Re: [Q] Lowest scheduling priority

From: Sci-col (grobyn_at_telus.net)
Date: 11/29/03


Date: Sat, 29 Nov 2003 07:16:13 GMT

Fredderic wrote:
> On Thu, 13 Nov 2003 13:16:33 -0500, Steve Ackman wrote:
>
>
>>>Is there a way to set the priority of a process so low that it only uses
>>>processor cycles that would have gone idle anyway?
>>
>> I don't think your problem is niceness, but RAM
>>shortage. I used to run seti at nice 19, and
>>experienced delays in higher priority processes. It
>>was due to memory management. Once I doubled my RAM,
>>there was enough memory for all the processes to have
>>their own, so there was no need for them to share it.
>>Everything sped up considerably.
>
>
> On my machine, I have more than enough RAM (until I started leaving
> Mozilla running for a week on end, I hadn't realized I'd stuffed up my
> fstab and it wasn't even mounting a swap).
>
> It seems as far as I can tell, that a n-19 task will still steal some
> processor time from other apps. One of my favorite little arcade
> style games consumes every ounce of available CPU power, keeping the frame
> rate as high as possible. With not much else in the way of active
> processes, it seems to run very close to 100%. With seti running, it gets
> a significantly lower amount of CPU time, with seti getting the rest.
>
> This is consistent with the scheduler trying to avoid allowing processes
> to become starved for CPU time (by a method known as aging). Since seti
> is a predominantly maths application, it gets handed control every so
> often by the scheduler (just in case), and runs for its full allowed
> time slice (whereas many nice'd processes will frequently end up doing
> disk I/O or similar that relinquishes the CPU back to the higher priority
> task almost immediately).
>
> What I (and the OP, as best I can tell) want, is a way to stop the
> scheduler from giving seti ANY time slices unless nothing else at all
> wants them. I have considered writing a small daemon to stop seti when
> the processor utilization goes above 30%, and restart it again once it
> drops below 10%, but the current situation isn't really that big of an
> issue anyhow.
>
>
> Fredderic

hint: During the scheduling process (best candidate to be scheduled) is
set to the idle task of this cpu. However, the goodness of this
candidate is set to a very low value (-1000), in hope that there is
someone better than that.



Relevant Pages

  • Re: [Q] Lowest scheduling priority
    ... Once I doubled my RAM, ... With seti running, it gets ... a significantly lower amount of CPU time, ... This is consistent with the scheduler trying to avoid allowing processes ...
    (comp.os.linux)
  • SETI hardware suggestions?
    ... a SETI processor. ... The SETI sites ... I'm also interested in error-correcting ram ... As I look at my Micro Center flyer, ...
    (rec.arts.sf.fandom)
  • TaskManager = Fridge light
    ... What with things slowing down a bit I've set the task manager to default appear ... I have CPU 1 set, or rather I have Seti set to use ONLY CPU 1. ... The page file usage shows no change related to the usage graph. ...
    (microsoft.public.windowsxp.newusers)
  • Re: FX-55 vs. X2 4800+
    ... >I am running with an FX-55 CPU on the ASUS A8n-SLI Deluxe board. ... Presumably the SETI ... the command rate is determined by the motherboard, ...
    (alt.comp.periphs.mainboard.asus)
  • Re: [SLE] For or against ..Hyperthreading.
    ... >> I'll take a single cpu any day over multiple cpu's that DON'T ADD UP ... So, for example, I'll get more SETI ... except on the most compute intensive task. ... > updatedb does nothing else except the updatedb. ...
    (SuSE)