Subject: Re: [linux-audio-dev] LAAGA - main components
From: Paul Davis (pbd_AT_Op.Net)
Date: Mon Apr 30 2001 - 23:09:09 EEST
>> >- if we need for some reasons to use poll(2) or select(2) between
>> >components of the full system, we have to use something useable with
>> >poll/select (by example AF_LOCAL sockets).
>>
>> thats the whole point. you can't have multiple process communicating
>> via OS objects if you want low latency (reliably). the components in
>> the system must be in the same process.
>
>Why do you think I want to? Nothing in pcm_shm/pcm_lbserver model forces
>multiple process, nothing forces to have OS objects for communicating.
then you'd better explain what you have in mind a little
further. because i cannot understand why or how any of what you've
alluded to so far would have any relevance to an in-process
system. the very terms of pcm_***shm*** and the loopback ***server***
are all connected with IPC.
>> if you don't agree with me, then i would like to see a working example
>> of this design with a 1.3msec deadline for audio processing.
>
>I've not a strong opinion for or against this statement, perhaps you're
>right, perhaps it's an urban myth, who knows ;-)
No, I've tried it. Its not an urban myth. On the x86, at the very
least, you can't get reliable execution within 1ms of more than 3
RT_FIFO processes if they touch more than a line or two of the cache.
>However it's definitely irrelevant as the model I propose *do not* force
>task scheduling every elapsed period.
In which case, how does anything happen if its not all in-process? and
where in the descriptions i've seen of the ALSA side of all this is
there any mention of in-process clients?
>And so what? We're discussing about the API for component communication,
>we are not discussing whether to use a single thread, multiple threads
>or multiple processes.
No, we *are* discussing that Abramo. There is a very big question
whether to support only in-process "components" ("plugins"),
only out-of-process components ("other processes"), or some
combination of both.
--p
This archive was generated by hypermail 2b28 : Mon Apr 30 2001 - 23:47:11 EEST