... But, it feels like Apple have pushed the idea of GPU-computing with the release of nMP
Apple is more so pushing the uniformity of the availability. Two GPUs standard on higher end MBP and MP ( and iMac although Apple appears to be suppressing that tandem a bit) should make it a more standard model for developers to plan for.
Besides that Apple isn't doing much that isn't already flowing out of the industry wide GPGPU trends as a whole.
The other part that Apple is pushing on is not getting into single vendor proprietary tar-pits with this new movement/trend-line. OpenCL as an option.
Broad application adoption of any form of OpenCL is going to give the overall GPGPU market more flexibility.
and since they dropped the nvidia/cuda option it should be in their interest to get opencl 2.0 working as fast as possible.
Not really. There are zero production deployed OpenCL 2.0 hardware right now. OpenCL 1.2? Yes. And they have since there is a significant fraction of the deployed/operational Mac hardware population that actually can support OpenCL 1.2 (e.g.,
http://www.anandtech.com/show/6886/intel-updates-opencl-driver-and-tools-for-ivy-bridge-and-haswell ). The millions of HD4000 and HD5000 Macs out there can take advantage with Mavericks' broad spectrum OpenCL 1.2 enablement.
Nvidia either didn't offer Apple a OpenCL 2.0 capable GPU card for the Mac Pro's previous design bake off or lost the desk bake off ( or both ). Nvidia is shooting itself in the head for Apple Design bake-offs by dragging their feet on OpenCL support isn't Apple's problem with OpenCL scheduling.
What Apple should be far more concerned about is better and more robust OpenCL tools. To be one of the founders of OpenCL Apple's supportive environment ( tools , education/support, etc.) around OpenCL seems at best "ordinary" and in some extent lagging.
To some extent not surprising since the "computation focused" GPGPU cards have largely been missing from the Mac and OS X line up while the Mac Pro product line was comatose for several years. So it is not all that surprising that the toolchains coupled to that computation focused hardware on other platforms is in a bit better shape than Apple's.
I would very much appreciate the option of sharing more complex data with the kernels which appears in opencl 2.0*. Therefore I am waiting for upgrade until they either have hardware that is supported by opencl 2.0 or cuda (i.e. nvidia).
It would be nice of the Mac Pro moved to OpenCL 2.0 capable hardware before any other Mac model, but given the track record over last 4 years a move within a same 6-10 month window would be progress. I think Apple is going to have to 'walk' before they start 'running'.
Did Nvidia ever get around to supporting OpenCL 1.2? As long as they play the "OpenCL has to loose for CUDA to win" game the number of design wins will continue to go down ( have disappeared from low end iMacs an MBP 15" models ).
OpenCL 2.0 is coming from Intel and AMD on next iteration parts. I don't think Apple is going to change upgrade timing just for OpenCL 2.0 though. The next set of hardware upgrades when delivered will probably have OpenCL 2.0 uniformity across the whole line up ( just not any particular role played there by Nvidia at all).