OpenSound - was Re: lockmgr patch
c.turner at 199technologies.org
Sat Jun 16 01:14:49 PDT 2007
Steve O'Hara-Smith wrote:
> On Fri, 15 Jun 2007 21:09:24 -0400
> Chris Turner <c.turner at 199technologies.org> wrote:
> This leads me to ask - have you ever tried playing a guitar through
> the DragonFly or FreeBSD sound system because it seems to me that the
> video recording A/V sync problem is about as demanding ?
I have not - part of the problem is that a lot of these apps are fairly
Linux/ALSA centric due to the (percieved?) needs of the above, so things
aren't yet compiling cleanly on DF - kind of a chicken and egg thing
that I'm chipping away at over time.
I suppose the standard answer from that perspective is that these are
still two separate types of scenarios - the TV card / audio record synch
is a passive, 'recording centric' type of thing that can be buffered
more before to catch all the hardware interrupts (e.g 1 second is ok),
whereas the music production case involves needing to hear things as
they happen in "real time" (TM) so that the input (i.e. the sound
resulting from a live performance) is "musically" in synch with the
output of other, possibly pre-recorded data (e.g. < ~10msec of
buffering/delay). So real-time monitoring latency is the key here due to
the human interaction component.
With the smaller buffering / stricter monitoring requirements,
scheduling deadlines become more of an issue as dropouts start to occur,
which is also possibly 'less acceptable' in the 'professional' case
(this is very use-case-dependent), and things may or may not be
compounded by heavy data processing, etc.
(e.g. playing / mixing 10 tracks from disk through 12 plugins while
recording 5 tracks and synthesizing 3 more kind of a thing)
More information about the Kernel