I'm new in OpenCL programming, I hope to use AMD GPU to accelerate some bioinformatics programs. I only have some experience on NVdia CUDA, although it also support OpenCL. Now I plan to extend my work to other GPUs, especially AMD-ATI's, so I decided to try OpenCL.
Here is my question,
OpenCL introduces a device-independent strategy to load run kernel programs, that is to load source code and compile in the GPU and run it. I am afraid it might be a time costing procedure in my future programs. Then I also found a function named "clCreateProgramWithBinary", which is said to be able to load a prebuild binary kernel program. However, I could not find any information of how to prebuild a binary kernel.
Anyone knows how?