We also give results concerning latency hiding: a technique which aims to reduce the impact of high latency networks by ensuring an adequate supply of work. Somewhat surprisingly, our results show that more aggressive work creation strategies are less effective for latency hiding especially at very high latencies.
Though the existing Prograph environment offers concurrent programming in the form of a threads package, to incorporate a seamless mechanism for true distributed computation would require a change to the Prograph system, which is currently infeasible. To overcome this problem, we have constructed a translation system which exports Prograph programs to some other representation; the parallel features are then added at this point.
We describe a mechanism for distributing Prograph programs via a translation to Lisp. The effort in distributing the program is then in the domain of the Lisp environment. This paper discusses the work-in-progress regarding the distribution of Prograph programs using Lisp as the parallel engine.
This paper presents the results of using this new profiling tool in the analysis of a number of Haskell programs. The overheads of the scheme are discussed and the benefits of this new system are considered. The paper also outlines how this approach can be modified to trace and debug Haskell programs.