On Mon, 19 Mar 2012 01:07:03 -0400 Matthew Mondor <mm_li...@pulsar-zone.net> wrote:
> On Mon, 19 Mar 2012 01:03:19 -0400 > Matthew Mondor <mm_li...@pulsar-zone.net> wrote: > > > On Mon, 19 Mar 2012 00:51:13 -0400 > > Matthew Mondor <mm_li...@pulsar-zone.net> wrote: > > > > > I suspect that boehm-gc could be at fault. I only did the tests on > > > NetBSD so far, but also intend to try on Linux. These tests were also > > > using the new NetBSD-6 TLS support, I'll also try with > > > --with-__threads=no. > > > > Well, the system seems quite faster using --with-__threads=no so far. > > I'll investigate about why the TLS path seems so problematic. I also > > don't experience the instable speed issue. > > Well I guess I spoke too fast though, as the same problem with the FDs > magically becomming invalid happened. But yet again that could be a > boehm-gc related issue. With the stable release I never got that exact > error, but a buggy GC can be a source of many random problems... An interesting thing is that despite it not being stable when I stress-test it using ab(8), the new locks seem more stable over a number of days with very low traffic than the old pthread-based implementation. I didn't yet investigate why performance and stability decreases with --with--__threads=yes yet. -- Matt ------------------------------------------------------------------------------ This SF email is sponsosred by: Try Windows Azure free for 90 days Click Here http://p.sf.net/sfu/sfd2d-msazure _______________________________________________ Ecls-list mailing list Ecls-list@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/ecls-list
