/* Copyright (C) 2005, 2007, 2008, 2009 Free Software Foundation, Inc.
|
/* Copyright (C) 2005, 2007, 2008, 2009 Free Software Foundation, Inc.
|
Contributed by Richard Henderson <rth@redhat.com>.
|
Contributed by Richard Henderson <rth@redhat.com>.
|
|
|
This file is part of the GNU OpenMP Library (libgomp).
|
This file is part of the GNU OpenMP Library (libgomp).
|
|
|
Libgomp is free software; you can redistribute it and/or modify it
|
Libgomp is free software; you can redistribute it and/or modify it
|
under the terms of the GNU General Public License as published by
|
under the terms of the GNU General Public License as published by
|
the Free Software Foundation; either version 3, or (at your option)
|
the Free Software Foundation; either version 3, or (at your option)
|
any later version.
|
any later version.
|
|
|
Libgomp is distributed in the hope that it will be useful, but WITHOUT ANY
|
Libgomp is distributed in the hope that it will be useful, but WITHOUT ANY
|
WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
|
WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
|
FOR A PARTICULAR PURPOSE. See the GNU General Public License for
|
FOR A PARTICULAR PURPOSE. See the GNU General Public License for
|
more details.
|
more details.
|
|
|
Under Section 7 of GPL version 3, you are granted additional
|
Under Section 7 of GPL version 3, you are granted additional
|
permissions described in the GCC Runtime Library Exception, version
|
permissions described in the GCC Runtime Library Exception, version
|
3.1, as published by the Free Software Foundation.
|
3.1, as published by the Free Software Foundation.
|
|
|
You should have received a copy of the GNU General Public License and
|
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;
|
a copy of the GCC Runtime Library Exception along with this program;
|
see the files COPYING3 and COPYING.RUNTIME respectively. If not, see
|
see the files COPYING3 and COPYING.RUNTIME respectively. If not, see
|
<http://www.gnu.org/licenses/>. */
|
<http://www.gnu.org/licenses/>. */
|
|
|
/* This file handles the (bare) PARALLEL construct. */
|
/* This file handles the (bare) PARALLEL construct. */
|
|
|
#include "libgomp.h"
|
#include "libgomp.h"
|
#include <limits.h>
|
#include <limits.h>
|
|
|
|
|
/* Determine the number of threads to be launched for a PARALLEL construct.
|
/* Determine the number of threads to be launched for a PARALLEL construct.
|
This algorithm is explicitly described in OpenMP 3.0 section 2.4.1.
|
This algorithm is explicitly described in OpenMP 3.0 section 2.4.1.
|
SPECIFIED is a combination of the NUM_THREADS clause and the IF clause.
|
SPECIFIED is a combination of the NUM_THREADS clause and the IF clause.
|
If the IF clause is false, SPECIFIED is forced to 1. When NUM_THREADS
|
If the IF clause is false, SPECIFIED is forced to 1. When NUM_THREADS
|
is not present, SPECIFIED is 0. */
|
is not present, SPECIFIED is 0. */
|
|
|
unsigned
|
unsigned
|
gomp_resolve_num_threads (unsigned specified, unsigned count)
|
gomp_resolve_num_threads (unsigned specified, unsigned count)
|
{
|
{
|
struct gomp_thread *thread = gomp_thread();
|
struct gomp_thread *thread = gomp_thread();
|
struct gomp_task_icv *icv;
|
struct gomp_task_icv *icv;
|
unsigned threads_requested, max_num_threads, num_threads;
|
unsigned threads_requested, max_num_threads, num_threads;
|
unsigned long remaining;
|
unsigned long remaining;
|
|
|
icv = gomp_icv (false);
|
icv = gomp_icv (false);
|
|
|
if (specified == 1)
|
if (specified == 1)
|
return 1;
|
return 1;
|
else if (thread->ts.active_level >= 1 && !icv->nest_var)
|
else if (thread->ts.active_level >= 1 && !icv->nest_var)
