To sum up what you so far have responded: For data distribution it seems, OpenMP and MPI are still the best implementation if you expect to have a constant runtime for each of these tasks. On the other hand, for task distribution (say, ms > 10 or 100, and variable runtimes), any of the higher languages (e.g., Python, Erlang, Clojure) are acceptable abstractions to ease their development.

There seems to be no language that truly would facilitate distributed computations (multicore & multiprocessor) significantly, however (maybe Haskell?). Is that really it, there is no language that is specifically apt for the new multi-core/-processor machines popping up everywhere that will make programming life easier for those architectures, especially when I want to do some algorithmic speed-ups? I am a bit disappointed, in a way.