Anytime Soon? Async mem transfers with Arrays?

Are there any plans in the near future to implement asynchronous memory transfers from host to device when using CUDA arrays (i.e.: when working with textures as opposed to global memory)

Also, is there any chance of this working with current gen hardware (GTX 280)?

Thanks :)

Also, there’s one point I’m confused about from the documentation …

If I perform a cudaMemcpyToArray with a cudaMemcpyDeviceToDevice flag, will that call be asynchronous (with host execution)?

I would tend to think that it would be, but I haven’t found a definitive answer so far.