_list (bool quiet)
+{
+  unsigned long i, j;
+
+  for (i = 0, j = 0; i < gomp_places_list_len; i++)
+    {
+      cpu_set_t *cpusetp = (cpu_set_t *) gomp_places_list[i];
+      bool nonempty = false;
+#ifdef CPU_AND_S
+      CPU_AND_S (gomp_cpuset_size, cpusetp, cpusetp, gomp_cpusetp);
+      nonempty = gomp_cpuset_popcount (gomp_cpuset_size, cpusetp) != 0;
+#else
+      unsigned long k, max = gomp_cpuset_size / sizeof (cpusetp->__bits[0]);
+      for (k = 0; k < max; k++)
+	if ((cpusetp->__bits[k] &= gomp_cpusetp->__bits[k]) != 0)
+	  nonempty = true;
+#endif
+      if (nonempty)
+	gomp_places_list[j++] = gomp_places_list[i];
+    }
+
+  if (j == 0)
+    {
+      if (!quiet)
+	gomp_error ("None of the places contain usable logical CPUs");
+      return false;
+    }
+  else if (j < gomp_places_list_len)
+    {
+      if (!quiet)
+	gomp_error ("Number of places reduced from %ld to %ld because some "
+		    "places didn't contain any usable logical CPUs",
+		    gomp_places_list_len, j);
+      gomp_places_list_len = j;
+    }
+  return true;
+}
+
+bool
+gomp_affinity_init_level (int level, unsigned long count, bool quiet)
+{
+  unsigned long i, max = 8 * gomp_cpuset_size;
+
+  if (gomp_cpusetp)
+    {
+      unsigned long maxcount
+	= gomp_cpuset_popcount (gomp_cpuset_size, gomp_cpusetp);
+      if (count > maxcount)
+	count = maxcount;
+    }
+  gomp_places_list = gomp_affinity_alloc (count, quiet);
+  gomp_places_list_len = 0;
+  if (gomp_places_list == NULL)
+    return false;
+  /* SMT (threads).  */
+  if (level == 1)
+    {
+      for (i = 0; i < max && gomp_places_list_len < count; i++)
+	if (CPU_ISSET_S (i, gomp_cpuset_size, gomp_cpusetp))
 	  {
-	    cpus++;
-	    gomp_cpu_affinity[widx++] = idx;
+	    gomp_affinity_init_place (gomp_places_list[gomp_places_list_len]);
+	    gomp_affinity_add_cpus (gomp_places_list[gomp_places_list_len],
+				    i, 1, 0, true);
+	    ++gomp_places_list_len;
 	  }
+      return true;
     }
   else
-    for (widx = idx = 0; idx < gomp_cpu_affinity_len; idx++)
-      if (gomp_cpu_affinity[idx] < CPU_SETSIZE
-	  && CPU_ISSET (gomp_cpu_affinity[idx], &cpuset))
+    {
+      char name[sizeof ("/sys/devices/system/cpu/cpu/topology/"
+			"thread_siblings_list") + 3 * sizeof (unsigned long)];
+      size_t prefix_len = sizeof ("/sys/devices/system/cpu/cpu") - 1;
+      cpu_set_t *copy = gomp_alloca (gomp_cpuset_size);
+      FILE *f;
+      char *line = NULL;
+      size_t linelen = 0;
+
+      memcpy (name, "/sys/devices/system/cpu/cpu", prefix_len);
+      memcpy (copy, gomp_cpusetp, gomp_cpuset_size);
+      for (i = 0; i < max && gomp_places_list_len < count; i++)
+	if (CPU_ISSET_S (i, gomp_cpuset_size, copy))
+	  {
+	    sprintf (name + prefix_len, "%lu/topology/%s_siblings_list",
+		     i, level == 2 ? "thread" : "core");
+	    f = fopen (name, "r");
+	    if (f != NULL)
+	      {
+		if (getline (&line, &linelen, f) > 0)
+		  {
+		    char *p = line;
+		    bool seen_i = false;
+		    void *pl = gomp_places_list[gomp_places_list_len];
+		    gomp_affinity_init_place (pl);
+		    while (*p && *p != '\n')
+		      {
+			unsigned long first, last;
+			errno = 0;
+			first = strtoul (p, &p, 10);
+			if (errno)
+			  break;
+			last = first;
+			if (*p == '-')
+			  {
+			    errno = 0;
+			    last = strtoul (p + 1, &p, 10);
+			    if (errno || last < first)
+			      break;
+			  }
+			for (; first <= last; first++)
+			  if (CPU_ISSET_S (first, gomp_cpuset_size, copy)
+			      && gomp_affinity_add_cpus (pl, first, 1, 0,
+							 true))
+			    {
+			      CPU_CLR_S (first, gomp_cpuset_size, copy);
+			      if (first == i)
+				seen_i = true;
+			    }
+			if (*p == ',')
+			  ++p;
+		      }
+		    if (seen_i)
+		      gomp_places_list_len++;
+		  }
+		fclose (f);
+	      }
+	  }
+      if (gomp_places_list == 0)
 	{
-	  if (! CPU_ISSET (gomp_cpu_affinity[idx], &cpusetnew))
-	    {
-	      cpus++;
-	      CPU_SET (gomp_cpu_affinity[idx], &cpusetnew);
-	    }
-	  gomp_cpu_affinity[widx++] = gomp_cpu_affinity[idx];
+	  if (!quiet)
+	    gomp_error ("Error reading %s topology",
+			level == 2 ? "core" : "socket");
+	  free (gomp_places_list);
+	  gomp_places_list = NULL;
+	  return false;
 	}
-
-  if (widx == 0)
-    {
-      gomp_error ("no CPUs left for affinity setting");
-      free (gomp_cpu_affinity);
-      gomp_cpu_affinity = NULL;
-      gomp_cpu_affinity_len = 0;
-      return;
+      return true;
     }
-
-  gomp_cpu_affinity_len = widx;
-  if (cpus < gomp_available_cpus)
-    gomp_available_cpus = cpus;
-  CPU_ZERO (&cpuset);
-  CPU_SET (gomp_cpu_affinity[0], &cpuset);
-  pthread_setaffinity_np (pthread_self (), sizeof (cpuset), &cpuset);
-  affinity_counter = 1;
+  return false;
 }
 
