I have a system of ODEs which I'd like to solve with many (thousands) different initial conditions. In the net there are articles that CUDA can speed up such computations, but I haven't found anything for Mathematica. So is there something around?

  • $\begingroup$ You can always compile in CUDA functions into MM. Once you established the communication network between GPU and MM, usage of CUDA functionality can be programmed. Here is a resource: mariomulansky.de/data/uploads/ncwg.pdf $\endgroup$ – Eisbär Dec 17 '20 at 7:22
  • $\begingroup$ You can easily write your solution with CUDA (in this case, an ODE solver), and then instead of using a main method make a lib, and then load it from M with LibraryLink. There is an example in the documentation that shows how to do that. LL is extremely fast, no latency, and it doesn't really care what you implement (in fact, doesn't even know that you're implementing something that uses CUDA). I use it all the time, it's best way to use CUDA from M! $\endgroup$ – Andreas Lauschke Mar 8 at 2:48

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.