On Wed, Apr 15, 2015 at 03:51:59AM -0600, Anthony J. Bentley wrote: > David Coppa writes: > > On Fri, Apr 3, 2015 at 9:47 PM, Anthony J. Bentley <anth...@cathet.us> > > wrote: > > > David Coppa writes: > > >> On Mon, Mar 23, 2015 at 3:14 AM, Michael <lesniewskis...@gmail.com> > > >> wrote: > > >> > So some have mentioned that epoch shouldn't be removed. Apologies as > > >> > following [0] it only mentions epoch once and doesn't note that it has > > >> > to stay on updates. > > >> > [0]: http://www.openbsd.org/faq/ports/guide.html > > >> > I also bumped the libary version as there were a lot of changes in the > > >> > new libtorrent version that mean it can't be used with the older > > >> > rtorrent application. > > >> > Attached are the updated patches, further comments welcome! > > >> > > >> I'm testing it right now... > > >> So it seems they finally solved the infamous problem with IPIs > > >> (Inter-processor interrupts) going sky-high (in the order of ~900000 > > >> ipi) and thus bringing the system down to its knees. > > > > > > This has not been fixed. Today I opened both rtorrent and firefox and my > > > system locked up within 10 minutes. > > > > Please, can you try again while looking at ipi with 'systat 1' > > Maybe I've been experiencing a different issue than you have. During a > bad slowdown that produced visible jitter and killed all my IRC > connections (nothing besides rtorrent, net/ii, and irssi was running), > IPIs were under 50 but re0 interrupts spiked to over 500000. Is that > something that should happen during even heavy rtorrent usage (80+ > torrents)? Currently it's holding at about 5600 re0 interrupts.
Fwiw, this rtorrent 0.9.4 update is now broken on powerpc... with the usual atomic stuff. .libs/libtorrent.so.21.0: undefined reference to `__sync_fetch_and_and_8' .libs/libtorrent.so.21.0: undefined reference to `__sync_add_and_fetch_8' Landry
