> >>My tests shows that increasing the scheduling frequency to HZ=1000 ....
> >
> >I feel wary of increasing HZ, but maybe I'll try this out and see how
> >it works for me.
>
> First clarification: increasing HZ has big payoffs for a program like
> your benchmark, which *isn't* burning CPU cycles generating audio in
> real time. But it can be problematic for an application that can use
> every available cycle for its own work, because the ten-fold increase
> in timer interrupt frequency is going to steal cycles. Does anyone
> have any idea how many (x86) cycles it takes to process a timer
> interrupt that doesn't result in any change to the tasks running on
> each CPU ?
quite a while ago I've raised HZ fron 100 to 1024 on an old DX2/66.
comparing kernel compile time (obviously CPU bound with 20MB ram
and ~95% CPU utilisation) there was a slowdown of ~1 min for
a total compile time of ~55 min. so you can see the effect,
but even for a DX2/66 it wasn't bad at all (and it did solve the
problems where I needed HZ >> 100).
Harald
--
All SCSI disks will from now on ___ _____
be required to send an email notice 0--,| /OOOOOOO\
24 hours prior to complete hardware failure! <_/ / /OOOOOOOOOOO\
\ \/OOOOOOOOOOOOOOO\
\ OOOOOOOOOOOOOOOOO|//
Harald Koenig, \/\/\/\/\/\/\/\/\/
Inst.f.Theoret.Astrophysik // / \\ \
koenig@tat.physik.uni-tuebingen.de ^^^^^ ^^^^^
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/