Subject: Re: [linux-audio-dev] stereo / multichannel plugins in LADSPA ?
From: David Olofson (david_AT_gardena.net)
Date: Sun Mar 26 2000 - 00:12:02 EET
On Thu, 23 Mar 2000, Paul Barton-Davis wrote:
> i can't think of any devices that accept 24 channels of data, process
> them all in the same way, and spit out 24 channels. so first of all,
> this isn't a very likely model. but even if it was - OK, just use 24
> ports, and let them all be connected independently. yes, its a
> performance hit, but thats OK, its a rather strange thing to be
> doing.
I have to agree here... As for the performance hit on a single
plugin-plugin connection, it's minimal for any normal number of
channels, and from the full net POV, unavoidable with any plugin
system. All you can do (and that means a whole lot!) is to have the
host reuse buffers in a smart way, to use hot buffers as much as
possible. And that actually becomes easier if most buffers are of the
same size.
> the performance hit from cache effects and/or loop iterations *HAS* to
> happen either in the host (see my next paragraph) or in the plugin, so
> its a wash eiher way.
...and I think the host has a greater chance of
>
> >> again, we need you thoughts about this !
>
> your scheme makes it hard/impossible to create a multichannel plugin that
> accepts data from different plugins. why ? because the data would have
> to interleaved, and that would mean that the host would have to do
> this. we want LADSPA hosts to, in the words of a good saying about
> OS's: provide mechanism, not policy. having the host interleave data
> from multiple plugins is a big step in the direction of policy, and
> should be left alone.
Yes... I could *possibly* accept a stereo data type as a performance
hack in some cases, but I'm afraid those plugins would just generate
lots of mono<->stereo converters in most people's processing nets.
Also, what is stereo? Analog mixers where the buses are transparently
usable as parts of a stereo signal with up to 8 channels have been
around for quite a while already. I wouldn't be surprized if some of
the >2 channel effect processors already can handle all channels as
channels of the same stereo signal.
Anyway, N-channel stereo *can* be done very easily with my
data-type-hack-in-progress, by adding suitable data types, in case it
turns out that it's really something that saves more power than it
wastes in many systems.
> >> With my proposed enhancements, LADSPA could easily become
> >> _THE_ plugin API (at least for audio streaming) for Linux.
>
> since there is no *other* plugin API, right now it appears to be _THE_
> plugin api by default :)
Very true! :-) Where's the competition? Will it be between different
mutations of LADSPA...? That would be cool, as long as they're
similar enough to merge into some final API that everyone can accept.
//David
.- M u C o S --------------------------------. .- David Olofson ------.
| A Free/Open Multimedia | | Audio Hacker |
| Plugin and Integration Standard | | Linux Advocate |
`------------> http://www.linuxdj.com/mucos -' | Open Source Advocate |
.- A u d i a l i t y ------------------------. | Singer |
| Rock Solid Low Latency Signal Processing | | Songwriter |
`---> http://www.angelfire.com/or/audiality -' `-> david_AT_linuxdj.com -'
This archive was generated by hypermail 2b28 : Sun Mar 26 2000 - 00:55:10 EET