Date: Mon, 2 Feb 2004 17:42:24 -0500 (EST) From: Jeff Roberson <jroberson@chesapeake.net> To: Don Lewis <truckman@freebsd.org> Cc: current@freebsd.org Subject: Re: nice not nice enough with ULE Message-ID: <20040202174022.S30170-100000@mail.chesapeake.net> In-Reply-To: <200402020201.i1221h7E098722@gw.catspoiler.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On Sun, 1 Feb 2004, Don Lewis wrote: > I'm seeing the following during the java library building phase of the > lang/gcc33 port: > > > last pid: 41215; load averages: 3.82, 3.60, 3.53 up 0+05:39:11 17:53:43 > 67 processes: 4 running, 63 sleeping > CPU states: 1.5% user, 78.2% nice, 20.3% system, 0.0% interrupt, 0.0% idle > Mem: 72M Active, 624M Inact, 125M Wired, 37M Cache, 111M Buf, 139M Free > Swap: 2055M Total, 2055M Free > > PID USERNAME PRI NICE SIZE RES STATE TIME WCPU CPU COMMAND > 531 setiathome 139 15 16692K 15888K RUN 195:45 75.00% 75.00% setiathome > 18973 dl 76 0 2200K 1328K RUN 0:41 0.00% 0.00% top > 508 uucp 5 0 1264K 924K ttyin 0:35 0.00% 0.00% newapc > 582 dl 76 0 6004K 2296K select 0:10 0.00% 0.00% sshd > 92710 root 8 0 11616K 10596K wait 0:08 0.00% 0.00% gmake > 19272 root 8 0 12456K 11980K wait 0:07 0.00% 0.00% ruby > 98135 root 8 0 9676K 8568K wait 0:06 0.00% 0.00% gmake > 19274 root -8 0 1124K 520K piperd 0:05 0.00% 0.00% tee > 510 uucp 76 0 1240K 888K select 0:04 0.00% 0.00% upsd > 18955 dl 76 0 6004K 2320K RUN 0:03 0.00% 0.00% sshd > 514 uucp 8 0 1244K 920K nanslp 0:02 0.00% 0.00% upsmon > 7129 root 76 0 1204K 784K select 0:02 0.00% 0.00% script > 430 root 76 0 1516K 1020K select 0:02 0.00% 0.00% ntpd > 16600 dl 4 0 1152K 664K kqread 0:01 0.00% 0.00% tail > 461 root 76 0 3252K 2056K select 0:01 0.00% 0.00% sendmail > 35166 dl 8 0 1188K 868K nanslp 0:01 0.00% 0.00% vmstat > > > The port build is running at nice 0, but setiathome at nice 15 is > getting 75% of the CPU. Something unusual about this phase of the port > build is that it is using libtool, which must spawn a lot of processes, > because "last pid" in increasing by about 30 to 40 per second. The port > build should be pretty close to CPU bound because I see very litle disk > I/O. ULE Does some things on fork to try to prevent rapidly forking "interactive" threads from overwhelming the system with new non-interactive tasks. This is exactly the case with make or sh. It may be a little too aggressive. I will reproduce this scenario this weekend. Thanks for the effort that you've put into this. Cheers, Jeff > > When larger files are compiled so that individual processes hang around > for a longer period of time, I see setiathome drop down to about 11% of > the CPU. > > The machine in question is has a single Athlon XP processor and is > running rev 1.98 of sched_ule.c. > > > _______________________________________________ > freebsd-current@freebsd.org mailing list > http://lists.freebsd.org/mailman/listinfo/freebsd-current > To unsubscribe, send any mail to "freebsd-current-unsubscribe@freebsd.org" >
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20040202174022.S30170-100000>