Subject: Re: [linux-audio-dev] matrix model vs. memory
From: Paul Davis (pbd_AT_Op.Net)
Date: Sat May 19 2001 - 20:36:11 EEST
>> >I'm not sure to understand, but why we can't have a separate
>> >"matrix_gain" plugin?
>>
>> what do you propose it would do?
>
>Given N inputs and M output and a matrix N*M of float values
>
> N
>out(M) = sum(in(k)*matrix(k,M))
> k=0
and how does it go about writing to "out" from "in" ? before answering
this, read on...
>Exactly like pcm_route in alsa-lib.
>
>> what do you suggest would happen if it was not present?
>
>I don't understand the question.
there has to be some underlying mechanism for moving data from an
output port to an input port. there has to be a definition of how
either a plugin or an engine causes the data to get from one place to
the other. it could be as simple as memcpy(), or something more
complex.
either this mechanism includes the possibility of a gain coefficient
to be used in the transfer or it does not.
if it does not, then your suggestion makes sense. but in that case,
the most common operation carried out on audio data (gain control)
becomes less efficient, since it has to pass through a plugin with its
own buffer(s).
either way, there is always an implicit or explicit matrix model
within the system, mapping the connections between ports. so the
questions are whether to (1) make the matrix model explicit and (2)
add the gain coefficient to this model.
--p
This archive was generated by hypermail 2b28 : Sat May 19 2001 - 20:52:55 EEST