 void
-gomp_init_thread_affinity (pthread_attr_t *attr)
+gomp_affinity_print_place (void *p)
 {
-  unsigned int cpu;
-  cpu_set_t cpuset;
-
-  cpu = __atomic_fetch_add (&affinity_counter, 1, MEMMODEL_RELAXED);
-  cpu %= gomp_cpu_affinity_len;
-  CPU_ZERO (&cpuset);
-  CPU_SET (gomp_cpu_affinity[cpu], &cpuset);
-  pthread_attr_setaffinity_np (attr, sizeof (cpu_set_t), &cpuset);
+  unsigned long i, max = 8 * gomp_cpuset_size, len;
+  cpu_set_t *cpusetp = (cpu_set_t *) p;
+  bool notfirst = false;
+
+  for (i = 0, len = 0; i < max; i++)
+    if (CPU_ISSET_S (i, gomp_cpuset_size, cpusetp))
+      {
+	if (len == 0)
+	  {
+	    if (notfirst)
+	      fputc (',', stderr);
+	    notfirst = true;
+	    fprintf (stderr, "%lu", i);
+	  }
+	++len;
+      }
+    else
+      {
+	if (len > 1)
+	  fprintf (stderr, ":%lu", len);
+	len = 0;
+      }
+  if (len > 1)
+    fprintf (stderr, ":%lu", len);
 }
 
 #else
diff --git a/libgomp/config/linux/bar.c b/libgomp/config/linux/bar.c
index 35baa886ab4..6b591e5a6c5 100644
--- a/libgomp/config/linux/bar.c
+++ b/libgomp/config/linux/bar.c
@@ -33,11 +33,11 @@
 void
 gomp_barrier_wait_end (gomp_barrier_t *bar, gomp_barrier_state_t state)
 {
-  if (__builtin_expect ((state & 1) != 0, 0))
+  if (__builtin_expect (state & BAR_WAS_LAST, 0))
     {
       /* Next time we'll be awaiting TOTAL threads again.  */
       bar->awaited = bar->total;
-      __atomic_store_n (&bar->generation, bar->generation + 4,
+      __atomic_store_n (&bar->generation, bar->generation + BAR_INCR,
 			MEMMODEL_RELEASE);
       futex_wake ((int *) &bar->generation, INT_MAX);
     }
@@ -66,7 +66,7 @@ void
 gomp_barrier_wait_last (gomp_barrier_t *bar)
 {
   gomp_barrier_state_t state = gomp_barrier_wait_start (bar);
-  if (state & 1)
+  if (state & BAR_WAS_LAST)
     gomp_barrier_wait_end (bar, state);
 }
 
@@ -81,40 +81,43 @@ gomp_team_barrier_wait_end (gomp_barrier_t *bar, gomp_barrier_state_t state)
 {
   unsigned int generation, gen;
 
-  if (__builtin_expect ((state & 1) != 0, 0))
+  if (__builtin_expect (state & BAR_WAS_LAST, 0))
     {
       /* Next time we'll be awaiting TOTAL threads again.  */
       struct gomp_thread *thr = gomp_thread ();
       struct gomp_team *team = thr->ts.team;
 
       bar->awaited = bar->total;
+      team->work_share_cancelled = 0;
       if (__builtin_expect (team->task_count, 0))
 	{
 	  gomp_barrier_handle_tasks (state);
-	  state &= ~1;
+	  state &= ~BAR_WAS_LAST;
 	}
       else
 	{
-	  __atomic_store_n (&bar->generation, state + 3, MEMMODEL_RELEASE);
+	  state &= ~BAR_CANCELLED;
+	  state += BAR_INCR - BAR_WAS_LAST;
+	  __atomic_store_n (&bar->generation, state, MEMMODEL_RELEASE);
 	  futex_wake ((int *) &bar->generation, INT_MAX);
 	  return;
 	}
     }
 
