Hi! I am working on a performance estimation based on OpenCL framework. For GPU architecture, there have been some works on the performance modeling. However, I am unclear about how does OpenCL utilizing multi-core CPUs. Can any one give me some information on this subject?
I am not sure i understood your question.
But openCL is meant to be cross platform and even compatible for heterogenous systems. You can find how CPU & GPU binaries are generated from chapter 2 of ATI OpenCL Programming Guide.