1 files changed, 61 insertions, 15 deletions

diff --git a/libgomp/config/linux/bar.h b/libgomp/config/linux/bar.h
index 57268585d8b..85150caf189 100644
--- a/libgomp/config/linux/bar.h
+++ b/libgomp/config/linux/bar.h
@@ -1,4 +1,4 @@
-/* Copyright (C) 2005 Free Software Foundation, Inc.
+/* Copyright (C) 2005, 2008 Free Software Foundation, Inc.
    Contributed by Richard Henderson <rth@redhat.com>.
 
    This file is part of the GNU OpenMP Library (libgomp).
@@ -36,40 +36,86 @@
 
 typedef struct
 {
-  gomp_mutex_t mutex;
-  unsigned total;
-  unsigned arrived;
-  int generation;
+  /* Make sure total/generation is in a mostly read cacheline, while
+     awaited in a separate cacheline.  */
+  unsigned total __attribute__((aligned (64)));
+  unsigned generation;
+  unsigned awaited __attribute__((aligned (64)));
 } gomp_barrier_t;
+typedef unsigned int gomp_barrier_state_t;
 
 static inline void gomp_barrier_init (gomp_barrier_t *bar, unsigned count)
 {
-  gomp_mutex_init (&bar->mutex);
   bar->total = count;
-  bar->arrived = 0;
+  bar->awaited = count;
   bar->generation = 0;
 }
 
 static inline void gomp_barrier_reinit (gomp_barrier_t *bar, unsigned count)
 {
-  gomp_mutex_lock (&bar->mutex);
+  __sync_fetch_and_add (&bar->awaited, count - bar->total);
   bar->total = count;
-  gomp_mutex_unlock (&bar->mutex);
 }
 
 static inline void gomp_barrier_destroy (gomp_barrier_t *bar)
 {
-  /* Before destroying, make sure all threads have left the barrier.  */
-  gomp_mutex_lock (&bar->mutex);
 }
 
 extern void gomp_barrier_wait (gomp_barrier_t *);
-extern void gomp_barrier_wait_end (gomp_barrier_t *, bool);
+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_end (gomp_barrier_t *,
+					gomp_barrier_state_t);
+extern void gomp_team_barrier_wake (gomp_barrier_t *, int);
 
-static inline bool gomp_barrier_wait_start (gomp_barrier_t *bar)
+static inline gomp_barrier_state_t
+gomp_barrier_wait_start (gomp_barrier_t *bar)
 {
-  gomp_mutex_lock (&bar->mutex);
-  return ++bar->arrived == bar->total;
+  unsigned int ret = bar->generation & ~3;
+  /* Do we need any barrier here or is __sync_add_and_fetch acting
+     as the needed LoadLoad barrier already?  */
+  ret += __sync_add_and_fetch (&bar->awaited, -1) == 0;
+  return ret;
+}
+
+static inline bool
+gomp_barrier_last_thread (gomp_barrier_state_t state)
+{
+  return state & 1;
+}
+
+/* All the inlines below must be called with team->task_lock
+   held.  */
+
+static inline void
+gomp_team_barrier_set_task_pending (gomp_barrier_t *bar)
+{
+  bar->generation |= 1;
+}
+
+static inline void
+gomp_team_barrier_clear_task_pending (gomp_barrier_t *bar)
+{
+  bar->generation &= ~1;
+}
+
+static inline void
+gomp_team_barrier_set_waiting_for_tasks (gomp_barrier_t *bar)
+{
+  bar->generation |= 2;
+}
+
+static inline bool
+gomp_team_barrier_waiting_for_tasks (gomp_barrier_t *bar)
+{
+  return (bar->generation & 2) != 0;
+}
+
+static inline void
+gomp_team_barrier_done (gomp_barrier_t *bar, gomp_barrier_state_t state)
+{
+  bar->generation = (state & ~3) + 4;
 }
 
 #endif /* GOMP_BARRIER_H */