José Orlando Pereira napsal(a): > On Friday 19 April 2013 12:36:15 Jan Friesse wrote: >> Jose, >> I was thinking about that problem a little deeper and what about >> implementing this: >> - Store buffer in cpg_inst (so alloc in cpg_initialize, free in >> cpg_finalize) >> - When dispatch is called, set variable in cpg_inst (something like >> is_in_dispatch) before callback and unset it after callback. >> - IF is_in_dispatch is set, allocate new buffer, else use buffer from >> cpg_inst >> >> Basically, this should solve all problems with (in practice) no perf >> hit, because 99.9999% code doesn't reenter *_dispatch function, and make >> rest of 0.0001% code work correctly. > > A tentative patch is attached, just for cpg for now. Freeing in finalize > is a bit tricky, as dispatch might still be running. > > But this seems to be too much trouble when there is a decent workaround > (i.e., changing JVM's stack size with a simple command line parameter) > for my initial problem. So, I'm not pushing very hard for this > to get accepted. > Thanks a lot for your patch. We are currently analyzing two different solutions: - libqb dispatch functions returns raw pointer to buffer (instead of copy). This would mean no stack needed (and maybe also little speedup improvement) - Disallow reentrant call of *_dispatch functions (so again no stack needed) I will definitively let you know results. Thanks again for patch. Regards, Honza _______________________________________________ discuss mailing list discuss@xxxxxxxxxxxx http://lists.corosync.org/mailman/listinfo/discuss