   generation = state;
+  state &= ~BAR_CANCELLED;
   do
     {
       do_wait ((int *) &bar->generation, generation);
       gen = __atomic_load_n (&bar->generation, MEMMODEL_ACQUIRE);
-      if (__builtin_expect (gen & 1, 0))
+      if (__builtin_expect (gen & BAR_TASK_PENDING, 0))
 	{
 	  gomp_barrier_handle_tasks (state);
 	  gen = __atomic_load_n (&bar->generation, MEMMODEL_ACQUIRE);
 	}
-      if ((gen & 2) != 0)
-	generation |= 2;
+      generation |= gen & BAR_WAITING_FOR_TASK;
     }
-  while (gen != state + 4);
+  while (gen != state + BAR_INCR);
 }
 
 void
@@ -122,3 +125,86 @@ gomp_team_barrier_wait (gomp_barrier_t *bar)
 {
   gomp_team_barrier_wait_end (bar, gomp_barrier_wait_start (bar));
 }
+
+void
+gomp_team_barrier_wait_final (gomp_barrier_t *bar)
+{
+  gomp_barrier_state_t state = gomp_barrier_wait_final_start (bar);
+  if (__builtin_expect (state & BAR_WAS_LAST, 0))
+    bar->awaited_final = bar->total;
+  gomp_team_barrier_wait_end (bar, state);
+}
+
+bool
+gomp_team_barrier_wait_cancel_end (gomp_barrier_t *bar,
+				   gomp_barrier_state_t state)
+{
+  unsigned int generation, gen;
+
+  if (__builtin_expect (state & BAR_WAS_LAST, 0))
+    {
+      /* Next time we'll be awaiting TOTAL threads again.  */
+      /* BAR_CANCELLED should never be set in state here, because
+	 cancellation means that at least one of the threads has been
+	 cancelled, thus on a cancellable barrier we should never see
+	 all threads to arrive.  */
+      struct gomp_thread *thr = gomp_thread ();
+      struct gomp_team *team = thr->ts.team;
+
+      bar->awaited = bar->total;
+      team->work_share_cancelled = 0;
+      if (__builtin_expect (team->task_count, 0))
+	{
+	  gomp_barrier_handle_tasks (state);
+	  state &= ~BAR_WAS_LAST;
+	}
+      else
+	{
+	  state += BAR_INCR - BAR_WAS_LAST;
+	  __atomic_store_n (&bar->generation, state, MEMMODEL_RELEASE);
+	  futex_wake ((int *) &bar->generation, INT_MAX);
+	  return false;
+	}
+    }
+
+  if (__builtin_expect (state & BAR_CANCELLED, 0))
+    return true;
+
+  generation = state;
+  do
+    {
+      do_wait ((int *) &bar->generation, generation);
+      gen = __atomic_load_n (&bar->generation, MEMMODEL_ACQUIRE);
+      if (__builtin_expect (gen & BAR_CANCELLED, 0))
+	return true;
+      if (__builtin_expect (gen & BAR_TASK_PENDING, 0))
+	{
+	  gomp_barrier_handle_tasks (state);
+	  gen = __atomic_load_n (&bar->generation, MEMMODEL_ACQUIRE);
+	}
+      generation |= gen & BAR_WAITING_FOR_TASK;
+    }
+  while (gen != state + BAR_INCR);
+
+  return false;
+}
+
+bool
+gomp_team_barrier_wait_cancel (gomp_barrier_t *bar)
+{
+  return gomp_team_barrier_wait_cancel_end (bar, gomp_barrier_wait_start (bar));
+}
+
+void
+gomp_team_barrier_cancel (struct gomp_team *team)
+{
+  gomp_mutex_lock (&team->task_lock);
+  if (team->barrier.generation & BAR_CANCELLED)
+    {
+      gomp_mutex_unlock (&team->task_lock);
+      return;
+    }
+  team->barrier.generation |= BAR_CANCELLED;
+  gomp_mutex_unlock (&team->task_lock);
+  futex_wake ((int *) &team->barrier.generation, INT_MAX);
+}
diff --git a/libgomp/config/linux/bar.h b/libgomp/config/linux/bar.h
index 69b97069647..914c86778e5 100644
--- a/libgomp/config/linux/bar.h
+++ b/libgomp/config/linux/bar.h
@@ -38,13 +38,25 @@ typedef struct
   unsigned total __attribute__((aligned (64)));
   unsigned generation;
   unsigned awaited __attribute__((aligned (64)));
+  unsigned awaited_final;
 } gomp_barrier_t;
+
 typedef unsigned int gomp_barrier_state_t;
 
+/* The generation field contains a counter in the high bits, with a few
+   low bits dedicated to flags.  Note that TASK_PENDING and WAS_LAST can
+   share space because WAS_LAST is never stored back to generation.  */
+#define BAR_TASK_PENDING	1
+#define BAR_WAS_LAST		1
+#define BAR_WAITING_FOR_TASK	2
+#define BAR_CANCELLED		4
+#define BAR_INCR		8
+
 static inline void gomp_barrier_init (gomp_barrier_t *bar, unsigned count)
 {
   bar->total = count;
   bar->awaited = count;
+  bar->awaited_final = count;
   bar->generation = 0;
 }
 