|
return 1;
|
return 1;
|
else if (thread->ts.active_level >= gomp_max_active_levels_var)
|
else if (thread->ts.active_level >= gomp_max_active_levels_var)
|
return 1;
|
return 1;
|
|
|
/* If NUM_THREADS not specified, use nthreads_var. */
|
/* If NUM_THREADS not specified, use nthreads_var. */
|
if (specified == 0)
|
if (specified == 0)
|
threads_requested = icv->nthreads_var;
|
threads_requested = icv->nthreads_var;
|
else
|
else
|
threads_requested = specified;
|
threads_requested = specified;
|
|
|
max_num_threads = threads_requested;
|
max_num_threads = threads_requested;
|
|
|
/* If dynamic threads are enabled, bound the number of threads
|
/* If dynamic threads are enabled, bound the number of threads
|
that we launch. */
|
that we launch. */
|
if (icv->dyn_var)
|
if (icv->dyn_var)
|
{
|
{
|
unsigned dyn = gomp_dynamic_max_threads ();
|
unsigned dyn = gomp_dynamic_max_threads ();
|
if (dyn < max_num_threads)
|
if (dyn < max_num_threads)
|
max_num_threads = dyn;
|
max_num_threads = dyn;
|
|
|
/* Optimization for parallel sections. */
|
/* Optimization for parallel sections. */
|
if (count && count < max_num_threads)
|
if (count && count < max_num_threads)
|
max_num_threads = count;
|
max_num_threads = count;
|
}
|
}
|
|
|
/* ULONG_MAX stands for infinity. */
|
/* ULONG_MAX stands for infinity. */
|
if (__builtin_expect (gomp_thread_limit_var == ULONG_MAX, 1)
|
if (__builtin_expect (gomp_thread_limit_var == ULONG_MAX, 1)
|
|| max_num_threads == 1)
|
|| max_num_threads == 1)
|
return max_num_threads;
|
return max_num_threads;
|
|
|
#ifdef HAVE_SYNC_BUILTINS
|
#ifdef HAVE_SYNC_BUILTINS
|
do
|
do
|
{
|
{
|
remaining = gomp_remaining_threads_count;
|
remaining = gomp_remaining_threads_count;
|
num_threads = max_num_threads;
|
num_threads = max_num_threads;
|
if (num_threads > remaining)
|
if (num_threads > remaining)
|
num_threads = remaining + 1;
|
num_threads = remaining + 1;
|
}
|
}
|
while (__sync_val_compare_and_swap (&gomp_remaining_threads_count,
|
while (__sync_val_compare_and_swap (&gomp_remaining_threads_count,
|
remaining, remaining - num_threads + 1)
|
remaining, remaining - num_threads + 1)
|
!= remaining);
|
!= remaining);
|
#else
|
#else
|
gomp_mutex_lock (&gomp_remaining_threads_lock);
|
gomp_mutex_lock (&gomp_remaining_threads_lock);
|
num_threads = max_num_threads;
|
num_threads = max_num_threads;
|
remaining = gomp_remaining_threads_count;
|
remaining = gomp_remaining_threads_count;
|
if (num_threads > remaining)
|
if (num_threads > remaining)
|
num_threads = remaining + 1;
|
num_threads = remaining + 1;
|
gomp_remaining_threads_count -= num_threads - 1;
|
gomp_remaining_threads_count -= num_threads - 1;
|
gomp_mutex_unlock (&gomp_remaining_threads_lock);
|
gomp_mutex_unlock (&gomp_remaining_threads_lock);
|
#endif
|
#endif
|
|
|
return num_threads;
|
return num_threads;
|
}
|
}
|
|
|
void
|
void
|
GOMP_parallel_start (void (*fn) (void *), void *data, unsigned num_threads)
|
GOMP_parallel_start (void (*fn) (void *), void *data, unsigned num_threads)
|
{
|
{
|
num_threads = gomp_resolve_num_threads (num_threads, 0);
|
num_threads = gomp_resolve_num_threads (num_threads, 0);
|
gomp_team_start (fn, data, num_threads, gomp_new_team (num_threads));
|
gomp_team_start (fn, data, num_threads, gomp_new_team (num_threads));
|
}
|
}
|
|
|
void
|
void
|
GOMP_parallel_end (void)
|
GOMP_parallel_end (void)
|
{
|
{
|
if (__builtin_expect (gomp_thread_limit_var != ULONG_MAX, 0))
|
if (__builtin_expect (gomp_thread_limit_var != ULONG_MAX, 0))
|
{
|
{
|
struct gomp_thread *thr = gomp_thread ();
|
struct gomp_thread *thr = gomp_thread ();
|
struct gomp_team *team = thr->ts.team;
|
struct gomp_team *team = thr->ts.team;
