lion roars? it doesn't appear so
lion roars? it doesn't appear so
- Subject: lion roars? it doesn't appear so
- From: Paul Davis <email@hidden>
- Date: Thu, 15 Sep 2011 09:13:53 -0400
we've begun to get email from customers using mixbus on lion. the
story doesn't look good. here's a summary of the two most significant
points:
1) increased CPU load when executing the same code with the same session.
- specifically, a mixbus session that runs at about 50% JACK
DSP load (*)
with Snow Leopard takes 65-80% on the same machine with Lion
2) plugins fail AU validation that pass it under Snow Leopard.
- specifically, only 2/3 of the Waves plugins in their V8 release pass
auval validation inside Logic under Lion, but they all pass under
Snow Leopard
In addition, certain plugins (eg. PSP NobleQ, or T-Racks LA2A) work
correctly with Snow Leopard but with more than 4 instances of either,
the host (mixbus) crashes under Lion. They also generate much higher
DSP loads under Lion.
The CPU load seems to me likely to be caused by the new audio server
that we can see signs of on Lion. Stephane Letz raised this and got no
response, but it appears that there is now an additional server,
presumably not entirely unlike JACK, interposing between hardware and
*all* software on Lion. This makes systems like JACK much less
efficient, probably. Or perhaps not. Since this new
feature/server/thing has not been mentioned/discussed/acknowledged by
Apple, its hard to know.
The plugin issues seem likely to be related to some tightening up on
the AU spec in Lion, but again without any obvious announcement from
Apple. Or maybe there was and they just never bothered to tell the
"coreaudio API" mailing list about it. Who knows.
Either way, it doesn't look like much of an improvement. Have other
host authors and/or plugin developers noticed similar problems with
Lion?
--p
(*) JACK DSP load is a measurement of how much of the time available
for processing a given chunk of audio is actually used. If the
processing took precisely zero seconds, the JACK DSP load would be
zero. If it took the entire time represented by the audio, the JACK
DSP load would be 100%. the measured time includes everything done
from when JACK is woken by the coreaudio driver to when it blocks
again waiting for the next driver wakeup. It is not directly
equivalent to CPU load, but is closely related to it. Generally on
both OS X and Linux, DSP loads of around 80% are where the system
becomes unreliable for audio.
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Coreaudio-api mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden