Summarising an experiment in parallel programming language design

摘要

Just before the skeleton/template concepts come into fashion in the parallel processing community, we started designing a new parallel programming language. This language (P3L) sports skeletons at the statement level and it has been implemented by writing a template-based compiler. With language, we performed a set of tests aimed at verifying different results: the suitability of P3L for massively parallel programming, its efficiency in generating "good" parallel code and the performance achieved with respect to traditional parallel programming languages. This paper aims at summarising the experience gained by our group with the "P3L experiment", and, in particular, at pointing out the main results achieved in one year of intensive experimentation with the P3L compiling tools.