Posts by Samuel Blomqvist

1) Message boards : Number crunching : One WU not respecting nice... (Message 309170)
Posted 18 May 2006 by Profile Samuel Blomqvist
Post:
Hi Martin,

the strange thing is he's seeing 25%us (User) load, and only 75%ni (Nice),
while all boinc tasks are at nearly 100% cpu.

There really should be just about 100%ni in the second line.


Regards Hans




Exactly. I know all the other stuff and BOINC is using almost 100% x4 CPUtime. And that is fine with me. However they should all be running with nice 19 and looking at top they do but still in gkrellm some WUs show up as not being nice. And when you look at top you see that there is only 75% or less nice when there should really be almost 100% nice... since nothing much else is running. When running with nice 19 BOINC shouldn't generate "user load" should it?

I just got another one it is the WU 22fe99aa.2560.546.959636.3.85

/LinuxSam
2) Message boards : Number crunching : One WU not respecting nice... (Message 308475)
Posted 17 May 2006 by Profile Samuel Blomqvist
Post:




A nice level of 19 means "idle priority". There's something weird going on.

Try increasing top's update rate with the "s" command, maybe you can spot something .

If you login as root, you can specify sub-second update rates like "0.1" or even "0.02"

Use the "i" command to only display active tasks.

Regards Hans


I don't see anything that is strange... (tried increasing the updaterate but didn't see anything then either) the only process that shows up as running is a 4.02 setiversion... Seti_enhanced allways shows up as sleeping for some reason.

/Samuel
3) Message boards : Number crunching : One WU not respecting nice... (Message 308443)
Posted 17 May 2006 by Profile Samuel Blomqvist
Post:
I now have another WU missbehaving...

It's the WU 07ja99aa.20918.7201.686074.3.57

Shows up on the CPU-load meters as not being "nice" but looking in top it shows with a nice 19.

/Samuel

4) Message boards : Number crunching : Cpu usage (Message 308024)
Posted 16 May 2006 by Profile Samuel Blomqvist
Post:
Seems I was wrong about not being able to limit the CPU-usage...

look here: http://setiathome.berkeley.edu/download_network.php

And check out Threadmaster could do what you are asking for.

The other facts about BOINC using spare CPU-cycles are correct though...

/LinuxSam
5) Message boards : Number crunching : Cpu usage (Message 306749)
Posted 15 May 2006 by Profile Samuel Blomqvist
Post:
i have used a program that tells me that BOINC for seti is using most of my cpu power. i just recently upgraded to a hyperthreading cpu, how do i get this to use less of the cpu?


BOINC is using all left over CPU-cycles... BOINC should be running at lowest priority... As soon as another program needs the CPU BOINC should let go of the resources... I don't know of a way to let it only use this or that much CPU...

With a hyperthreading CPU you could ofcourse tell it only to use one CPU since it sees hyperthreading as two cpus... I think there is a setting somewhere "On multi CPU systems use this many CPUs" Check the settings. Setting that to 1 might make it use less CPU-time on your computer or you could simply manually turn it off if you feel it is degrading performance when you do something important.

/LinuxSam
6) Message boards : Number crunching : CPU Running Hot (Message 306729)
Posted 15 May 2006 by Profile Samuel Blomqvist
Post:
I'm running a system of Dual Opteron 265 (dualcore). And my CPU-tempratures at full load are between 38C and 43C depending on room temprature...

My CPU-coolers are: Scythe Ninja Plus Heatpipe and I have used Arctic Silver 5 They each have a 120mm fan running about 1300rpm and I guess I could turn them down a bit if I wanted to.

My case is a Lian-Li PC-V2100 Plus.




/LinuxSam
7) Message boards : Number crunching : One WU not respecting nice... (Message 306615)
Posted 15 May 2006 by Profile Samuel Blomqvist
Post:
Can you check what the process priorities actually are for that process and any associated processes? (Use "top" in a terminal window?)

Can you change that process priority down to "nice 19"? (Use "renice"?)

Or have you got strange settings for that one CPU meter? (Double check its options/properties?)

Let us know what you find,

Good luck,
Martin



Here is part of what top shows:
Tasks: 148 total, 2 running, 146 sleeping, 0 stopped, 0 zombie
Cpu(s): 26.9% us, 0.6% sy, 72.3% ni, 0.0% id, 0.0% wa, 0.1% hi, 0.2% si
Mem: 2075172k total, 2006832k used, 68340k free, 37300k buffers
Swap: 2634620k total, 160k used, 2634460k free, 1237836k cached

PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
11380 boinc 34 19 59240 37m 3420 S 100 1.9 167:47.01 setiathome-5.12
955 boinc 34 19 60240 38m 3420 S 99 1.9 414:59.22 setiathome-5.12
11808 boinc 39 19 18968 15m 1252 R 97 0.8 129:52.83 setiathome_4.02
956 boinc 34 19 59144 37m 3420 S 94 1.9 415:46.37 setiathome-5.12
14347 samuel 16 0 28556 14m 11m S 5 0.7 0:03.49 boincmgr


taking into account that that process has about 2h 50min of CPU-time I'd say it's the topmost one...

the 100% CPU-load moves between the cores so I don't think it has anything to do with settings in the CPU-meters (I'm using gkrellm)

shut down the processes one by one and when I shut down that one the 100% load disapeared.


/LinuxSam
8) Message boards : Number crunching : One WU not respecting nice... (Message 306479)
Posted 15 May 2006 by Profile Samuel Blomqvist
Post:
This is strange... I have dual system with a total of four cores and suddenly saw one of the CPU-meters indicating 100% load (It shouldn't show "nice") and it turns out it is a special WU (30mr99aa.29392.16146.492302.3.196_3) that is causing this... I have no idea if this has any real effect on my computer or not but all the other WUs are running fine with nice 19 and not showing up on the CPU-load meter.

Anybody's got an idea?

BTW running BOINC under Debian GNU/Linux with a 2.6.16-kernel.

/Samuel





 
©2025 University of California
 
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.