4

I'm looking for a programming language that would scale well on multiprocessors and distributed systems, and is able to work well with the GPU for number crunching. What do you think, is Erlang and CUDA a good match?

LE: I want to use it for image processing: feature detection, bundle adjustment and scene reconstruction; so it's fairly parallel. The GPU would do the computational intensive part and Erlang would just manage the tasks and shuffle data around.

titus
  • 5,512
  • 7
  • 25
  • 39
  • 3
    It depends entirely on what that "number crunching" is. Is it a problem that can be done in parallel? A good way to know is asking yourself if the operations depend on each other in any way? For example, the obvious gcd alg relies on the previous step. So you can't parallelise gcd itself as per that algorithm (although other, parallel, algorithms exist) but you could do lots of those gcds in parallel. Then the question becomes which is more efficient, another question altogether, again depending on what you were doing. –  Apr 11 '11 at 20:59
  • sounds it. Since that is very pertinent to the question I've edited it in; you've got an edit link just underneath the tags if you're not happy/want to reword/add more detail etc. –  Apr 11 '11 at 21:50
  • Seems that I'll just have to try how Erlang and CUDA work together and write a book about it – titus Apr 12 '11 at 14:14

0 Answers0