@@ -62,27 +74,55 @@ extern void gomp_barrier_wait (gomp_barrier_t *);
 extern void gomp_barrier_wait_last (gomp_barrier_t *);
 extern void gomp_barrier_wait_end (gomp_barrier_t *, gomp_barrier_state_t);
 extern void gomp_team_barrier_wait (gomp_barrier_t *);
+extern void gomp_team_barrier_wait_final (gomp_barrier_t *);
 extern void gomp_team_barrier_wait_end (gomp_barrier_t *,
 					gomp_barrier_state_t);
+extern bool gomp_team_barrier_wait_cancel (gomp_barrier_t *);
+extern bool gomp_team_barrier_wait_cancel_end (gomp_barrier_t *,
+					       gomp_barrier_state_t);
 extern void gomp_team_barrier_wake (gomp_barrier_t *, int);
+struct gomp_team;
+extern void gomp_team_barrier_cancel (struct gomp_team *);
 
 static inline gomp_barrier_state_t
 gomp_barrier_wait_start (gomp_barrier_t *bar)
 {
-  unsigned int ret = __atomic_load_n (&bar->generation, MEMMODEL_ACQUIRE) & ~3;
+  unsigned int ret = __atomic_load_n (&bar->generation, MEMMODEL_ACQUIRE);
+  ret &= -BAR_INCR | BAR_CANCELLED;
   /* A memory barrier is needed before exiting from the various forms
      of gomp_barrier_wait, to satisfy OpenMP API version 3.1 section
      2.8.6 flush Construct, which says there is an implicit flush during
      a barrier region.  This is a convenient place to add the barrier,
      so we use MEMMODEL_ACQ_REL here rather than MEMMODEL_ACQUIRE.  */
-  ret += __atomic_add_fetch (&bar->awaited, -1, MEMMODEL_ACQ_REL) == 0;
+  if (__atomic_add_fetch (&bar->awaited, -1, MEMMODEL_ACQ_REL) == 0)
+    ret |= BAR_WAS_LAST;
+  return ret;
+}
+
+static inline gomp_barrier_state_t
+gomp_barrier_wait_cancel_start (gomp_barrier_t *bar)
+{
+  return gomp_barrier_wait_start (bar);
+}
+
+/* This is like gomp_barrier_wait_start, except it decrements
+   bar->awaited_final rather than bar->awaited and should be used
+   for the gomp_team_end barrier only.  */
+static inline gomp_barrier_state_t
+gomp_barrier_wait_final_start (gomp_barrier_t *bar)
+{
+  unsigned int ret = __atomic_load_n (&bar->generation, MEMMODEL_ACQUIRE);
+  ret &= -BAR_INCR | BAR_CANCELLED;
+  /* See above gomp_barrier_wait_start comment.  */
+  if (__atomic_add_fetch (&bar->awaited_final, -1, MEMMODEL_ACQ_REL) == 0)
+    ret |= BAR_WAS_LAST;
   return ret;
 }
 
 static inline bool
 gomp_barrier_last_thread (gomp_barrier_state_t state)
 {
-  return state & 1;
+  return state & BAR_WAS_LAST;
 }
 
 /* All the inlines below must be called with team->task_lock
@@ -91,31 +131,37 @@ gomp_barrier_last_thread (gomp_barrier_state_t state)
 static inline void
 gomp_team_barrier_set_task_pending (gomp_barrier_t *bar)
 {
-  bar->generation |= 1;
+  bar->generation |= BAR_TASK_PENDING;
 }
 
 static inline void
 gomp_team_barrier_clear_task_pending (gomp_barrier_t *bar)
 {
-  bar->generation &= ~1;
+  bar->generation &= ~BAR_TASK_PENDING;
 }
 
 static inline void
 gomp_team_barrier_set_waiting_for_tasks (gomp_barrier_t *bar)
 {
-  bar->generation |= 2;
+  bar->generation |= BAR_WAITING_FOR_TASK;
 }
 
 static inline bool
 gomp_team_barrier_waiting_for_tasks (gomp_barrier_t *bar)
 {
-  return (bar->generation & 2) != 0;
+  return (bar->generation & BAR_WAITING_FOR_TASK) != 0;
+}
+
+static inline bool
+gomp_team_barrier_cancelled (gomp_barrier_t *bar)
+{
+  return __builtin_expect ((bar->generation & BAR_CANCELLED) != 0, 0);
 }
 
 static inline void
 gomp_team_barrier_done (gomp_barrier_t *bar, gomp_barrier_state_t state)
 {
-  bar->generation = (state & ~3) + 4;
+  bar->generation = (state & -BAR_INCR) + BAR_INCR;
 }
 
 #endif /* GOMP_BARRIER_H */
diff --git a/libgomp/config/linux/proc.c b/libgomp/config/linux/proc.c
index cbb773e6e90..d4ae116e239 100644
--- a/libgomp/config/linux/proc.c
+++ b/libgomp/config/linux/proc.c
@@ -30,6 +30,7 @@
 #endif
 #include "libgomp.h"
 #include "proc.h"
+#include <errno.h>
 #include <stdlib.h>
 #include <unistd.h>
 #ifdef HAVE_GETLOADAVG
@@ -39,19 +40,28 @@
 #endif
 
