Le 1 juin 09 à 14:16, Benno Senoner a écrit :
Hi Stephane,
what kind of latencies did you achieve with jack on windows ?
some time ago I tried to pipe LinuxSampler Win32 and Kontakt2 through
jack using the ASIO router and ASIO backend for jack (or does it go
thorugh portaudio which in turn uses ASIO ? I don' remember exactly,
it was 8-12 months ago)
and compared to LinuxSampler or Kontakt using ASIO directly jack was
much worse, each few secs I got a dropout under load.
And the latency of the asio router for jack was always 512 frames.
JackRouter reflect the currently chosen JACK server buffer-size for
ASIO clients.
I
don't remember what number of frames I used for jack out but
I tried several combinations and in no way I could match the latency
and reliability of using ASIO directly.
As hardware I was using a Dell laptop with ASIO4ALL (I think it's
intel hidef audio chip)
Now the question is: was the weak link the ASIO router for jack, was
it portaudio or is it the WIN32 IPC which is unreliable ?
AFAIK multithreaded VST hosts do use IPC calls to to support multicore
CPUs and it seems capable of achieving low latencies.
Difficult to say without more precise "profiling" and tests...
There is some discussion going on on the LinuxSampler forum about the
upcoming LinuxSampler version (the beta is already available)
and an user asked it the next windows version will come with jack
support compiled in. (it works with jack win32 too)
http://bb.linuxsampler.org/viewtopic.php?
f=2&t=339&sid=ed3163f7f87cfe0e45030e4bbf802fab&start=20
I did not compare the performance of LinuxSampler win32 using jack
directly as opposed piping it through the ASIO JackRouter.
This would be interesting to test, to see if the JackRouter component
is the "weak" part or not...
What do you think Stephane, can native jack clients on
windows achieve
performance which is almost at par as native ASIO apps ?
If we release LinuxSampler with jack support we have probably to ship
libjack (otherwise the app does not start) with the sampler and it
could be that it conflicts with an
already installed jack.
Well I also recently had this kind of "weak" link requirement for
libjack on Linux. I think OSX supports some kind of weak linking with
any compiled framework, but the situation is less clear on Windows on
Linux. A possible solution would be to provide a special
"libweakjack" library with the appropriate bahaviour for that.
So probably at this point the official LS will
be released without jack compiled in but it would be nice if jack will
soon become an established method audio/MIDI routing
on windows too.
Stephane, what are your plans regarding the windows platform ? I see
that jack1.9 for windows can be downloaded from
jackaudio.org download
page
so is it regarded as stable (built from the same codebase as the linux
version?) ?
Same codebase yes.
did you get any positive/negative feedback from
windows
users ?
No "stablility problems" feeback AFAICS, but people are maybe not
using the system in real low-latencies situations.
Stephane