xtensor-fftw

xtensor-fftw copied to clipboard

Use xexpression as input for fft functions

16

I don't know if it makes sense and if it's easy to achieve it, but it would be nice to not have to evaluate the xarrays for a call of...

martinRenou

Implement high performance interface

3

The simple numpy-style interface (#6) will by default hide some of FFTW's features, opting for simplicity and default settings for decent performance. We will also implement a high performance interface...

egpbos

Replace template specialization workaround with `= delete` syntax

2

We use template specializations to make sure we get the correct precision. The problem, for instance, with a non-template `f(const xt::xarray &input)` is that it will also compile when you...

egpbos

Build demo

5

A cling Jupyter notebook showing interactive usage. Something with an image, something with filtering, maybe something with audio. Maybe there are numpy.fft notebooks that we could draw inspiration from as...

egpbos

Follow some good suggestions from https://youtu.be/bsXLMQ6WgIk. - "Use `add_library` for installation of header only libraries", which xtensor-fftw is (xtensor itself as well). - Set `CXX_CLANG_TIDY=ON` by default.

egpbos

Use e.g. [licecap](https://www.cockos.com/licecap/). Put it on Twitter and maybe on the documentation page under Interactive examples.

egpbos

Put cling pragma for loading fftw3 in headers

1

As in https://github.com/QuantStack/xtensor-io/blob/a57e3b90ca1025a180c5813c56b8a344ec2d8165/include/xtensor-io/xaudio.hpp#L20-L21, we can put the library loading step in the headers themselves, which saves cling/notebook users from doing it.

egpbos

Real FFTs on arrays where the last number in the shape is an odd number require special care. In particular: an inverse rFFT cannot automatically determine whether the output should...

egpbos

Row major format

3

Data arrays in FFTW must be in row major format (http://www.fftw.org/fftw3_doc/Row_002dmajor-Format.html). Currently, xtensor-fftw defines output arrays as row major, but it does not check for row major-ness on the input...

egpbos

Document destructive inverse real FFTs

1

The default behavior for inverse real FFT functions (*c2r*) in FFTW is to destroy the input arrays during the calculation. This is probably very unexpected for most users. We should...

egpbos