After implementing a mostly lock-less version of the multi-core
support and optimizing it nicely (locks are only held for thread
synchronization, which occurs once per fluid_synth_one_block), I went
ahead and committed my changes. Basically all the voices are
synthesized to their own dedicated audio buffers, which turns out to
be not all that much data (FLUID_BUFSIZE * 256 for polyphony=256 is
64K for 32 bit float data) and simplified the logic greatly (no need
to handle buffer starvation). Once all voices have been synthesized,
the primary synthesis thread mixes them all to the
left/right/chorus/reverb buffers. At this point, I think the overhead
is pretty minimal. I ran up to 8 cores on my single core laptop and
saw fairly small overall CPU increase compared to single core. So I'm
much happier with this now.
I tested it on my little brother's core duo laptop and saw -F fast
rendering increases of anywhere from 11-25%. A decrease of 25% in
render time doesn't seem all that bad, if you consider that the
maximum theoretical reduction of 2 CPUs would be 50%. I imagine some
of the other code paths (reverb/chorus, etc) are probably rather
significant and could perhaps also benefit from multi-cores (perhaps
do reverb and chorus separately or something).
I also added audio.realtime=yes/no, midi.realtime=yes/no,
audio.realtime-prio=1-99 and midi.realtime-prio=1-99 as I discussed
before. The core threads take advantage of this, to try and be the
same priority as the audio threads. The possible priority
inconsistency is still an issue for programs embedding FluidSynth
though, but it would still work, just perhaps not optimally.
A couple of bug fixes got committed too, which I caught with valgrind.
The -F fast render option is really useful for benchmarking and
debugging. Thanks for adding this! :)
I'd like to hear of other user's experiences with the new multi core
support, on multi-core machines (a quad core or higher would be nice
to get results on). In summary, its useful if you want to maximize
the number of potential FluidSynth voices on your system or -F fast
render quicker to an audio file. The largest benefit is seen when
there are a lot of voices in a MIDI composition or sequence. To try
it, just pass -o synth.cpu-cores=N, where N is something like 2, 4, 8
or 256 ;)
Cheers.
Josh
_______________________________________________
fluid-dev mailing list
[email protected]
http://lists.nongnu.org/mailman/listinfo/fluid-dev