 #ifdef HAVE_PTHREAD_AFFINITY_NP
+unsigned long gomp_cpuset_size;
+static unsigned long gomp_get_cpuset_size;
+cpu_set_t *gomp_cpusetp;
+
 unsigned long
-gomp_cpuset_popcount (cpu_set_t *cpusetp)
+gomp_cpuset_popcount (unsigned long cpusetsize, cpu_set_t *cpusetp)
 {
-#ifdef CPU_COUNT
-  /* glibc 2.6 and above provide a macro for this.  */
-  return CPU_COUNT (cpusetp);
+#ifdef CPU_COUNT_S
+  /* glibc 2.7 and above provide a macro for this.  */
+  return CPU_COUNT_S (cpusetsize, cpusetp);
 #else
+#ifdef CPU_COUNT
+  if (cpusetsize == sizeof (cpu_set_t))
+    /* glibc 2.6 and above provide a macro for this.  */
+    return CPU_COUNT (cpusetp);
+#endif
   size_t i;
   unsigned long ret = 0;
-  extern int check[sizeof (cpusetp->__bits[0]) == sizeof (unsigned long int)];
+  extern int check[sizeof (cpusetp->__bits[0]) == sizeof (unsigned long int)
+		   ? 1 : -1];
 
-  (void) check;
-  for (i = 0; i < sizeof (*cpusetp) / sizeof (cpusetp->__bits[0]); i++)
+  for (i = 0; i < cpusetsize / sizeof (cpusetp->__bits[0]); i++)
     {
       unsigned long int mask = cpusetp->__bits[i];
       if (mask == 0)
@@ -70,16 +80,63 @@ void
 gomp_init_num_threads (void)
 {
 #ifdef HAVE_PTHREAD_AFFINITY_NP
-  cpu_set_t cpuset;
+#if defined (_SC_NPROCESSORS_CONF) && defined (CPU_ALLOC_SIZE)
+  gomp_cpuset_size = sysconf (_SC_NPROCESSORS_CONF);
+  gomp_cpuset_size = CPU_ALLOC_SIZE (gomp_cpuset_size);
+#else
+  gomp_cpuset_size = sizeof (cpu_set_t);
+#endif
 
-  if (pthread_getaffinity_np (pthread_self (), sizeof (cpuset), &cpuset) == 0)
+  gomp_cpusetp = (cpu_set_t *) gomp_malloc (gomp_cpuset_size);
+  do
     {
-      /* Count only the CPUs this process can use.  */
-      gomp_global_icv.nthreads_var = gomp_cpuset_popcount (&cpuset);
-      if (gomp_global_icv.nthreads_var == 0)
-	gomp_global_icv.nthreads_var = 1;
-      return;
+      int ret = pthread_getaffinity_np (pthread_self (), gomp_cpuset_size,
+					gomp_cpusetp);
+      if (ret == 0)
+	{
+	  unsigned long i;
+	  /* Count only the CPUs this process can use.  */
+	  gomp_global_icv.nthreads_var
+	    = gomp_cpuset_popcount (gomp_cpuset_size, gomp_cpusetp);
+	  if (gomp_global_icv.nthreads_var == 0)
+	    break;
+	  gomp_get_cpuset_size = gomp_cpuset_size;
+#ifdef CPU_ALLOC_SIZE
+	  for (i = gomp_cpuset_size * 8; i; i--)
+	    if (CPU_ISSET_S (i - 1, gomp_cpuset_size, gomp_cpusetp))
+	      break;
+	  gomp_cpuset_size = CPU_ALLOC_SIZE (i);
+#endif
+	  return;
+	}
+      if (ret != EINVAL)
+	break;
+#ifdef CPU_ALLOC_SIZE
+      if (gomp_cpuset_size < sizeof (cpu_set_t))
+	gomp_cpuset_size = sizeof (cpu_set_t);
+      else
+	gomp_cpuset_size = gomp_cpuset_size * 2;
+      if (gomp_cpuset_size < 8 * sizeof (cpu_set_t))
+	gomp_cpusetp
+	  = (cpu_set_t *) gomp_realloc (gomp_cpusetp, gomp_cpuset_size);
+      else
+	{
+	  /* Avoid gomp_fatal if too large memory allocation would be
+	     requested, e.g. kernel returning EINVAL all the time.  */
+	  void *p = realloc (gomp_cpusetp, gomp_cpuset_size);
+	  if (p == NULL)
+	    break;
+	  gomp_cpusetp = (cpu_set_t *) p;
+	}
+#else
+      break;
+#endif
     }
+  while (1);
+  gomp_cpuset_size = 0;
+  gomp_global_icv.nthreads_var = 1;
+  free (gomp_cpusetp);
+  gomp_cpusetp = NULL;
 #endif
 #ifdef _SC_NPROCESSORS_ONLN
   gomp_global_icv.nthreads_var = sysconf (_SC_NPROCESSORS_ONLN);
@@ -90,15 +147,14 @@ static int
 get_num_procs (void)
 {
 #ifdef HAVE_PTHREAD_AFFINITY_NP
-  cpu_set_t cpuset;
-
-  if (gomp_cpu_affinity == NULL)
+  if (gomp_places_list == NULL)
     {
       /* Count only the CPUs this process can use.  */
-      if (pthread_getaffinity_np (pthread_self (), sizeof (cpuset),
-				  &cpuset) == 0)
+      if (gomp_cpusetp
+	  && pthread_getaffinity_np (pthread_self (), gomp_get_cpuset_size,
+				     gomp_cpusetp) == 0)
 	{
-	  int ret = gomp_cpuset_popcount (&cpuset);
+	  int ret = gomp_cpuset_popcount (gomp_get_cpuset_size, gomp_cpusetp);
 	  return ret != 0 ? ret : 1;
 	}
     }
diff --git a/libgomp/config/linux/proc.h b/libgomp/config/linux/proc.h
index cba7f4a09e6..bdc85dba99e 100644
--- a/libgomp/config/linux/proc.h
+++ b/libgomp/config/linux/proc.h
@@ -28,7 +28,10 @@
 #include <sched.h>
 
