Subject: Re: [linux-audio-dev] Reverbs... and the usual whinings
From: Benno Senoner (sbenno_AT_gardena.net)
Date: Thu Sep 28 2000 - 03:39:14 EEST
On Wed, 27 Sep 2000, Kai Vehmanen wrote:
> Answering to an older msg...
>
> On Sat, 23 Sep 2000, Benno Senoner wrote:
>
> > no the low-latency patch alone does not ensure a drop-out free operation.
> [...]
> > So the key is to provide enough buffering to overcome the 1-2sec stall
> > of the disk thread.
>
> Hmm, is the disk i/o latency (over 1sec stalls possible) something of an
> accepted compromise (from disk-i/o design perspective), or is it something
> you should expect to vary (possibly a lot) from impl. to impl.?
do you mean varies in function of application implementation or kernel
implementation ?
These stalls are mainly when large buffers are synced to disk during writes.
> How do
> 2.2.x and 2.4.x kernels compare here?
A few months ago I as in touch with Jens Axboe, and
he had some experimental patches which showed up considerably better
values when running hdrbench.
I guess I should simplify the usage of hdrbench (quite messy right now), so that
linux kernel hackers working on the disk subsystem can use it to lower
these stalls.
I think max stalls of 0.5 sec is quite realistic and would mean halving
(or reducing even more) the disk buffers.
we will see ....
PS: pleasant EVO news coming tomorrow .... :-)
Benno.
This archive was generated by hypermail 2b28 : Thu Sep 28 2000 - 02:11:15 EEST