From alsa-devel-owner@alsa.jcu.cz  Mon Jan 11 18:38:57 1999
Received: from nic.funet.fi (nic.funet.fi [128.214.248.6])
	by marvin.jcu.cz (8.9.1a/8.9.1) with ESMTP id SAA04117
	for <alsa-devel@alsa.jcu.cz>; Mon, 11 Jan 1999 18:36:29 +0100
Received: from localhost (user: 'kouhia', uid#241) by nic.funet.fi id <2470-2277>; Mon, 11 Jan 1999 19:36:08 +0200
From: Juhana Sadeharju <kouhia@nic.funet.fi>
To: alsa-devel@alsa.jcu.cz
In-reply-to: <006d01be3d6c$87bb1660$47010180@pcbg> (golinvaux@benjamin.net)
Subject: Re: Scheduler issues
Message-Id: <19990111173610Z2470-2277+2066@nic.funet.fi>
Date: 	Mon, 11 Jan 1999 19:36:08 +0200
Reply-To: alsa-devel@alsa.jcu.cz
Sender: alsa-devel-owner@alsa.jcu.cz
Precedence: list

>From:	"Benjamin GOLINVAUX" <golinvaux@benjamin.net>
>
>2) isn't it just what KURT (or RTLinux ?) promise ? From KURT manual :
>"when in real-time mode, [the kernel] executes only those processes that are
>marked as real-time)... Looks promising, isn't it ?

I think the above (or running only one process) is not a good way
because the programming becomes a difficult task. How to make a GUI
to work at the same time?

We just should have a mechanism that makes the wanted processes to
schedule N times per second.

Should it work as "run these processes at each k*1/N (k = 0,1,...N-1) tick"?
Perhaps it would practically done such that rtfork() would make the above
kind of rt-processes and those processes would be executed in the creation
order. Or there could be some reordering calls to alter the execution
order of rt-processes. So, first rt-process executes at 0, 1/N, etc.
and second process at 0+epsilon, 1/N+epsilon, etc. where epsilon is the
run time of the first process.

The above might mean that all rt-processes should be of the form which does
a tiny thing and then stops. For example, process next M samples and wait
for the next turn.

All other processes are executed if there is a free time after rt-processes
and before the next tick.

Thank you Steve for the explanation. But are driver codes scheduled? For
example, how disk or A/D reading happens?


>BTW2, do you have a clue of which OS the last Fairlight series III machines
>(from 1990) ran ? (might be what we are looking for)

I'm interested in to any info about their OS & software...

Yours,

Juhana

