Subject: Re: [linux-audio-user] Re: linux-audio-user] [ANN]
From: John Check (j4strngs_AT_bitless.net)
Date: Tue Aug 10 2004 - 10:46:49 EEST
On Tuesday 10 August 2004 03:28 am, David Baron wrote:
> > > Problem with Jamin is that is a process to process thingie. Another
> > > program, eating precious CPU cycles, must be playing and pre-processing
> > > the audio to feed Jamin. I just do not have the CPU guts to run this
> > > way. Under that other OS, I can run this type of software as a
> > > standalone (file-to-file) or DX/VST plugin OK. The three-process
> > > (playing app, jack, Jamin, jack) system is just not efficient.
> >
> > While the JACK overhead is measurable, I doubt it's your main problem.
> >
> > JAMin uses an FFT for linear-phase filtering. This is quite expensive
> > in CPU, but sounds great. We made that tradeoff consciously, choosing
> > sound quality over CPU cost, recognizing that some older CPUs would
> > have trouble keeping up. Moore's Law is rapidly fixing that problem
> > even as we speak. JAMin only uses about 25% of my relatively old
> > Athlon XP 1800+.
> >
> > IIUC, most Windows mastering applications use lower-cost non-linear
> > filters, so they run comfortably on low-end hardware. That is a
> > reasonable business tradeoff for them to make.
>
> Yup. I use several FFT based plugins and the eat it up quite nicely. They
> get some improvement by using assembler rather than C++ for the math but
> they still eat it up. BUT, I can use them. If I use the worst ones or too
> many, then I need to destructively apply. Great examples are "CloneBoy" and
> "CloneEnsemble". I can make MIDI choirs sing (and well!), but not "live" if
> I am using both of these.
>
> > If your machine is close to being able to hack it, try using a large
> > JACK buffer size (-p2048 or -p4096). This reduces both JACK and FFT
> > overhead. Mastering does not require low-latency operation, anyway.
>
> My Qjackctl show 46ms latency now. I run my Windows junk at > 100ms
> routinely so as not to run afoul of plugins that neglect to use the
> lookahead calls. What do I set that here as well and give it a try.
>
> > > A standalone or LDASCP Jamin would be worthwhile for those of us with
> > > older equipment.
> >
> > You're welcome to contribute one yourself. The GUI is far too complex
> > for LADSPA, but there's nothing particularly complicated about adding
> > file I/O to JAMin, itself. We just didn't feel like working on that.
> > There are so many good JACK-based solutions already available.
>
> I might just try it. I fail to compile anything that wants QT3 (I do have
> it and cannoct figure out why the ./configure cannot find it) but other
Set QTDIR
> stuff will usually compile. If I do and can get that working, how do I
> contribute it to the project?
This archive was generated by hypermail 2b28 : Tue Aug 10 2004 - 10:56:51 EEST