>Does the alsa midi RX code timestamp the incomming data? That
>would make the processing much easier.
No, it doesn't. It might not be a bad idea, but it implies the use of
something other than a raw MIDI stream, since there is nowhere to
"put" a timestamp in the actual MIDI data. The ALSA sequencer itself
does this, but thats done later than when the interrupt actually
occured. And besides, quite a few MIDI interfaces have FIFO queues
themselves, and any truly accurate timestamping needs to be done
before the data goes into their FIFOs.
>In fact this trick could be used to sync (in software), multiple
>consumer cards....., Going to have to have a play with that,
>select on all the fd's, when a card unblocks, make a note of the
>TSC, then compute the drift and sample double as needed to maintain
>lock, the devil is in the detail, but it might work.
it might work, but is it worth it ? a cheap card with 2 output
channels costs US$40-80. If you wanted 16 channels, thats
$320-640. This is firmly in the range of *real* multichannel
cards. The only place i can see godawful hacks like this being useful
is if for some reason you really want 4-8 channels, but even, i
suspect there are some sensible h/w options that make more sense.
--p
