From alsa-devel-owner@alsa.jcu.cz  Fri Jan 15 17:29:29 1999
Received: from renoir.op.net (root@renoir.op.net [209.152.193.4])
	by marvin.jcu.cz (8.9.1a/8.9.1) with ESMTP id RAA26114
	for <alsa-devel@alsa.jcu.cz>; Fri, 15 Jan 1999 17:27:31 +0100
Received: from someip.ppp.op.net (d-bm2-17.ppp.op.net [209.152.194.55]) by renoir.op.net (o1/$Revision: 1.18 $) with ESMTP id LAA01083 for <alsa-devel@alsa.jcu.cz>; Fri, 15 Jan 1999 11:26:42 -0500 (EST)
Message-Id: <199901151626.LAA01083@renoir.op.net>
To: alsa-devel@alsa.jcu.cz
Subject: Re: Linux scheduler issues. 
In-reply-to: Your message of "Fri, 15 Jan 1999 12:35:26 +0100."
             <9901151135.AA05985@tele2> 
Date: Fri, 15 Jan 1999 11:28:04 -0500
From: Paul Barton-Davis <pbd@op.net>
Reply-To: alsa-devel@alsa.jcu.cz
Sender: alsa-devel-owner@alsa.jcu.cz
Precedence: list

In message <9901151135.AA05985@tele2>you write:
>Benjamin GOLINVAUX writes:
> > However, there is a point I don't clearly understand about Steve's claims
> > about RT language : don't you think the heaviest workload to have real-time
> > performance is in kernel-mode code ?
> >
>
>Of course the main work is in the kernel. As far as a language goes, I
>already have the language. Everything I have mentioned so far is
>already implemented in Erlang. What I/we need is a real time MUSIC
>kernel to run on.

First of all, I don't agree at all that the heaviest workload is in
the kernel. If you run a software synthesizer or sophisticated DSP
emulation, the vast majority of cpu cycles will be spent doing
floating point math. The time you spend in the kernel is limited, in
terms of this particular application, to copying data to and from the
audio output device, and this will take vastly less time than the
sound generation. If this wasn't true, then my old 486 would have been
able to run Csound in real time but not, for example, play WAV files. 

So, although we certainly want a speedy, low-latency kernel and
driver, I think its more or equally critical that we have even more
efficient applications for this kind of thing.

Erlang has its merits, but for me falls into a long line of attempts
to solve with a language problems that exist at the application and/or
OS level. I can see no justification for Erlang's subsumption of OS
duties, other than Ericsson presumably wanting a platform independent
environment for telephony and related software development.

Yes, C and C++ are clunky, low on facilities and confusing to many
people. But when I want my DSP simulator running at warp speed, I know
how I'm going to get there.

I think, Steve, that if you really want a real-time music kernel, you
should not be looking at regular Linux. 

--p

