> Here's the detailed write up I did that describes the SyncFence > semantics and how those map directly to Linux Futexes: [...]
This makes SyncFences sound like condition variables under another name, to me. Surely pthreads has condition variables? (I don't know pthreads....) However, I'd give a very close to look to the synchronization code. SyncFences look like condition variables _without_ the associated mutexes. The mutex is there for very good reason, though, leading me to suspect that SyncFence-using code is full of race conditions. (Suspect. Since I haven't read it, I'm not sure; there probably exist uses for mutex-free condvars, and this may be one of them....) In any case, the mutex does not have to actually cause any trouble, if present; it can be acquired and released immediately around the condvar calls. /~\ The ASCII Mouse \ / Ribbon Campaign X Against HTML mo...@rodents-montreal.org / \ Email! 7D C8 61 52 5D E7 2D 39 4E F1 31 3E E8 B3 27 4B _______________________________________________ xorg-devel@lists.x.org: X.Org development Archives: http://lists.x.org/archives/xorg-devel Info: http://lists.x.org/mailman/listinfo/xorg-devel