Short Courses of the 21st Regional School of High Performance Computing from Southern Brazil
Keywords:
High PerformanceSynopsis
This book presents textual versions, in the form of book chapters, of six short courses accepted and presented at the 21st Regional School of High-Performance Computing from Southern Brazil (ERAD/RS). The mini-courses deal with technical aspects related to the following themes: parallel programming for different architectures, tools for optimizing parallel application performance, and how to conduct reproducible experiments.
In the first chapter of this book, "Unraveling the Use of Hardware Counters to Optimize Artificial Intelligence Applications", the authors explain how hardware counters from Intel Xeon Cascade Lake and NEC SX-Aurora TSUBASA architectures help in the performance analysis of AI applications. In the second chapter, "Profiling and Tuning OpenACC Code", the authors present optimization techniques that can be used to increase OpenACC-based applications, which are suitable for both multicore and GPU architectures. In the third chapter, "Are you root? Reproducible Experiments in User Space", the authors present techniques to create environments for reproducible experiments using the Spack package manager and two containerization tools: Docker and Singularity. In the fourth chapter, "Beyond simply: #pragma omp parallel for", the authors address some new and less widespread OpenMP features, going further than the commonly seen introductory OpenMP courses. In the fifth chapter, "Private Cloud Computing Environment for Testing and Developing Parallel Programs", the authors offer an introduction to the deployment of private clouds, demonstrating the benefits of that for developing and testing parallel applications. Finally, in the sixth chapter, "Development of Task-Based Applications with OpenMP Tasks", the authors deal with the task-oriented paradigm and how to build parallel applications with tasks in OpenMP.
Chapters
-
1. Desvendando o Uso de Contadores de Hardware para Otimizar Aplicações de Inteligência Artificial
-
2. Otimização de Programas Paralelos com uso do OpenACC
-
3. Are you root? Experimentos Reprodutíveis em Espaço de Usuário
-
4. Além de Simplesmente: #pragma omp parallel for
-
5. Ambiente de Nuvem Computacional Privada para Teste e Desenvolvimento de Programas Paralelos
-
6. Desenvolvimento de Aplicações Baseadas em Tarefas com OpenMP Tasks
Downloads
