We have just got a copy of the MATLAB compiler and C Math Library for our UNIX workstation. The main reason was in an attempt to increase the speed of some code we have.
I tried it all out on a command in the Hi-Spec toolbox, called bispecd.m which took 40 seconds to run as a m file, 50 as a mex file and 70 seconds as a stand-alone C application. The code is highly vectorised, and needs complex values......
Heres the question!!! Does anybody know if by rewriting the code to use nested loops instead of vectorised code, will the resulting C stand-alone executable (generated with the compiler) be faster than everything else????
I tried the optimizations that were mentioned, namely bounding of matrices, but there was little improvement.
Any ideas on how to speed-up this code
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ ~~~~ Angus Murray ~~~~~~~~ ~~~~ Condition Monitoring Research Group,~~~~~~~~ ~~~~ Department of Engineering, ~~~~~~~~ ~~~~ University of Aberdeen. ~~~~~~~~ ~~~~ Tel: 01224-272986. Fax: 01224-272986~~~~~~~~ ~~~~ http://www.eng.abdn.ac.uk/cmrg ~~~~~~~~ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~