Can I use Parallella for nuclear research at home

[hamster]

Could I use the Parallella for a particle simulation that models the neutron flux in a spherical lump of fissile material, that is undergoing explosive compression?

As a pinch I would guess that you would need particles that would travel on average the mean free path of the neutron (3D location, 3D velocity - would a floating point engine be useful for that?) and then use math like that in ray tracing to calculate intersection of the ray with the surface of the different layers (apparently Ogres are not the only thing like Onions...).

Should the path end outside the fissile material then the neutron is lost, should it end within the material then you have the option of it being absorbed, bouncing off of or splitting the virtual atom it has collided with. To get the whole ball rolling you would need some way to model either the natural decay of the atoms involved (e.g. x random events per second) or and external source of neutron flux. Can anybody tell me the flux density of a beam from a a neutron gun?

By modeling a compression wave squeezing the layers of the device I hope to play with virtual critical assemblies in my spare time...

Or would I be better off doing this on the two and a half Raspberry Pis that I can buy with the $100 of pocket money I have saved for my nuclear research budget?

[glasspelican]

Ok lets not look at the epiphany chip yet.
with the parallella you get a dual core arm a7 cpu and 1 gig of ram
the raspberrypie gives you a single core arm a6 with 1/2 gig of ram

The a7 core is clock for clock faster then the a6.
given just that you should be able to do better on the parallella then a pair of raspberrys.
Lets say that running linux takes 256meg of ram,
each raspberry needs to run its own copy, leaving your application with 512m spread over 2 devices
on the parallella you only need to run one copy leaving 768m

Then we have the epiphany coprocessor on the parallella platform.
physics simulations tend to be floating point heavy.
so assuming that your project can be made to execute in parallel, the execution speed could approach that of my laptop. (theoretical performance of this part challenges my desktop, and trounces my laptop)
not bad for 5 watts.

If it is a single threaded application, you would not be able to take advantage of the second raspberry, or the second arm core
but you would still be better off with the parallella because of the faster single core speed.

Edit: if your under a time constraint, your probably better off developing your application on a laptop

[ysapir]

I think we've found our "killer" app....