OpenCL parallel memory transfers

“Since compute capability 1.1, it is possible to overlap compute with one memcopy to or from host. Compute capability 2.0 improves on this by enabling a second parallel copy operation in the opposite direction at full speed (PCIe is symmetric)”.

Can OpenCL do this at the moment? Are there any other Fermi features OpenCL does not yet support? (i.e concurrent kernel execution).

Thanks in advance for any help.