|
if (team && team->nthreads > 1)
|
if (team && team->nthreads > 1)
|
{
|
{
|
#ifdef HAVE_SYNC_BUILTINS
|
#ifdef HAVE_SYNC_BUILTINS
|
__sync_fetch_and_add (&gomp_remaining_threads_count,
|
__sync_fetch_and_add (&gomp_remaining_threads_count,
|
1UL - team->nthreads);
|
1UL - team->nthreads);
|
#else
|
#else
|
gomp_mutex_lock (&gomp_remaining_threads_lock);
|
gomp_mutex_lock (&gomp_remaining_threads_lock);
|
gomp_remaining_threads_count -= team->nthreads - 1;
|
gomp_remaining_threads_count -= team->nthreads - 1;
|
#endif
|
#endif
|
}
|
}
|
}
|
}
|
gomp_team_end ();
|
gomp_team_end ();
|
}
|
}
|
|
|
|
|
/* The public OpenMP API for thread and team related inquiries. */
|
/* The public OpenMP API for thread and team related inquiries. */
|
|
|
int
|
int
|
omp_get_num_threads (void)
|
omp_get_num_threads (void)
|
{
|
{
|
struct gomp_team *team = gomp_thread ()->ts.team;
|
struct gomp_team *team = gomp_thread ()->ts.team;
|
return team ? team->nthreads : 1;
|
return team ? team->nthreads : 1;
|
}
|
}
|
|
|
int
|
int
|
omp_get_thread_num (void)
|
omp_get_thread_num (void)
|
{
|
{
|
return gomp_thread ()->ts.team_id;
|
return gomp_thread ()->ts.team_id;
|
}
|
}
|
|
|
/* This wasn't right for OpenMP 2.5. Active region used to be non-zero
|
/* This wasn't right for OpenMP 2.5. Active region used to be non-zero
|
when the IF clause doesn't evaluate to false, starting with OpenMP 3.0
|
when the IF clause doesn't evaluate to false, starting with OpenMP 3.0
|
it is non-zero with more than one thread in the team. */
|
it is non-zero with more than one thread in the team. */
|
|
|
int
|
int
|
omp_in_parallel (void)
|
omp_in_parallel (void)
|
{
|
{
|
return gomp_thread ()->ts.active_level > 0;
|
return gomp_thread ()->ts.active_level > 0;
|
}
|
}
|
|
|
int
|
int
|
omp_get_level (void)
|
omp_get_level (void)
|
{
|
{
|
return gomp_thread ()->ts.level;
|
return gomp_thread ()->ts.level;
|
}
|
}
|
|
|
int
|
int
|
omp_get_ancestor_thread_num (int level)
|
omp_get_ancestor_thread_num (int level)
|
{
|
{
|
struct gomp_team_state *ts = &gomp_thread ()->ts;
|
struct gomp_team_state *ts = &gomp_thread ()->ts;
|
if (level < 0 || level > ts->level)
|
if (level < 0 || level > ts->level)
|
return -1;
|
return -1;
|
for (level = ts->level - level; level > 0; --level)
|
for (level = ts->level - level; level > 0; --level)
|
ts = &ts->team->prev_ts;
|
ts = &ts->team->prev_ts;
|
return ts->team_id;
|
return ts->team_id;
|
}
|
}
|
|
|
int
|
int
|
omp_get_team_size (int level)
|
omp_get_team_size (int level)
|
{
|
{
|
struct gomp_team_state *ts = &gomp_thread ()->ts;
|
struct gomp_team_state *ts = &gomp_thread ()->ts;
|
if (level < 0 || level > ts->level)
|
if (level < 0 || level > ts->level)
|
return -1;
|
return -1;
|
for (level = ts->level - level; level > 0; --level)
|
for (level = ts->level - level; level > 0; --level)
|
ts = &ts->team->prev_ts;
|
ts = &ts->team->prev_ts;
|
if (ts->team == NULL)
|
if (ts->team == NULL)
|
return 1;
|
return 1;
|
else
|
else
|
return ts->team->nthreads;
|
return ts->team->nthreads;
|
}
|
}
|
|
|
int
|
int
|
omp_get_active_level (void)
|
omp_get_active_level (void)
|
{
|
{
|
return gomp_thread ()->ts.active_level;
|
return gomp_thread ()->ts.active_level;
|
}
|
}
|
|
|
ialias (omp_get_num_threads)
|
ialias (omp_get_num_threads)
|
ialias (omp_get_thread_num)
|
ialias (omp_get_thread_num)
|
ialias (omp_in_parallel)
|
ialias (omp_in_parallel)
|
ialias (omp_get_level)
|
ialias (omp_get_level)
|
ialias (omp_get_ancestor_thread_num)
|
ialias (omp_get_ancestor_thread_num)
|
ialias (omp_get_team_size)
|
ialias (omp_get_team_size)
|
ialias (omp_get_active_level)
|
ialias (omp_get_active_level)
|
|
|
