AMD has an open source SDK called "HIP" that's specifically designed for porting CUDA to C++:
https://rocm-documentation.readthedocs.io/en/latest/Installation_Guide/HIP.html
This would also theoretically be useful for getting it to run on CPUs which isn't so crazy nowadays seeing as we have the likes of the Threadripper 3990X (64core/128thread) or even the similarly-configured 64c/128t Epyc 7742 which can be ran in a dual-socket config for a whopping 128cores/256threads in total.