Optimize random generator about sfizz HOT 5 CLOSED

MSVC is pretty bad too...

I'm good with this, I'd like to bench it against the std::normal after the box-muller transform or do you have a faster method for this?

On a side note, visually, the white noise I generate now looks quite different from sforzando's. I need to plot some histograms to understand what's going on.

from sfizz.

I think the question to ask as: does it sound the same? if so, fine enough.

Visually, I have run a plot as I implemented and the result was like this.

from sfizz.

I'm moving this up to the next milestone for the SIMD version and the Gaussian one.

from sfizz.

It's funny that you comment this now, I was researching gauss and I have a decent one at hand.
I'm going to attach the plots.

from sfizz.

There is a distribution which matches c++ normal_distribution with various values of variance.
It's nothing original, it's no.gnoise of faust, I just make a plot from it.
It accepts a quality settings that ranges from 2 (looks triangle-ish) to any integer values.

from sfizz.