CUDA.NET or OpenCL.NET?
Anyone using CUDA.NET or OpenCL.NET? I’m thinking that a few GTX 480′s each with 480 cores will provide some power in crunching structural risk. Anyone know of any pros/cons of using either CUDA or OpenCL?
Advertisement
Like this:
One blogger likes this post.
Not specific to the .Net implementations but Dr. Mark Joshi has done some ground breaking work in derivatives & CUDA. http://sourceforge.net/projects/kooderive/files/
This is the Kooderive project (“kooderive”)
The objective of this project is to create a library of code for pricing financial derivatives products using CUDA to achieve GPU programming.
He uses the thrust library to simplify (relatively) his coding, found here–> http://code.google.com/p/thrust/
Vince Fulco
Hi, Mark,
I can compare CUDA and PC cluster. I’m using CUDA because of underlying parallel model more universal than OpenGL. PC clusters lose efficiency fast with growing number of processors even for optimized communication graph. I can send my report ..
I used both for portfolio optimization for non convex criteria and non convex area. Simple example: find optimal portfolio around given that sum of changes (increasing or decreasing investments) is less than 5%.
I found that CUDA is perfect for such tasks just because of almost 0 delay for getting access to data. Even for linear optimization I got 100 times acceleration on 32 core CUDA (CUDA registry memory remarkably faster than PC memory). For real life linear optimization is not much useful..For portfolio optimization I’m using Taboo search with some improvement for finding feasible points.
@Renat
Please, send me your report : gil_software@yahoo.co.uk
thanks in advance