JuliaApproximation / FastTransformsForwardDiff.jl

A Julia package to support forward-mode auto-differentiation for fast transforms
MIT License
13 stars 0 forks source link

Should FFTW work on Duals via method overloading? #5

Open michaelsachs opened 2 weeks ago

michaelsachs commented 2 weeks ago

FFTW currently does not work with Dual numbers.

using FFTW, DualNumbers
duals = [Dual(1,1), Dual(2,2)]
fft(duals)

This code returns ERROR: MethodError: no method matching plan_fft(::Vector{Dual{Int64}}, ::UnitRange{Int64}).

In this conversation, it sounds like FastTransformsForwardDiff intends to overload the required methods to make this snippet work. However, I am still getting the same error with the code below. Am I interpreting the discourse replies correctly?

using FFTW, DualNumbers, FastTransformsForwardDiff
duals = [Dual(1,1), Dual(2,2)]
fft(duals)

ERROR: MethodError: no method matching plan_fft(::Vector{Dual{Int64}}, ::UnitRange{Int64})

dlfivefifty commented 2 weeks ago

I think this package should be mothballed in favour of extensions....

But note this package only supports ForwardDiff, not DualNumbers