Re: [linux-audio-dev] questions to be resolved

New Message Reply About this list Date view Thread view Subject view Author view Other groups

Subject: Re: [linux-audio-dev] questions to be resolved
From: Richard C. Burnett (burnett_AT_tality.com)
Date: Tue May 22 2001 - 18:36:44 EEST


Thinking about this for a while, let me present a situation where I think
there are two opposing situtations that limit the 'parameter changing'
idea.

First, I have two different effects units, one that is routable and fully
configurable in any order, combination, that is my 2101 DSP From
Digitech. The other, which is not, is the RP7 also from Digitech.

When I switch 'patches' which is really a large configuration of different
effects, there is a short delay on my RP7 followed by an abrupt change to
the new settings.

On my 2101, there are two processors, so when I switch, it goes from one
to the other and they are mixed together until the first one fades. This
of course only works for those patches that require only one of the SDISC
processors.

I know there has been discussion about reorganization of connections for
optimized performance. What I do not see is how the reorganization could
be so complex as to require that much time. If it were that complex, then
I would think that it wouldn't be able to be real time also.

So I see two choices:

1. you go quite until the reorganization occurs
2. you keep 2 copies and keep processing the old one until the new one is
   ready, then switch.

Think of it as a configuration dual buffer, you start with two copies and
then a change is propagated through one, you switch to that one and then
bring the other one up to the same configuration. I think this is a great
approach, if:

1. The overhead of having two is small
2. The design of the connectivity allows it

For instance, if you have buffer reuse, and if I understand it correctly,
both copies could use the same buffers, since the changing copy is only
reorganizing itself or adding, etc, etc, then some sort of smart buffer
management could be used to keep track of that portion of it.

Well, it sounds good in my head :) I am sure that you can point out all
the flaws of it, and I hope you do :)

Please note that you start with two copies, not create a copy right before
a change, if that makes sense.

Rick

On Tue, 22 May 2001, Paul Davis wrote:

> >I've two copy of a topology area every time user change it, this makes
> >you worry?
>
> Yes. That means that be safe, you need to copy every Port, and every
> Port-owning object. Moreover, in the scheme you propose, its the
> reader (the audio thread) that makes the copy. And since you still can
> cause blocking behaviour in the audio thread, i don't see what its
> solving.
>
> --p
>

+------------------------+-----------------------+
| T a l i t y | +------+ |
+------------------------+ +----+-+ | |
| Richard Burnett | +-+ | |
| Senior Design Engineer +---+ +----+ |
| burnett_AT_tality.com | | |
| | | |
| Phone: 919.380.3014 | |
| Fax: 919.380.3903 | | |
+------------------------------------------------+


New Message Reply About this list Date view Thread view Subject view Author view Other groups

This archive was generated by hypermail 2b28 : Thu May 24 2001 - 06:06:25 EEST