Giloo

The code is [here](https://github.com/GillesDuvert/gdl/blob/triangulation/src/triangulation.cpp) for the calling part, and [here](https://github.com/GillesDuvert/gdl/blob/triangulation/src/delaunator/delaunator.cpp) for the delaunator.cpp changes. Basically I did [this](https://github.com/GillesDuvert/gdl/commit/3c7a6b3fab3f8bac2a10f76247fd58dc9d16fd8e#diff-22c35d7493b44c6e4e5c02b6101e7f2b), (1.618... being the golden number). It is just a "works for me"...

I'm pretty sure it will not always work, sorry if I seemed to say so. Concerning the golden number, I experimented that the triangulation was always OK for square 'arrays'...

As I said, a 1602x408 grid was enough. That is, the x values were [0./1601...1601./1601] and y=[0./1601..407/1601], double precision. It gives this:  The problem does not occur (for this...

Nice finding, a bug!

dSFMT outputs need to be in aligned memory variables. Until now (C++11) this was only ascertained if Eigen:: was used, since in this case only all memory allocations are made...

Appleman does show a figure, but it will be dark if the device is in decomposed state. I've forced device to be in decomposed=0 in #1261

@slayoo, I believe APPLEMAN is not useful as a test and was not used to this purpose until now. Not that the Mandelbrot Set is not nice, but if we...

That's the point of the interactive_tests, no? But indeed they are, er, interactive, so somebody has to run them and check. Far from perfect indeed...

closing and restarting to check influence of https://github.com/gnudatalanguage/gdl/pull/1337

Thanks @alaingdl As always, there is something in the external libraries or installs that create discrepancies between my personal experience and those of other users. Concerning the performance of surface...