Implicitly-threaded parallelism in Manticore

Abstract

The increasing availability of commodity multicore processors is making parallel computing available to the masses. Traditional parallel languages are largely intended for large-scale scientific computing and tend not to be well-suited to programming the applications one typically finds on a desktop system. Thus we need new parallel-language designs that address a broader spectrum of applications. In this paper, we present Manticore, a language for building parallel applications on commodity multicore hardware including a diverse collection of parallel constructs for different granularities of work. We focus on the implicitly-threaded parallel constructs in our high-level functional language. We concentrate on those elements that distinguish our design from related ones, namely, a novel parallel binding form, a nondeterministic parallel case form, and exceptions in the presence of data parallelism. These features differentiate the present work from related work on functional data parallel language designs, which has focused largely on parallel problems with regular structure and the compiler transformations -most notably, flattening -that make such designs feasible. We describe our implementation strategies and present some detailed examples utilizing various mechanisms of our language.

Parallel Haskell implementations of the N‐body problem

“…It allows to specify parallelism on several levels in a large-scale applications, typically using explicit synchronisation and coordination on the top level [45] and combining it with implicit, automatically managed, fine-grained threads on lower levels [17].…”

Programming in Manticore, a Heterogenous Parallel Functional Language

“…These include transformations that implement the threading policies for the implicitly-parallel constructs, which we describe briefly below (Section 5.3); a more detailed account of these transformations is given in other publications [75,32]. Lastly, we compile nested patterns to simpler decision trees using Pettersson's technique [58].…”

Program Parallelization Using Synchronized Pipelining

“…However, after partial completion of the first iteration, the while (j < length) { c = j; ... while (c < j + 2*i){ ... c++; [2,3]. The parallelization technique we propose allows the second loop iteration to gradually progress in parallel with the first one (and successive ones), introducing synchronization primitives in order to preserve the original semantics.…”

“…... while (j < length){ while (c-j<2 3 We use superscripts to denote which loop variables belong to and subscripts to refer to the value of the variable at a given iteration of its loop. Our function λ essentially maps m → r m .…”

“…In a different line of work, the Manticore project is developing a parallel programming language for heterogeneous multi-core processor systems [3]. A main feature of the language is the support for both implicit and explicit threading.…”