I am interested in using MKL in VxWorks. I am aware from marketing material that latest VxWorks releases contain the Intel C Compiler and IPP libraries, however my current project will have non-power-of-2 FFTs which are only available in MKL. Is MKL ported to VxWorks? Is it possible to use an existing MKL Linux install on VxWorks (I see some posts here about linking in Intel-built libraries before formal support was introduced)?
Hello, did you know that IPP does have support for Discrete Fourier Transform Functions where the length of the vector can be arbitrary? Take a look at Page 401 of the Intel Integrated Performance Primitives for Intel Architecture Reference Manual, Volume 1. Let me know if that will work for you. Could you help me understand your target architecture and use case as well? In the meantime, I'm going to do some research on VxWorks support for MKL to give you the latest and greatest answer.
Hi Noah, Thanks for the response. I am aware of the DFT support, but isn't that O(n^2) instead of O(n log n) (ref: http://software.intel.com/en-us/articles/mkl-ipp-choosing-an-fft)? MKL provides O(n log n) for all sizes.
My platform is an Intel i7-3612QE -based VME 2-socket embedded board. It's a realtime application so we need the O(n log n) for a set of sizes, not all power of 2's.
Yes, that KB article is up to date. Parallel Studio XE which includes IPP doesn't support VxWorks but I am still investigating your options. Stay tuned.
I got the official word: the currently available Intel® MKL versions do not support Wind River VxWorks, but there are still some things we can discuss about embedded tools offline. I'll be in touch with a meeting.