Jens M Andreasen jens.andreasen at comhem.se
Fri Nov 6 20:09:39 UTC 2009


I e-mailed you several times ... Am I blacklisted somehow?

Anyways, the new fancy €70 card (Gigabyte GT220 OC) is in place here
now. It is specced at approx 220 (bogo)flops. So far I have only been
able to milk it for something like 50 Gflops using previous work naively
scaled up to fill out the void.

This at 3x16 samples @ 96kHz (no, not a typo), 128 channels in/out
oldstyle slow PCIe 1.a, runlevel 2 and no monitor (this is important!).
256 channels works if you can do away with 3x64 (and the cost of the
roundtrip to the GPU being 0.3ms rather than 0.16ms)


On Tue, 2009-08-11 at 14:43 +0200, Adrian Knoth wrote:
> On Tue, Aug 11, 2009 at 01:54:39PM +0200, Jens M Andreasen wrote:
> > > It's not ideal, but assembling all the jack buffers into one big one  
> > > is not going to be that much load on the CPU.
> > OK .. Adrian Knoth showed some interest and says he knows his way around
> > in jackd as well as a colleague involved with CUDA. If the idea after
> > evaluation does not appear to be worth the effort, then we'll just drop
> > it by then.
> Speaking from the user's point of view, LV2 would be the way to go. It's
> just convenient to have all the settings saved in an ardour session,
> nice GUIs and so on.
> This could boil down to have a collecting thread somewhere and just
> small control plugins for getting the actual data.
> But I'm sure we'll find out. ;)
> > That would have to be a collection of generally useful plugins, at least
> > 32 channels wide to be worth it. A mega plugin so to say. This ain't no
> > lawn-mover you can turn around on a platter. Doing little things here
> > and there /only/ would be very difficult in general.
> I could imagine a generic all-in-wonder channel strip. (to the channel,
> it looks like a channel strip, but it's actually 32 or more channels in
> parallel).
> Which would mean: dynamics section (compressor, gate, gain), EQ, perhaps
> some fancy stuff like your rubberband or pitch correction in general,
> perhaps FFT analysis or at least FFT transformation, so subsequent
> plugins can operate in the frequency domain.
> Or whatever. ;)
> > processor it is running on. Or else you'll end up with 640 identical
> > channel-strips rather than something like a synth-collection, 64 fully
> Doesn't sound too bad to me. ;) Though I could perfectly live with 128
> identical channel strips plus your synth running, if switching kernels
> is feasible.
> > To put things in some economical perspective, I am talking about
> > upgrading this tiny desktop-machine to having bandwidth and processing
> > power twice that of a current top-of-the-line Intel Nehalem for less
> > than $200, maybe around Christmas.
> Christmas? That's ambitious, but hey, I guess we could "borrow" lots of
> code from existing plugins and chain them together.
> Anyway, it's a cool project.

More information about the Linux-audio-dev mailing list