Port Verdict to GPUs - Githubissues

Use C math functions instead of C++ ones for best compatibility with CUDA/HIP (you can use them directly in device code)
Use constexpr for constants whenever possible so that they can be used directly in device code
For non-constexpr constants, wrap in functions instead
Add VERDICT_HOST_DEVICE macro to apply __host__ __device__ specifiers to all functions

sandialabs / verdict