 #ifdef HAVE_PTHREAD_AFFINITY_NP
-extern unsigned long gomp_cpuset_popcount (cpu_set_t *);
+extern unsigned long gomp_cpuset_size attribute_hidden;
+extern cpu_set_t *gomp_cpusetp attribute_hidden;
+extern unsigned long gomp_cpuset_popcount (unsigned long, cpu_set_t *)
+     attribute_hidden;
 #endif
 
 #endif /* GOMP_PROC_H */
diff --git a/libgomp/config/posix/affinity.c b/libgomp/config/posix/affinity.c
index ac3d14e830c..e7f97ab08d6 100644
--- a/libgomp/config/posix/affinity.c
+++ b/libgomp/config/posix/affinity.c
@@ -32,7 +32,84 @@ gomp_init_affinity (void)
 }
 
 void
-gomp_init_thread_affinity (pthread_attr_t *attr)
+gomp_init_thread_affinity (pthread_attr_t *attr, unsigned int place)
 {
   (void) attr;
+  (void) place;
+}
+
+void **
+gomp_affinity_alloc (unsigned long count, bool quiet)
+{
+  (void) count;
+  if (!quiet)
+    gomp_error ("Affinity not supported on this configuration");
+  return NULL;
+}
+
+void
+gomp_affinity_init_place (void *p)
+{
+  (void) p;
+}
+
+bool
+gomp_affinity_add_cpus (void *p, unsigned long num,
+			unsigned long len, long stride, bool quiet)
+{
+  (void) p;
+  (void) num;
+  (void) len;
+  (void) stride;
+  (void) quiet;
+  return false;
+}
+
+bool
+gomp_affinity_remove_cpu (void *p, unsigned long num)
+{
+  (void) p;
+  (void) num;
+  return false;
+}
+
+bool
+gomp_affinity_copy_place (void *p, void *q, long stride)
+{
+  (void) p;
+  (void) q;
+  (void) stride;
+  return false;
+}
+
+bool
+gomp_affinity_same_place (void *p, void *q)
+{
+  (void) p;
+  (void) q;
+  return false;
+}
+
+bool
+gomp_affinity_finalize_place_list (bool quiet)
+{
+  (void) quiet;
+  return false;
+}
+
+bool
+gomp_affinity_init_level (int level, unsigned long count, bool quiet)
+{
+  (void) level;
+  (void) count;
+  (void) quiet;
+  if (!quiet)
+    gomp_error ("Affinity not supported on this configuration");
+  return NULL;
+}
+
+void
+gomp_affinity_print_place (void *p)
+{
+  (void) p;
 }
diff --git a/libgomp/config/posix/bar.c b/libgomp/config/posix/bar.c
index 06a3185c286..bdf3978caee 100644
--- a/libgomp/config/posix/bar.c
+++ b/libgomp/config/posix/bar.c
@@ -42,6 +42,7 @@ gomp_barrier_init (gomp_barrier_t *bar, unsigned count)
   bar->total = count;
   bar->arrived = 0;
   bar->generation = 0;
+  bar->cancellable = false;
 }
 
 void
@@ -72,7 +73,7 @@ gomp_barrier_wait_end (gomp_barrier_t *bar, gomp_barrier_state_t state)
 {
   unsigned int n;
 
-  if (state & 1)
+  if (state & BAR_WAS_LAST)
     {
       n = --bar->arrived;
       if (n > 0)
@@ -113,12 +114,14 @@ gomp_team_barrier_wait_end (gomp_barrier_t *bar, gomp_barrier_state_t state)
 {
   unsigned int n;
 
-  if (state & 1)
+  state &= ~BAR_CANCELLED;
+  if (state & BAR_WAS_LAST)
     {
       n = --bar->arrived;
       struct gomp_thread *thr = gomp_thread ();
       struct gomp_team *team = thr->ts.team;
 
