A performance-oriented comparative study of the Chapel high-productivity language to conventional programming environments

Abstract

The increase in complexity, diversity and scale of high performance computing environments, as well as the increasing sophistication of parallel applications and algorithms call for productivity-aware programming languages for high-performance computing. Among them, the Chapel programming language stands out as one of the more successful approaches based on the Partitioned Global Address Space programming model. Although Chapel is designed for productive parallel computing at scale, the question of its competitiveness with well-established conventional parallel programming environments arises. To this end, this work compares the performance of Chapel-based fractal generation on shared-and distributed-memory platforms with corresponding OpenMP and MPI+X implementations. The parallel computation of the Mandelbrot set is chosen as a test-case for its high degree of parallelism and its irregular workload. Experiments are performed on a cluster composed of 192 cores using the French national testbed Grid'5000. Chapel as well as its default tasking layer demonstrate high performance in shared-memory context, while Chapel competes with hybrid MPI+OpenMP in distributed-memory environment.

Parallel distributed productivity‐aware tree‐search using Chapel

“…Chapel is proven $$$ 3.7\times $$$ faster on average for BFS, but $$$ 1.3\times $$$ slower for PageRank. In addition, we investigated in Reference 13 the parallel generation of Mandelbrot's fractals on shared‐ and distributed‐memory platforms. We demonstrate the competitiveness of the Chapel implementation against its OpenMP and MPI+X counterparts.…”

“…Getting a feedback on the suitability and limits of conventional and new languages targeting exascale supercomputers becomes critical. In the same spirit as References 10‐15, this paper aims at providing a useful data point to help practitioners gauge the difficult question of whether or not to invest time and effort into learning and using the Chapel programming language.…”

“…The DistBag data structure has been redefined in two different ways: (1) we propose a new scheduling policy of its elements as well as a new synchronization mechanism using non‐blocking split deques, and (2) we redefine the underlying WS mechanism. The revised data structure is referred to as DistBag‐DFS 20 . Experimental comparisons between DistBag and DistBag‐DFS are done in Section 4.4.1.…”

“…Algorithm 1 shows the pseudo‐code of our Chapel‐based Productivity‐ and Performance‐aware Parallel Distributed DFS algorithm, namely P3D‐DFS 20 . The algorithm starts by creating the root node, using the problem‐specific data/parameters (line 5).…”

Possibilities of Use for Fractal Techniques as Parameters of Graphic Analysis

“…Nowadays, the most popular methods for parallel programming are implemented using the message-passing interface and open multi-processing (Open MP), with an example of mixing of these two being Chapel, which uses a hybrid system in a distributed-memory environment [24]. A variant of the generation of the Mandelbrot set using the two technologies is presented in [25], with this involving the use of an AMD Ryzen™ 5 2500U Quad-Core system analyzed in detail at the processor level.…”