I have a program currently written in CUDA and wish to translate it to OpenCL.
My intention was to make a gradual conversion, where one kernel at the time is ported from CUDA to OpenCL. Unfortunately, I have not been able to figure out how to share data between CUDA and OpenCL, i.e. how to create a cl_mem object that contains the same area of global memory as a float * created with cudaMalloc, or, equivalently, get a float * I can pass to a CUDA kernel from a cl_mem object.
Is this possible?