.Net Standard 2.0 implementation of DSP FFT
Performance numbers for FFT in the digital signal processing library were based on a native implementation of the algorithms in .Net Framework. The OSS release of Trill and all of its features must be .Net Standard 2.0 compliant, and it cannot rely on any library for which there is not proper licensing. We need to do one of the following:
- Implement our own FFT library over .Net Standard 2.0 (not preferable)
- Find a .Net Standard library over which we can take a dependency (proper licensing, good performance)
- Find a native library over which we can take a dependency and implement .Net Standard interop (proper licensing, good performance) (not preferable)
One option to consider: https://github.com/mathnet/mathnet-numerics.
I especially like that the library you mention has idiomatic F# extensions. Great idea.
For what it's worth, we've made the branch that we're developing the DSP extensions public in case the community would like to contribute or comment.