I've seen articles on the web site which state that the Linux version of the MKL can be used with VxWorks. Is there a document that describes how to do this?
Would you please give the link to this article? We are not ready to work on that OS right now. Please check the Release Notes first in the case of similar "news".
Here is the link: https://software.intel.com/en-us/forums/topic/344794
It looks like there was a phone conversation that may have offerred some suggestions.
IPP supports VxWorks and for your case you can try to use Discrete FFT mode from IPP. The performance would be ~O(nLog(N)).
For more complete information about compiler optimizations, see our Optimization Notice.