Re: FFT proposal for gsoc 2021

From: Janek Kozicki (janek.listy.mailowe_at_[hidden])
Date: 2021-03-31 22:19:18

Stefan Seefeld via Boost said:     (by the date of Wed, 31 Mar 2021 15:54:57 -0400)

> Hi Eduardo,
> 
> On 2021-03-31 2:58 p.m., Eduardo Quintana via Boost wrote:
> > Hi Janek,
> >
> >> I think this is a mistake. Best approach is to not duplicate existing
> >> work, especially when it is written by the experts in the field.
> > You are right. But I cannot say that the main goal is to design an FFTW wrapper.
> 
> I agree, that wouldn't sound very appealing. I think the best way to 
> look at it is to design a modern C++ FFT API, with a big "I", i.e. an 
> emphasis on "interface", and to test that API by binding it to one (or 
> two, if time permits) existing libraries.
> 
> Testing such an interface has two sides:
> 
> 1) Is the API accepted by users, i.e. is it convenient to use ?
> 
> 2) Is the API implementable, i.e. can it be seamlessly and efficiently 
> bound to existing implementations ? (To test this carefully this 
> requires more than one backend, to avoid the API adopting idiosyncrasies 
> from the first backend it was developed around.)
> 
> In a nutshell, the above involves a substantial amount of work. 
> Describing it as "design an FFTW wrapper" wouldn't do it justice. :-)

Yes! Make the interface general. Test with various examples (e.g.
matrices) and/or libraries. With the future goal to allow performing
Fourier transforms of everything that provides Nth root of unity.
Maybe even a template check + a compiler error when such root is not
found.

This can help in creating a sound design of the interface.

best regards
Janek

--
Janek Kozicki, PhD. DSc. Arch. Assoc. Prof.
Gdañsk University of Technology
Faculty of Applied Physics and Mathematics
Department of Theoretical Physics and Quantum Information
--
http://yade-dem.org/
http://pg.edu.pl/jkozicki (click English flag on top right)