Talking with nickoppen in this thread; I decided to get a few base libraries up and running, but especially OpenCL and interop with GLFW for data visualization. Please let me know if anyone wants other broad-domain libraries linked up with at least a hello world or basic unit output.
Anyway, I was thinking about projects that would result in some helpful code and had this idea. Given the way the Epiphany cores communicate with each other, the routing pattern/speeds etc.. Do you think would be a good demonstration of how OpenCL kernels can be tasked with more optimal patterns?
I'm thinking that during the process of re-implementing the game to match epiphany (and the commuting of data in the game is quite similar) I could play with the base rules to make them match Epiphany to do constant-time operations on full memory banks. I think the end result of what I'll need to make could next be used to map standard NDRanges into a more 'grid-over-time' optimization.
Hopefully this can be a good entry-point demo for people making a transition from more common 1D/2D linear array optimization style to Epiphany optimization style.
Looking for feedback as to whether this might work out to be fairly enlightening on the architecture usage through OpenCL, if you know of any other/better demos or good tips for getting this to work =)