At ICFP/FHPNC, Hans-Wolfgang Loidl suggested a joint "challenge" for the various projects and efforts within the FP/performance/parallelism intersection. The purpose is on one hand to demonstrate that there are actually some pretty solid projects going on, and on the other hand to investigate differences in approaches. Essentially something very similar to the old SICSA MultiCore Challenge.
We use the term "challenge" rather than "competition" because the point is not to make a quantitative contest like is done for SAT and SMT solvers, or the Computer Language Benchmarks Game, as our languages/platforms/domains/projects are far too diverse for such a contest to make much sense. (How would one compare e.g. a synthesied low-power FPGA design with a distributed Haskell program?) Of course, when it does make sense, direct runtime comparisons can also be interesting.
Concretely, the challenge will take the form of a problem description that is implemented by the various participants, with the results presented at FHPNC, and maybe complemented by a special issue of some journal with relevant papers (this is what happened with the SICSA MultiCore Challenge).
For example problems, we have looked at the Problem Based Benchmark Suite, with comes with existing (Cilk) C++ implementations as a reference. We are considering either rayCast or nearestNeighbors. Especially the latter is straightforward to implement naively, but is interesting to handle efficiently.
Interested in participating, or do you have comments about how we should do it? Comment below or send an email to athas@sigkill.dk.