On Sun, Oct 6, 2019 at 11:03 AM Vladimir Nikishkin <xxxxxx@gmail.com> wrote:
>
> >>Oh, oh! Before we go real radio, something more like Matlab / Octave
> >>first would be great. Of the top of my head: matrix calculation,
> >>plotting, fast fourier transform, cross correlation, convolution...
> >>This is very useful for experiment with radio signal processing.
>
> Doesn't slib include pure-scheme matrix algebra and fft already?

Ah I didn't know (not a big slib user)

> To do anything fast nowadays you need gpu acceleration, and I'm not sure it's easy to rewrite from scratch. So the potential srfi would need to not just include all the functions, but possibly support different backends, i.e. pure scheme, blas, gpu-blas, etc.

To me it's more of making sure that replacing pure scheme with
accelerated backends is possible. So that the same Scheme code could
run in accelerated mode without modification. The actual blas,
gpu-blas, fftw... support can't be part of the srfi reference
implementation because it involves ffi anyway.
--
Duy