Re: -mm seems significanty slower than mainline on kernbench

Con Kolivas wrote: 
On Wed, 11 Jan 2006 01:38 pm, Peter Williams wrote:

Con Kolivas wrote:
> I guess we need to check whether reversing this patch helps.

It would be interesting to see if it does.

If it does we probably have to wear the cost (and try to reduce it) as
without this change smp nice support is fairly ineffective due to the
fact that it moves exactly the same tasks as would be moved without it.
 At the most it changes the frequency at which load balancing occurs.


I disagree. I think the current implementation changes the balancing according to nice much more effectively than previously where by their very nature, low priority tasks were balanced more frequently and ended up getting their own cpu.

I can't follow the logic here and I certainly don't see much difference in practice.

No it does not provide firm 'nice' handling that we can achieve on UP configurations but it is also free in throughput terms and miles better than without it. I would like to see your more robust (and nicer code) solution incorporated but I also want to see it cost us as little as possible. We haven't confirmed anything just yet...

Yes, that's true. I must admit that I wouldn't have expected the increased overhead to be very big. In general, the "system" CPU time in a kernbench is only about 1% of the total CPU time and the extra overhead should be just a fraction of that.

It's possible that better distribution of niceness across CPU leads to more preemption within a run queue (i.e. if there all the same priority they won't preempt each other much) leading to more context switches. But you wouldn't expect that to show up in kernbench as the tasks are all the same niceness and usually end up with the same dynamic priority.

Peter
--
Peter Williams                                   pwil3058@xxxxxxxxxxxxxx

"Learning, n. The kind of ignorance distinguishing the studious."
 -- Ambrose Bierce
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Relevant Pages

  • Changing Program Process Priority Defaults
    ... changing the program priority. ... It was distributed to limit the CPU time that the SETI ... >My kids love playing a Disney Lama Game. ... >automatically force the default to normal priority? ...
    (microsoft.public.windowsxp.general)
  • Re: Price of CPU seconds
    ... understand why the 0.5% CPU time is a matter: ... If it is the first then the cost is substantially higher than it appears when just ... looking at raw numbers - even to their mindset. ... For IBM-MAIN subscribe / signoff / archive access instructions, ...
    (bit.listserv.ibm-main)
  • Re: Windows service Vs Console app performance
    ... but priority is a funny thing. ... other open windows, task bar applets, mouse, etc. ... > can have as much CPU time as it wants, bar what is needed for the OS, so ... > Maybe he has another processor intensive task running at the same time? ...
    (microsoft.public.dotnet.framework)
  • Re: DirectShow Thread Priority
    ... > I'm not persuaded that it does run at a higher priority. ... other CPU-intensive tasks going on during video rendering, ... will take most of the CPU time anyway. ... difference in performance by the encoder. ...
    (microsoft.public.win32.programmer.directx.video)
  • Re: Possible fix for Windows Update systems hangs
    ... This posting is provided "AS IS" with no warranties, ... > was using 99% of the CPU time. ... > From TaskManager set the Priority of the EXPLORER ... > Run ADD/REMOVE PROGRAM from control panel. ...
    (microsoft.public.windowsupdate)