Complex arithmetic in CUDA

Question

I often use complex arithmetic in CUDA and I need to define my own implementations, for example, transcendental functions ( sin , cos , exp , ...) on complex numbers. I want to definitely solve the problem by writing a library of such functions in order to consider the “large” number of cases and avoid being stopped at any time for new implementations.

I know that for some of them writing code may not be trivial in terms of the trade-off between accuracy and computational complexity. However, I also know that C ++ has its own sophisticated math library .

Is it possible to "rework" existing C ++ solutions for CUDA purposes?

Answer 1

You probably already found the answer, but here's mine:

There is an opportunity to "recycle" sources for a complex C ++ ISO library

UPD:

CUDA related: cusp :: complex <float> Structure Template Reference

I think you can try collaborating with the author:

C ++ std :: complex implementation for CUDA devices (i.e. compilation with nvcc)