+      team->work_share_cancelled = 0;
       if (team->task_count)
 	{
 	  gomp_barrier_handle_tasks (state);
@@ -128,7 +131,7 @@ gomp_team_barrier_wait_end (gomp_barrier_t *bar, gomp_barrier_state_t state)
 	  return;
 	}
 
-      bar->generation = state + 3;
+      bar->generation = state + BAR_INCR - BAR_WAS_LAST;
       if (n > 0)
 	{
 	  do
@@ -141,13 +144,18 @@ gomp_team_barrier_wait_end (gomp_barrier_t *bar, gomp_barrier_state_t state)
   else
     {
       gomp_mutex_unlock (&bar->mutex1);
+      int gen;
       do
 	{
 	  gomp_sem_wait (&bar->sem1);
-	  if (bar->generation & 1)
-	    gomp_barrier_handle_tasks (state);
+	  gen = __atomic_load_n (&bar->generation, MEMMODEL_ACQUIRE);
+	  if (gen & BAR_TASK_PENDING)
+	    {
+	      gomp_barrier_handle_tasks (state);
+	      gen = __atomic_load_n (&bar->generation, MEMMODEL_ACQUIRE);
+	    }
 	}
-      while (bar->generation != state + 4);
+      while (gen != state + BAR_INCR);
 
 #ifdef HAVE_SYNC_BUILTINS
       n = __sync_add_and_fetch (&bar->arrived, -1);
@@ -162,6 +170,81 @@ gomp_team_barrier_wait_end (gomp_barrier_t *bar, gomp_barrier_state_t state)
     }
 }
 
+bool
+gomp_team_barrier_wait_cancel_end (gomp_barrier_t *bar,
+				   gomp_barrier_state_t state)
+{
+  unsigned int n;
+
+  if (state & BAR_WAS_LAST)
+    {
+      bar->cancellable = false;
+      n = --bar->arrived;
+      struct gomp_thread *thr = gomp_thread ();
+      struct gomp_team *team = thr->ts.team;
+
+      team->work_share_cancelled = 0;
+      if (team->task_count)
+	{
+	  gomp_barrier_handle_tasks (state);
+	  if (n > 0)
+	    gomp_sem_wait (&bar->sem2);
+	  gomp_mutex_unlock (&bar->mutex1);
+	  return false;
+	}
+
+      bar->generation = state + BAR_INCR - BAR_WAS_LAST;
+      if (n > 0)
+	{
+	  do
+	    gomp_sem_post (&bar->sem1);
+	  while (--n != 0);
+	  gomp_sem_wait (&bar->sem2);
+	}
+      gomp_mutex_unlock (&bar->mutex1);
+    }
+  else
+    {
+      if (state & BAR_CANCELLED)
+	{
+	  gomp_mutex_unlock (&bar->mutex1);
+	  return true;
+	}
+      bar->cancellable = true;
+      gomp_mutex_unlock (&bar->mutex1);
+      int gen;
+      do
+	{
+	  gomp_sem_wait (&bar->sem1);
+	  gen = __atomic_load_n (&bar->generation, MEMMODEL_ACQUIRE);
+	  if (gen & BAR_CANCELLED)
+	    break;
+	  if (gen & BAR_TASK_PENDING)
+	    {
+	      gomp_barrier_handle_tasks (state);
+	      gen = __atomic_load_n (&bar->generation, MEMMODEL_ACQUIRE);
+	      if (gen & BAR_CANCELLED)
+		break;
+	    }
+	}
+      while (gen != state + BAR_INCR);
+
+#ifdef HAVE_SYNC_BUILTINS
+      n = __sync_add_and_fetch (&bar->arrived, -1);
+#else
+      gomp_mutex_lock (&bar->mutex2);
+      n = --bar->arrived;
+      gomp_mutex_unlock (&bar->mutex2);
+#endif
+
+      if (n == 0)
+	gomp_sem_post (&bar->sem2);
+      if (gen & BAR_CANCELLED)
+	return true;
+    }
+  return false;
+}
+
 void
 gomp_team_barrier_wait (gomp_barrier_t *barrier)
 {
@@ -176,3 +259,40 @@ gomp_team_barrier_wake (gomp_barrier_t *bar, int count)
   while (count-- > 0)
     gomp_sem_post (&bar->sem1);
 }
+
+bool
+gomp_team_barrier_wait_cancel (gomp_barrier_t *bar)
+{
+  gomp_barrier_state_t state = gomp_barrier_wait_cancel_start (bar);
+  return gomp_team_barrier_wait_cancel_end (bar, state);
+}
+
+void
+gomp_team_barrier_cancel (struct gomp_team *team)
+{
+  if (team->barrier.generation & BAR_CANCELLED)
+    return;
+  gomp_mutex_lock (&team->barrier.mutex1);
+  gomp_mutex_lock (&team->task_lock);
+  if (team->barrier.generation & BAR_CANCELLED)
+    {
+      gomp_mutex_unlock (&team->task_lock);
+      gomp_mutex_unlock (&team->barrier.mutex1);
+      return;
+    }
+  team->barrier.generation |= BAR_CANCELLED;
+  gomp_mutex_unlock (&team->task_lock);
+  if (team->barrier.cancellable)
+    {
+      int n = team->barrier.arrived;
+      if (n > 0)
+	{
+	  do
+	    gomp_sem_post (&team->barrier.sem1);
+	  while (--n != 0);
+	  gomp_sem_wait (&team->barrier.sem2);
+	}
+      team->barrier.cancellable = false;
+    }
+  gomp_mutex_unlock (&team->barrier.mutex1);
+}
diff --git a/libgomp/config/posix/bar.h b/libgomp/config/posix/bar.h
index 1a16ca86bba..9fcd4da6723 100644
--- a/libgomp/config/posix/bar.h
+++ b/libgomp/config/posix/bar.h
@@ -43,9 +43,20 @@ typedef struct
   unsigned total;
   unsigned arrived;
   unsigned generation;
+  bool cancellable;
 } gomp_barrier_t;
+
 typedef unsigned int gomp_barrier_state_t;
 
