Hi Jan, On 8/25/07, Jan Ciger <[EMAIL PROTECTED]> wrote: > Yes, that has me puzzled too, but you have to keep in mind that Linux > threading is implemented in separate processes. There isn't much > difference between a process and a thread on Linux. Even on a single > core machine one thread holding a lock could be scheduled away from the > CPU while another one is executed. On a multicore machine this only > happens sooner.
Its not quite the same w.r.t single core and multi-core, on a single core machine each thread/process gets a scheduled block on time CPU and during this period nothing else runs, nothing else can read and write to memory. In a multi-core system things properly happen in parallel - so contention over reading and writing of memory become a much greater issue, this is why its not uncommon to find multi-threaded code that runs "robustly" on a single core breaks once you throw more cores at it. > If the multicore system is fast enough to run through the critical > section causing the deadlock before the other thread has a chance to > acquire the offending lock, you will not see the bug. I guess there is chance that is happening, but I would expect that occasionally we'd see hangs on multi-core systems if timing was an issue. > OK, let me know if you find something out. Haven't tested yet... Robert. _______________________________________________ osg-users mailing list [email protected] http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
