I want to develop an application with OpenCL to run on multiGPU. At some point, data from one GPU should be transferred to another one. Is there any way to avoid transferrin
