'Q'
11-28-2007, 11:36 AM
Hi group,
I thought, what could I build with lots of fast floating point hardware and the vector math & random number libraries? Compound my 'what would I build' by having access to reasonable number of ClearSpeed boards. This needs to be both massively scalable and yet trivial to comprehend.
A Monte Carlo numerical integration example code is posted at:
http://developer.clearspeed.com/resources/samplecode/
MCT scales across many boards (tested with eight in one box so far, that's 1536 processing elements), and shows end-to-end deployment from Excel through to a socket-based-server executing the Monte Carlo engine.
MCT shows how you can present massively parallel machines in an implicit manner; along with all the system software stack.
Enjoy!
Q
I thought, what could I build with lots of fast floating point hardware and the vector math & random number libraries? Compound my 'what would I build' by having access to reasonable number of ClearSpeed boards. This needs to be both massively scalable and yet trivial to comprehend.
A Monte Carlo numerical integration example code is posted at:
http://developer.clearspeed.com/resources/samplecode/
MCT scales across many boards (tested with eight in one box so far, that's 1536 processing elements), and shows end-to-end deployment from Excel through to a socket-based-server executing the Monte Carlo engine.
MCT shows how you can present massively parallel machines in an implicit manner; along with all the system software stack.
Enjoy!
Q