I got my first Parallella around 1 month ago, and dove into it immediately.
I am a software developer who loves the subject of astronomy, so the N-body simulations appeal to me.
I first located some Nbody C language code on the net for a CPU and modified it to run the calculations on the 16 Epiphany cores.
This is one of the early videos (sorry for the poor iPad video quality, wish I could record the cdmi output directly)
This one has a super massive star in the middle of the screen and one of the smaller stars coded to leave orbit trail, kind of interesting to see some of the orbits the stars fall into around the massive one. In this one all the stars are white and equal mass, except for the super massive one in the middle of the screen. It is red but the iPad doesn't seem to pic up reds well.
Then I had seen a video of a simulation from a University where they color coded the stars by mass and had small trails behind each star to give a better indication of where it is going, so I copied that. The smallest mass stars are blue, medium mass white, large mass red (these show up as white in the poor quality video)
These stars are all starting at zero velocity all on the surface of the sphere, the gravity of the system collapses onto itself.
These videos all look better in person, I am not the best cameraman
Here is a cube randomly filled with stars, again zero velocity and it collapses onto itself. I find the star system collapses the most visually appealing.
I am now working on adding graphics the N-body code from the US Army lab github. If I can get the Epiphany cores dma copying the stars directly into the frame buffer like the mandelbrot example does, it should be a pretty fast demo I think. Will post a video of that here when I get it working satisfactory.