I'm a newbee of using AMD GPU. Recently I got an AMD W7100 card and wanted to implement the DirectGMA on the GPU data transfer. As I searched on the net, the DirectGMA is first configured via some commands like:
1. clCreateBuffer() with CL_MEM_BUS_ADDRESSABLE_AMD flag
2. Make that buffer “resident” by using clEnqueueMakeBuffersResidentAMD()
3. Wait for other GPU with clEnqueueWaitSignalAMD()
As I tested it on my GPU, the return value of the function clEnqueueMakeBuffersResidentAMD() is -5, which means CL_OUT_OF_RESOURCES. However, as I refer to the description of the API function clEnqueueMakeBuffersResidentAMD() in link AMDBusAddressableMemory (LWJGL 3.1.2 SNAPSHOT) (this is the only one that seems the most official), there is no return error with CL_OUT_OF_RESOURCES. Therefore I don't know what is the problem.
Could anyone kindly help about this?
Thanks a lot!