Is it possible to make a Jitter/OpenGL setup on a MAC Pro with 4 x NVIDIA GeForce GT 120 using 1 monitor for PATCH/GUI and the 7 remaining screen outputs for individual OpenGL destinations. I’m not planning on stitching the displays together.

Is a setup of this kind possible? How is processing shared among the various cards when using multiple renders?

thanks in advance

Piotr

you will want to set the @rect attribute of each of your windows to be the exact coords of the two displays. any overlap of a jit.window onto another display/graphics card will significantly impair performance. jit.displays is your friend for this information.

another thing that i’ve found to help, is after loading your patch, and before you start drawing, re-instantiate each of the gl contexts. i don’t know why, but this seems to help. you can do this easily by toggling either the @fsaa or @floating attributes of jit.window.

post a patch for further help.

http://www.vimeo.com/4293627

this thing will run for 5 years , started 7 months ago
i only did the technical stuff not artistic.
it runs in max4

first put the windows on the right screen and then init the gl worlds and the shared_contexts
make sure you share contexts with two screens that are on the same gpu
if you share between two gpu’s kernel panics will follow…..

until now i did not find any way to automaticaly see wich screen is on which gpu
so i had to manualy find out the right positions

each output of the videocards first go to a vga amplifier and then to a projector
if a projector is taken out for maintenance and the mac has to reinit the shared_contexts it still sees 8 video outputs . if it would not see the 8 but 7 or less osx will reshuffle the screens to other positions and the patch is not able to share the right contexts with the right screens/gpu’s. gpu crash will follow….

-mj

excellent info for anybody considering a multi GPU patch in Jitter.

-pp