[Date Index] [Thread Index] [Author Index]
Re: compile speed
On Mar 21, 8:42 am, "Boson" wrote: > > since i need high values of nx,ny (~5000) and the loop scales as > nx*ny^2, But if you really want to run this for 5000x5000 matrices, I suggest you write the program in some performance oriented compiled language like C/C++ or Fortran. I doubt that you can get acceptable performance with Mathematica for this specific problem. If you really want to use Mathematica, you can try improving the Fold[ ] part in testnocomp3 by working backwards in the list and calling Random only as many times as necessary.