Subject: Re: [Alsa-devel] Re: [linux-audio-dev] Re: more bad low latency results
From: Takashi Iwai (tiwai_AT_suse.de)
Date: Thu Nov 15 2001 - 18:48:03 EET
At Thu, 15 Nov 2001 17:28:19 +0100,
Maarten de Boer wrote:
>
> > There are VM tuning parameters in /proc/sys/vm/vm_*_ratio files.
> > Lowering these values tend to bring better latency results.
>
> Any recommended reading about this?
I've used this setting:
echo 6 > /proc/sys/vm/vm_scan_ratio
echo 2 > /proc/sys/vm/vm_mapped_ratio
echo 4 > /proc/sys/vm/vm_balance_ratio
> I reinstalled my system (needed to do that for a while now anyway)
> to be crispy clean. Look at the latest results.
>
> http://www.iua.upf.es/~mdeboer/latency)
>
> (I will put new tests here instead of on my testing machine)
>
> It starts to look better. However, as you can see there are some ugly
> peaks remaining during disk write test. What would be the way to trace
> the cause of these peaks?
I experienced this, too. Perhaps the tuning above helps for such VM
throttle cases.
> Meanwhile, the alsa-lib/test/latency test still sounds like crap, and any
> sample rate, and in both modes. I really believe this only happens when
> I compile my kernel for 386/586 instead of AMD K7.
That's really weird.. How about to measure the time of read/write
responce in latency.c so that we can know at least whether it's
related with the kernel latency.
> It would be really nice if somebody could repeat my tests on identical or
> similar hardware (AMD Athlon, Trident 4DWave NX), with the same versions
> of kernel, patch, and alsa (latest cvs that is). I will try with a es1371
> to see what happens with the alsa I/O...
I'll try SB live with Athlon later..
ciao,
Takashi
This archive was generated by hypermail 2b28 : Thu Nov 15 2001 - 18:46:05 EET