I haven't for a few reasons. The first is that there's a pretty good learning curve to the OpenCL interface. Granted, it is more flexible in terms of processor targets but that also adds complexity.
Your question is even more interesting when LabVIEW is in the mix because LabVIEW's execution system facilitates much of the concurrent execution abilities offered by OpenCL with an added benefit - it does so at both the algorithm and application levels. OpenCL is designed to handle the former. That's not meant to be a knock on OpenCL just my personal observation.
Now that I've offered my opinion, let me clarify a few things. This does NOT mean that we are unwilling to target OpenCL in the future thru a feature such as the GPU Computing module. The current NICompute architecture is fully capable of interfacing to OpenCL and it's execution engine(s). However, such support would be driven by customer needs. No development plans to support OpenCL are in place at this time.
