diff options
| author | rpj <rpj> | 2011-03-30 13:41:17 +0000 |
|---|---|---|
| committer | rpj <rpj> | 2011-03-30 13:41:17 +0000 |
| commit | 17933a9b9276debeda3d1ee1326d6b3c8a0bba56 (patch) | |
| tree | d682f683cec49dea402ba849588992f00ca434b6 | |
| parent | e07da6f5a812f95799de4f361faa4db47f77526a (diff) | |
*** empty log message ***
| -rw-r--r-- | NEWS | 26 | ||||
| -rw-r--r-- | tests/SIZES.VSE | 40 |
2 files changed, 31 insertions, 35 deletions
@@ -44,16 +44,11 @@ POSIX TSD (TLS) destructors. See README.NONPORTABLE for descriptions of these routines. - Ramiro Polla -Robust mutexes are implemented within the PROCESS_PRIVATE scope. Robust -mutexes are mandatory (SUSv3 2008) for implementations that support -PSHARED objects and discretionary for those that don't. We still don't -have PSHARED support but wanted to be source code compatible and they -will hopefully nevertheless be useful for some applications. NOTE that -pthread_mutex_*() functions may return different values for robust -mutexes than they otherwise do in normal usage, e.g. -pthread_mutex_unlock() is required to check ownership for all mutex -types when the mutex is robust, whereas this does not occur for the -"normal" non-robust mutex type. +Robust mutexes are implemented within the PROCESS_PRIVATE scope. NOTE that +pthread_mutex_* functions may return different error codes for robust +mutexes than they otherwise do in normal usage, e.g. pthread_mutex_unlock +is required to check ownership for all mutex types when the mutex is +robust, whereas this does not occur for the "normal" non-robust mutex type. - Ross Johnson pthread_getunique_np is implemented for source level compatibility @@ -88,11 +83,10 @@ Removed potential NULL pointer reference. - Robert Kindred Removed the requirement that applications restrict the number of threads -calling pthread_barrier_wait() [on the same barrier] to just the barrier -count in order to avoid contention and dead lock. Also reduced the -contention between barrier_wait and barrier_destroy. This change will -have slowed barriers down slightly but halves the number of semaphores -consumed per barrier to one. +calling pthread_barrier_wait to just the barrier count. Also reduced the +contention between barrier_wait and barrier_destroy. This change will have +slowed barriers down slightly but halves the number of semaphores consumed +per barrier to one. - Ross Johnson Fixed a handle leak in sched_[gs]etscheduler. @@ -115,7 +109,7 @@ WINCE is defined (because RJ is unsure of the dependency there). - Ramiro Polla Several static POSIX mutexes used for internal management were replaced by -MCS queue-based locks to reduce resource overheads, particularly Win32 +MCS queue-based locks to reduce resource consumption, in particular use of Win32 objects. - Ross Johnson diff --git a/tests/SIZES.VSE b/tests/SIZES.VSE index edc6427..709114d 100644 --- a/tests/SIZES.VSE +++ b/tests/SIZES.VSE @@ -1,19 +1,21 @@ -Sizes of pthreads-win32 structs -------------------------------- - pthread_t_ 68 - pthread_attr_t_ 28 - sem_t_ 4 - pthread_mutex_t_ 44 - pthread_mutexattr_t_ 8 - pthread_spinlock_t_ 8 - pthread_barrier_t_ 24 - pthread_barrierattr_t_ 4 - pthread_key_t_ 16 - pthread_cond_t_ 32 - pthread_condattr_t_ 4 - pthread_rwlock_t_ 28 - pthread_rwlockattr_t_ 4 - pthread_once_t_ 8 - ptw32_cleanup_t 12 - sched_param 4 -------------------------------- +Sizes of pthreads-win32 structs
+-------------------------------
+ pthread_t 8
+ ptw32_thread_t 96
+ pthread_attr_t_ 28
+ sem_t_ 12
+ pthread_mutex_t_ 28
+ pthread_mutexattr_t_ 12
+ pthread_spinlock_t_ 8
+ pthread_barrier_t_ 36
+ pthread_barrierattr_t_ 4
+ pthread_key_t_ 16
+ pthread_cond_t_ 32
+ pthread_condattr_t_ 4
+ pthread_rwlock_t_ 28
+ pthread_rwlockattr_t_ 4
+ pthread_once_t_ 16
+ ptw32_cleanup_t 12
+ ptw32_mcs_node_t_ 16
+ sched_param 4
+-------------------------------
|