+/* The generation field contains a counter in the high bits, with a few
+   low bits dedicated to flags.  Note that TASK_PENDING and WAS_LAST can
+   share space because WAS_LAST is never stored back to generation.  */
+#define BAR_TASK_PENDING	1
+#define BAR_WAS_LAST		1
+#define BAR_WAITING_FOR_TASK	2
+#define BAR_CANCELLED		4
+#define BAR_INCR		8
+
 extern void gomp_barrier_init (gomp_barrier_t *, unsigned);
 extern void gomp_barrier_reinit (gomp_barrier_t *, unsigned);
 extern void gomp_barrier_destroy (gomp_barrier_t *);
@@ -55,22 +66,47 @@ extern void gomp_barrier_wait_end (gomp_barrier_t *, gomp_barrier_state_t);
 extern void gomp_team_barrier_wait (gomp_barrier_t *);
 extern void gomp_team_barrier_wait_end (gomp_barrier_t *,
 					gomp_barrier_state_t);
+extern bool gomp_team_barrier_wait_cancel (gomp_barrier_t *);
+extern bool gomp_team_barrier_wait_cancel_end (gomp_barrier_t *,
+					       gomp_barrier_state_t);
 extern void gomp_team_barrier_wake (gomp_barrier_t *, int);
+struct gomp_team;
+extern void gomp_team_barrier_cancel (struct gomp_team *);
 
 static inline gomp_barrier_state_t
 gomp_barrier_wait_start (gomp_barrier_t *bar)
 {
   unsigned int ret;
   gomp_mutex_lock (&bar->mutex1);
-  ret = bar->generation & ~3;
-  ret += ++bar->arrived == bar->total;
+  ret = bar->generation & (-BAR_INCR | BAR_CANCELLED);
+  if (++bar->arrived == bar->total)
+    ret |= BAR_WAS_LAST;
+  return ret;
+}
+
+static inline gomp_barrier_state_t
+gomp_barrier_wait_cancel_start (gomp_barrier_t *bar)
+{
+  unsigned int ret;
+  gomp_mutex_lock (&bar->mutex1);
+  ret = bar->generation & (-BAR_INCR | BAR_CANCELLED);
+  if (ret & BAR_CANCELLED)
+    return ret;
+  if (++bar->arrived == bar->total)
+    ret |= BAR_WAS_LAST;
   return ret;
 }
 
+static inline void
+gomp_team_barrier_wait_final (gomp_barrier_t *bar)
+{
+  gomp_team_barrier_wait (bar);
+}
+
 static inline bool
 gomp_barrier_last_thread (gomp_barrier_state_t state)
 {
-  return state & 1;
+  return state & BAR_WAS_LAST;
 }
 
 static inline void
@@ -85,31 +121,37 @@ gomp_barrier_wait_last (gomp_barrier_t *bar)
 static inline void
 gomp_team_barrier_set_task_pending (gomp_barrier_t *bar)
 {
-  bar->generation |= 1;
+  bar->generation |= BAR_TASK_PENDING;
 }
 
 static inline void
 gomp_team_barrier_clear_task_pending (gomp_barrier_t *bar)
 {
-  bar->generation &= ~1;
+  bar->generation &= ~BAR_TASK_PENDING;
 }
 
 static inline void
 gomp_team_barrier_set_waiting_for_tasks (gomp_barrier_t *bar)
 {
-  bar->generation |= 2;
+  bar->generation |= BAR_WAITING_FOR_TASK;
 }
 
 static inline bool
 gomp_team_barrier_waiting_for_tasks (gomp_barrier_t *bar)
 {
-  return (bar->generation & 2) != 0;
+  return (bar->generation & BAR_WAITING_FOR_TASK) != 0;
+}
+
+static inline bool
+gomp_team_barrier_cancelled (gomp_barrier_t *bar)
+{
+  return __builtin_expect ((bar->generation & BAR_CANCELLED) != 0, 0);
 }
 
 static inline void
 gomp_team_barrier_done (gomp_barrier_t *bar, gomp_barrier_state_t state)
 {
-  bar->generation = (state & ~3) + 4;
+  bar->generation = (state & -BAR_INCR) + BAR_INCR;
 }
 
 #endif /* GOMP_BARRIER_H */