Hi all, Does anybody have experience about optimizing for speed for the GSL ODE45 solver?
I am facing a need of calling the ODE45 solver 10^9 times. I am thinking of the following options: 1. Convert GSL ODE45 program into assembly code, optimized for Intel Architecture; 2. Convert the performance critical part of GSL ODE45 code into assembly code. Could anybody tell me roughly, from your experience, which part of the code is the most time consuming part? 3. Does Intel Math Kernel Library have a ODE45 solver that I can use? 4. Will other third party libraries help in this regard? Thanks a lot Michael.
