Subject: Re: [linux-audio-dev] MAIA status
From: David Olofson (david_AT_gardena.net)
Date: Sun Jan 14 2001 - 22:42:26 EET
On Sunday 14 January 2001 01:33, Guillaume Badaut wrote:
> Hi !
>
> I would like to know how much is MAIA completed at the moment. And
> is there some source code available ? (more recent than MuCos
> 0.6.0)
I was just about to hack away for a few hours, but decided to have a
quick look in the mail box. :-)
Currently, the version is 0.7.0, and the protocol headers look quite
a bit different from 0.6.0. (Apart from the logo being updated with
some new ASCII art. :-) Some major issues have been sorted out, and
although I'm still unsure about the best way to do some things, I'm
probably just a few hours from running the first MAIA plugin with
some properties and some LADSPA/VST style audio ports.
As to design, I *think* I have varispeed streaming ports (streaming
host interface/protocol included) figured out, but the best way to
find out is probably to implement it.
Anyway, the basic idea is that a varispeed stream channel is an
abstraction of a file, which a plugin or client can stream from. The
streaming is handled by a "Streaming Server" (which can be a thread
of the local host, a separate process or running on another machine
in a network), and the interface on that level is a basic streaming
API with Quality Of Service guarantees.
Plugins will use an interface that is based on abstract objects
called "Stream Points". A stream channel can have one or more of
these, and each one represents a contigouos region inside the stream.
When initializing a Stream Point, one has to specify it's size,
maximum forward seek speed and maximum backward seek speed. (This is
needed for the Streaming Server to set up it's buffering.) The
maximum seek speeds can be zero, which means that the window will
just be cached locally at init, and guaranteed to always be
accessible, until the Stream Point is removed.
As an example, a direct-from-disk sampler (without looping) would use
two Stream Points per voice; one for the sample start, big enough to
handle the "relocate latency" of the: current play position Stream
Point. (The relocate latency depends on the source media and the
Streaming Server, and is one of a few performance characteristics
parameters that Streaming Servers have to provide. Hosts will provide
a basic toolkit for calculating buffer sizes and other stuff using
this datas.)
When a note is triggered, a voice plugin will first tell the second
Stream Point to move to the first sample after the end of the window
of the first Stream Point. Then it will ask the host to gather the
data it needs for the current buffer. (That should be it local RAM,
as specified by the first Stream Point.)
When the plugin hits the end of the window managed by the first
Stream Point, it'll start bumping the position of the second Stream
Point to follow the current play position, so that the data will
always be in local RAM when the plugin asks for it.
Basically, the only thing the Plugin has to worry about is that at
least one Stream Point covers the part of the stream that the Plugin
will need for the next frame, and that it doesn't exceed the seek
speeds for any points without taking the relocate latency in account.
Then it can just grab the data using a host function call. (Well,
plugins could mess with it themselves, but *mess* would probably be
an all to proper word... The underlying logic is essentially a disk
cache on steroids.)
As to the protocols I'm going to play with right now (Property and
Buffer, vaguely corresponding to LADSPA control ports and audio
ports, respectively), I'm still thinking about whether or not to take
the full multiple datatype system into the Properties, how to deal
with process_add() and some other things. Need to hack some actual
plugin and host code before I can decide on any lower level details.
//David
.- M A I A -------------------------------------------------.
| Multimedia Application Integration Architecture |
| A Free/Open Source Plugin API for Professional Multimedia |
`----------------------> http://www.linuxaudiodev.com/maia -'
.- David Olofson -------------------------------------------.
| Audio Hacker - Open Source Advocate - Singer - Songwriter |
`--------------------------------------> david_AT_linuxdj.com -'
This archive was generated by hypermail 2b28 : Sun Jan 14 2001 - 23:09:15 EET