Do you know of applications that use the hybrid approach of using OpenMP and MPI for running on SMP based clusters? I see a lot of apps written and studied in academia, but I am interested in commercial applications.
I'm not surprised that you haven't gotten any responses, yet. Running hybrid codes like you describe requires specific hardware configurations to be most effective. It may be a "chicken and egg" situation, but there aren't many end users that would purchase such systems. Those that can (e.g., automakers, pharmaceuticals, aerospace, etc.) don't add up to a number significant enough for commercial ISV's to devote the resources needed to develop and maintain a hybrid version of the applications along with just MPI or just threaded.
WIth dual-core processors, this might change since each node of even a small cluster (4 to 8 nodes) will be able to run threads on even a "single" processor node. Should this situation materialize, I think we'll still see a lag of time before ISV's catch on and are able to ramp up efforts to thread MPI codes.