Re: [linux-audio-dev] defending simplicity

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

Subject: Re: [linux-audio-dev] defending simplicity
From: Juhana Sadeharju (kouhia_AT_nic.funet.fi)
Date: Mon Mar 27 2000 - 22:23:06 EEST


>From: Kai Vehmanen <kaiv_AT_wakkanet.fi>
>> You define a new type 123. Your plug-ins knows that type and can
>> discuss together. You may help the engine by telling how much bytes
>> your type is.
>
>Hmm, we are not connecting plugins to each other. We are defining
>a standard API for a callback function that modifies an area of memory.
>The host calling this function must know the exact data format.
>I think it's unrealistic to assume that the whole processing network
>(inputs, processors, outputs) is based on LADSPA-plugins.

I have not assumed so.
The host can query if the datas are float or double or whatever
but that never happens at runtime in the core audio engine.

Don't assume the host can handle and understand the audio data!
It doesn't happen in my engine, for example.

But whatever, only disagreement is that you're allowing only float
(or double) type in the plug-in. As you say, we could have several
plug-in formats but that really makes the mess. Another type
plug-ins cannot be made by recompiling but they have to be rewritten.

>If we don't use a standard format, we'll end up in a situation where
>there are LADPSA-plugins for ecasound (STL-vectors of sample data),
>LADSPA plugins for Paul's latest Ardour version (26channel 24bit-in-32bit
>data), LASDPA plugins for aRts/MCOP, etc etc ... Yep, sure we have a standard
>plugin API, but most plugin-host combinations are incompatible.

But we have can automatically find out are they compatible or not!
That is perfectly ok.

And the above kind of mess will never happen --- if only you have brains,
that is. Naturally 26 channel data is as soon as possible converted
to 26 x 1 channel data with a plug-in provided by Paul.

>And 32bit vs 64bit vs 128bit is not an issue. You can just compile for
>multiple targets, without any changes in plugin or host code.

Exactly, it is that easy --- and I wonder why you think it causes
extra trouble for plug-in programmers.

Juhana


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

This archive was generated by hypermail 2b28 : Mon Mar 27 2000 - 23:30:03 EEST