A new approach to parallel functional programming
G Marino, Giancarlo Succi · 2002
Functional languages exhibit 'implicit parallelism' which has been used in attempts to achieve an efficient execution on parallel architectures; as a matter of the fact, the work has been focussed on process parallelism, leaving almost unexplored the field of data parallelism despite it seeming quite promising. To exploit data parallelism properly it is necessary to design a suitable collection. To do this the authors introduce a new data structure, the bag, and they give an intuitive explanation, a formal definition and some examples of it, evidencing its nondeterministic aspects and showing its impact on an effective implementation on a parallel architecture. Afterwards some practical justification of the usage of bags are presented. Finally they discuss some experimental results, which are remarkably encouraging, and draw some conclusions.>