An approach to task-based parallel programming for undergraduate students

This paper presents the description of a compulsory parallel programming course in the bachelor degree in Informatics Engineering at the Barcelona School of Informatics, Universitat Politècnica de Catalunya UPC-BarcelonaTech. The main focus of the course is on the shared-memory programming paradigm, which facilitates the presentation of fundamental aspects and notions of parallel computing. Unlike the “traditional” loop-based approach, which is the focus of parallel programming courses in other universities, this course presents the parallel programming concepts using a task-based approach. Tasking allows students to explore a broader set of parallel decomposition strategies, including linear, iterative and recursive strategies, and their implementation using the current version of OpenMP (OpenMP 4.5), which offers mechanisms (pragmas and intrinsic functions) to easily map these strategies into parallel programs. Simple models to understand the benefits of a task decomposition and the trade-offs introduced by different kinds of overheads are included in the course, together with the use of tools that allow an easy exploration of different task decomposition strategies and their potential parallelism (Tareador) and instrumentation and analysis of task parallel executions on real machines (Extrae and Paraver).This work has been supported by the grant SEV-2015-0493 of the Severo Ochoa Program, awarded by the Spanish Gov- ernment, by the Spanish Ministry of Science and Innovation (contract TIN2015-65316-P) and by Generalitat de Catalunya (contracts 2014-MOOC-00057 and 2014-SGR-1051). We also thank the anonymous reviewers and editor for their comments during the review process, other professors that have been in- volved in the implementation of the course and Paul Carpenter at BSC for his corrections and suggestions to improve the text.Postprint (published version