Nick

@spectre-ns : Is this only nlogn for power of 2 lengths?

> FFT for non-2^N sizes is really slow. Hmm. My measurements of FFTW show that power of 3 FFTs are faster than power of 2: ``` fftw_r2c/2 6.65 ns 6.65...

I have now improved the benchmark: ``` #include #include #include #include #include static void xtensor_fft(benchmark::State& state) { std::random_device rd; std::mt19937_64 mt(rd()); std::uniform_real_distribution unif(0.01, 1); size_t n = state.range(0); auto x...

Do we have a paid github account? I got hit with a charge at a 1GB using git-lfs with github.

Now that `DebugMode` is deprecated, what should become of the empty constructor? The [documentation](https://adios2.readthedocs.io/en/latest/components/components.html#components-overview) (which appears slightly broken, btw) still says ```cpp // Do not use () for empty constructor....

@pnorbert : What's your opinion on the C++14 [`deprecated` tag](https://en.cppreference.com/w/cpp/language/attributes/deprecated). Is there any hope for compiler support on the systems we need to run on?

@gpoore : Well, in any case, your efforts on this project are highly appreciated. This is a really useful project.

[Here's](https://math.stackexchange.com/a/2486241/24355) a definition of the mollifier. The idea is that as x->0, you should consider absolute error, and when |x|>>1, the relative error is more natural. The mollifier interpolates between...

I have pulled some bugs out of the features by doing some [nan checks](https://raw.githubusercontent.com/boostorg/math/b14975fa7f8eb49aec89aa8f51fb1615afe126b5/test/math_unit_test.hpp) I would happily write the documentation and submit a PR for this, but I'm a bit...

@imneme : If I wrote a proposal for this, would you be okay with me writing "An Apache licensed implementation exists at github, with explicit approval from the author to...