runningwild / go-fftw Goto Github PK

Is there a single precision version of FFTW for GO? Or what would it take to convert to single precision. In "C" the single precision version of FFTW is almost 4 times faster.

This is not an issue just a commit and question.
I have been using your go-fftw in my signal processing programs since you first posted it. I think it is a great addition to Go. It adds great value to Go for DSP applications that need FFTs. However, I prefer the older Array definition, Alloc2d(N1,N2) over the latest. The former allows easy access to Go's normal slice and array definitions, namely I used make(X[] complex128 , N1*N2) or slices of slices such as X[N1][N2] types in all of my programs. This made it easy to call and return these slices in my functions with go-fftw The current slice or array means I have to use some kind of conversion to get back to original slice or array format. Can you give me an example of how o easily convert back and forth between the two formats. Or better yet how to use the Alloc2d function in the present implementation of go-fftw?
I like your current implementation except for the inconvenient Array definition.

I would appreciate any advice that would make this current version easier to interface

GCSpec fails because memory allocated using fftw_malloc is never freed.

I replaced the " Measure" option in place of "Estimate" , In the interest of measuring the improvement in speed. It did not give the same results. I was wondering if that flag was be ignored.

Can this code be set up for 2d ffts by simply substitute Alloc2d? Then use the 2dfft call that is documented in FFTW?