topic Call objective function in Parallel in jacobi matrix calculation function (djacobix) in Intel® oneAPI Math Kernel Library

Call objective function in Parallel in jacobi matrix calculation function (djacobix)

Maosi_C_ — Mon, 15 Sep 2014 14:51:43 GMT

I'm learning to use the Non-linear optimization in the Intel Math Kernel Library. My starting point is the the example code "ex_nlsqp_bc_c_x.c" in the directory "examples_core\solverc\source".

In its RCI loop (Line 202-294), the jacobi matrix calculation function (djacobix) is called. Since it will call the objective function (e.g. extended_powell in the example) many times and my real objective function is time-consuming, I am wondering are the evaluations of the objective function performed in parallel? How do I make it happen? Do I need to change the code? Or Only some project property settings is needed?

Pls look at the extended

Gennady_F_Intel — Mon, 15 Sep 2014 16:08:30 GMT

Pls look at the extended_powell routine -- you can manually threaded this circle and also this can be easy vectorized.

Gennady, Thank you for your

Maosi_C_ — Mon, 15 Sep 2014 17:11:17 GMT

Gennady, Thank you for your reply.

I understand the extended_powell routine is a simple forward model, which can be easily threaded or vectorized. However, the real forward model I am about to use is much more complicated and impossible to threaded or vectorized (accumulated variables depending on the previous time step). That's why I need to know does djacobix call the forward model in parallel and how.