As for now we have a lot of possible OpenCL implementations - from FPGA and special accelerators like FirePro and Xeon Phi to mobile GPU and pure CPU (pocl for example) - today no more needs to have cpu-only implementation for any code that can be efficiently run with OpenCL.
Welcome to our forums! I've approved your topic, which should enable you to post freely in the future.
Main reasons for it not developing:
- Our devs (which is me + povmaniac) can only work in spare time. povmaniac tends to put a lot of effort.
- OpenCL is not very easy.
- Raytracing is not too easy on GPGPU either.
It will come, one day.