comments on late joyrides re: rt performance

Benjamin M. Schwartz bmschwar at fas.harvard.edu
Sat Aug 16 16:14:48 EDT 2008


-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

victor wrote:
...
| Comparing the RT audio performance of the two, I must say
| I am disappointed with how worse it has got.
...
| One of the things I have been playing with is a little MIDI file player
| (one of the test/example activities in
devel.laptop.org/activities/csndsugui),
| which uses a soundfont synthesis engine.

While diminished RT performance is a serious problem, it occurs to me that
a properly designed MIDI file player should not require low latency at
all.  The ALSA audio design makes it easy for applications to request long
buffers..  For example, the Distance activity plays audio with

subprocess.call(["/usr/bin/aplay", "--buffer-time=10000000", fname])

to request a maximum-size buffer, precisely because of the latency issues
that I observed in testing.  (Yes, this is incredibly ugly, but using
gstreamer in a multithreaded python app is uglier.)

Interactive sound synthesis is a different matter, but for pure playback,
perhaps you just need a longer buffer.

- --Ben
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.9 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iEYEARECAAYFAkinNTgACgkQUJT6e6HFtqSoMACeJLmArHDWvm5v3ngHh8xII/4i
ShIAn2Oour0zr/4YkBE0CR4ubDiRLlXz
=PbFO
-----END PGP SIGNATURE-----


More information about the Devel mailing list