Boinc linux version 5.8.16 wrong cpu effinity einstien
16 Sep 2007 18:13:07 UTC
Topic 193146
(moderation:
when running either run_manager or run_client with einstien the cpu effinity is set to "nice 19" which results in 60% loading on the cpu. seti app runs at "nice -10". is there some .xml config file i can fix this in???
Boinc linux version 5.8.16 wrong cpu effinity einstien
)
Nice 19 is the right setting for a background application.
Definitely! It's the very
)
Definitely!
It's the very idea of BOINC to use only idle cycles, and this means ==> max nice value (19) for all science apps! If there are other tasks consuming CPU cycles, the einstein and seti apps must give way.
It would be interesting to see a screenshot of "top" which shows the different nice value for the SETI app. Is it an official app or an "optimized" app?
CU
H-BE
RE: Definitely! It's the
)
My SETI@home optimized for Linux on a PII Deschutes with MMX runs at nice 19.
Tullio
[pre]top - 23:30:53 up 3
)
[pre]top - 23:30:53 up 3 days, 11:56, 2 users, load average: 4.28, 4.18, 4.06
Tasks: 164 total, 4 running, 160 sleeping, 0 stopped, 0 zombie
Cpu(s): 51.2%us, 0.6%sy, 47.3%ni, 0.0%id, 0.0%wa, 0.3%hi, 0.5%si, 0.0%st
Mem: 1033792k total, 945560k used, 88232k free, 171256k buffers
Swap: 2031608k total, 72132k used, 1959476k free, 177200k cached
PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
14518 root 10 -10 53916 33m 1560 R 100 3.3 138:10.98 setiathome-5.15
14637 root 10 -10 52976 32m 1560 R 100 3.2 72:08.62 setiathome-5.15
4061 root 39 19 78252 41m 2572 S 100 4.1 2056:18 einstein_S5R2_4
14519 root 39 19 174m 89m 28 R 92 8.8 92:51.41 rosetta_beta_5.
3686 root 20 0 43756 21m 8336 S 3 2.1 135:51.08 boincmgr
3119 root 20 0 73584 28m 6732 S 3 2.8 75:53.14 Xorg
14822 root 20 0 2240 1064 804 R 1 0.1 0:00.14 top
3404 root 20 0 26896 8540 3592 S 0 0.8 2:38.94 compiz
3438 root 20 0 18428 1184 1140 S 0 0.1 1:29.37 escd
1 root 20 0 2140 600 560 S 0 0.1 0:05.01 init
2 root 15 -5 0 0 0 S 0 0.0 0:00.00 kthreadd
3 root RT -5 0 0 0 S 0 0.0 0:01.63 migration/0
4 root 15 -5 0 0 0 S 0 0.0 0:00.29 ksoftirqd/0
5 root RT -5 0 0 0 S 0 0.0 0:00.00 watchdog/0
6 root RT -5 0 0 0 S 0 0.0 0:00.00 migration/1
7 root 15 -5 0 0 0 S 0 0.0 0:00.12 ksoftirqd/1
8 root RT -5 0 0 0 S 0 0.0 0:00.00 watchdog/1 [/pre]
my guess is when three nice -10's are pared up with one nice 19, the nice 19 doesn't run much.
seti app is chicken 2.4
RE: [pre]top - 23:30:53 up
)
Thanks very much, very interesting indeed. Well, either this is a strange new feature (or bug) of BOINC I don't know about, or those optimized SETI@Home apps are kind of not following standard BOINC patterns (note I did not say "cheating").
BOINC should never start more science apps at the same time than there are physical CPUs (or more precisiely, CPU cores) present in the system. So whether this makes a difference or not depends on what other processes are running on your PC. From the TOP output above I get the impression that, roughly speaking, the two SETI-apps will effectively get each one core almost exclusively for their own while Rosetta and Einstein are playing by the rules and are each using the idle cycles only of one core each.
CU
H-BE
RE: [pre] ... PID USER
)
well, einstein and rosetta ARE using CPU cycles (see above). So I guess it is a 4 core system... seti attached to 2 cores, einstein and rosetta (and the rest of the system) running on the other 2 cores.
Udo
Udo
RE: well, einstein and
)
Yup, which is exactly what I wrote, or wanted to write, at least. Still I see no reason why the seti stuff is behaving as if it was "more equal" than other BOINC apps :-).
CU
BRM
RE: RE: well, einstein
)
OK, this seems to be a "bug" in the chicken apps that was corrected in the 2.4L versions:
See this messge in the SETI forums.
Don't let root run user
)
Don't let root run user applications.