On Tue, Dec 3, 2013 at 2:18 PM, Pavel Levshin <[email protected]> wrote:
> 03.12.2013 14:52, Rainer Gerhards: > > > It's just a configure option. If --enable-jemalloc is given, we link to >> its libs. That's it. >> >>> Any new measurements in terms of performance compared to using normal >>> malloc? >>> >>> none done myself, but Pavel posted some impressive numbers in the past >> week >> or two (that actually made me switch). So it's probably best to search the >> ML archive. >> > > Those numbers depend heavily on configuration and environment, so take > them with a grain of salt. Maybe some platform is using more suitable > malloc by default; mine is not. > > Yeah, after you made me aware of it, I read up on jemalloc. The perception was overall very well. Especially for highly threaded apps as rsyslog a big win is expected, and seen by other apps as well. Of course, it all heavily depends, but I'd say it's a plus in any case (except the BSDs which, as far as I understand, use jemalloc in any case). > I did not recompile rsyslog to use jemalloc when was doing the test, I > used LD_PRELOAD instead. > > > yup, but the net effect should be the same. If --enable-jemalloc is NOT specified (default), rsyslog just builds again the standard malloc handler. But, as I wrote, we have enabled jemalloc builds in our packages. Just to clarify. Rainer _______________________________________________ rsyslog mailing list http://lists.adiscon.net/mailman/listinfo/rsyslog http://www.rsyslog.com/professional-services/ What's up with rsyslog? Follow https://twitter.com/rgerhards NOTE WELL: This is a PUBLIC mailing list, posts are ARCHIVED by a myriad of sites beyond our control. PLEASE UNSUBSCRIBE and DO NOT POST if you DON'T LIKE THAT.
