I know I’ve asked about this before, but it would be cool…
Some of us have been running 2 instances of vvvv and assigning them to different cores of multi core machines.
This works well to some extents, last time I did it, the patch went from 20-30 fps running 1 instance to 60fps by asigning 3 cores to the tracking and 1 to the output.
The issue we had however was that if you try and fullscreen the output and it has a higher resolution than 1024x768 it stopped the direct functioning in the tracking patch.
If there was a way to stop this happening it would add great possiblities to vvvv seeing as you tell us we’re not going to get multi-threading support anytime soon…
there is certainly nothing built into vvvv that would prevent this. could this be a vram limit? so you have two instances of vvvv with each having some ex9 renderers? and one instance is going fullscreen with one of its renderers? are those all on the same graphiccard? in span or split mode? could you try with different graphiccards for the two instances?
so you have two instances of vvvv with each having some ex9 renderers? and one instance is going fullscreen with one of its renderers?
Yes, both are on the same gfx card, in split mode, so maybe thats the thing, dont have a dual gfx system at the moment, its on my list when I get some prototyping time again!
And yes Im using set process.
Thanks for your questions, got some things to try :)
tried it on a system with 2 gfx cards (2x 8800GTX) … pretty much same thing.
Running multiple instances with multiple renderers works fine as longa as all renderers are in windowed mode. As soon as switching one of them to fullscreen, things become strange/unstable.
For example:
one Renderer on each graphics card / patch on graphics card 1
renderer on gfx 1 to fullscreen - windowed renderer on 2 keeps working
renderer on gfx 2 to fullscreen - windowed renderer on 1 stops working
trying more than one renderer fullscreen results in very slow renderer on gfx1 and grey renderer on gfx2
… hmm i think i have to use boygrouping again and there is not much to do for three of the four cores