Arnur Nigmetov

Interesting. I ran Hera code with q = 5 on the attached diagrams and it worked for default relative error 0.01. I understand that these files were generated from a...

@mrzv If I replace `double` with `float`, my executable gets stuck as well. It seems to be the familiar numerical issue: the epsilon in epsilon-scaling gets too small to guarantee...

Could we do something similar to numpy, where functions take additional dtype argument? By default it uses float, but we can also support double and long double, with the same...

I just realized what you were asking. Well, having input in float and doing computations in double cannot help in all situations, because in the equation for epsilon we need...