Trusted High-Performance Computing in the Classroom

Abstract

A well-designed high-performance computing (HPC) course not only presents theoretical parallelism concepts but also includes practical work on parallel systems. Today's machine models are diverse and as a consequence multiple programming models exist. The challenge for HPC course lecturers is to decide what to include and what to exclude, respectively. We have experience in teaching HPC in a multi-paradigm style. The practical course parts include message-passing programming using MPI, directive-based shared memory programming using OpenMP, partitioned global address space based programming using Chapel, and domain-specific programming using a high-level framework. If these models are taught in an isolated mode, students would have problems in assessing the strengths and weaknesses of the approaches presented. We propose a projectbased approach which introduces a specific problem to be solved (in our case a stencil computation) and asks for solutions using the programming approaches introduced. Our course has been successfully taught several times but a major problem has always been checking the individual student solutions, especially to decide which performance results reported one can trust. In order to overcome these deficiencies, we have built a pedagogical tool which enhances the trust in students' work. In the paper we present the infrastructure and tools that make student experiments easily reproducible by lecturers. We introduce a taxonomy for general benchmark experiments, describe the distributed architecture of our development and analysis environment, and, as a case study, discuss performance experiments when solving a stencil problem in multiple programming models.

Mapeamento Sistemático do Ensino Teórico e Prático de Programação Paralela

“…[ Burkhart et al 2014] mostram uma abordagem baseada em projetos para resolver problemas específicos. Neste artigo, duas abordagens são definidas, autônoma e integrada.…”

Scientific Computing, High-Performance Computing and Data Science in Higher Education

“…A number of academic programs that aim to address this issue have emerged at universities across the world (a few examples are [3,4]). Some of these grew out of the training efforts of High Performance Computing (HPC) centres and organizations (e.g.…”

Teaching Parallel Programming in Containers: Virtualization of a Heterogeneous Local Infrastructure

“…The last approach included advanced topics such as MapReduce and the General Purpose Computing on Graphics Processing Units (GPGPU) 4 . Burkhart, Guerrera and Maffia (2014) show a project-based approach for solving specific problems. In their paper, two methods are defined, standalone and integrated.…”

“…Chart 10 presents teaching parallel programming strategies for practical laboratory classes. The vast majority of the papers address the use of OpenMP and MPI for code parallelization (JOINER et al, 2006;MAROWKA, 2008;IVICA;RILEY;SHUBERT, 2009;ARROYO, 2013;SHAFI et al, 2014;BURKHART;GUERRERA;MAFFIA, 2014;SHAMSI;DURRANI;KAFI, 2015;SCHMIDT, 2015;CUENCA;GIMÉNEZ, 2016;EIJKHOUT, 2016;MOORE;DUNLOP, 2016). For the execution of the programs, the largest of the works had a cluster infrastructure (LIU, 2016;MOORE;DUNLOP, 2016;BALDWIN et al, 2016).…”

“…Still, many other works use Proprietary Software (PS) for the practice of parallel programming, so few infrastructure details are shown (JOINER et al, 2006;IVICA;RILEY;SHUBERT, 2009;ARROYO, 2013;SHOOP, 2013;FERNER;WILKIN-SON;HEATH, 2013;FERNER;WILKINSON;HEATH, 2014;BURKHART;GUERRERA;MAFFIA, 2014;UL-ABDIN;SVENSSON, 2015;SCHMIDT, 2015;BALDWIN et al, 2016;GARDNER;B., 2017;GROSSMAN et al, 2017a). Also, we highlight that the Java language is constantly mentioned in papers Although CUDA is not an open-source tool, it is still preferred for teaching due to its low complexity imposed by OpenCL, such as discovering platforms and devices, setting the queues and, compiling the kernel (ADAMS, 2021).…”