This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

MCSDK HPC 3.x FFTW Performance: Is it FFTW 3.3.4 or FFTW 3.0.0?

I've been looking around at various documentation that gauges the performance of FFTW on Keystone devices. I'm particularly interested in its performance when only executing on the ARM core(s). This page seems to provide a lot of helpful information, but there's a potential discrepancy that I want to resolve. Near the top of the Wiki page that I referenced, it says "TI FFTW library build depends on a compiled, installed FFTW 3.3.4 in single/double precision for ARM/Linux." However, later when talking about the benchmarking that was performed, you'll also see "In FFTW 3.0.0 release, the following FFTW functions/types are being accelerated by DSP." So, which version of FFTW was really used to generate those performance results? This is very important because FFTW did not begin including support for ARM's NEON acceleration until version 3.3.1. Thus, if FFTW 3.0.0 was in fact used to do those benchmarks, then those results don't reflect acceleration that should come as result of the NEON extensions. Thanks for any clarity you can offer!

Doug Hogan