The 14th Super Computing Camp – Cartagena de Indias - 2023
Robinson Rivas – OpenMP Tutorial
OpenMP Tutorial
Robinson Rivas Suárez
Universidad Central de Venezuela
INTRODUCTION TO OpenMP
Prof. Robinson Rivas-Suárez
UNIVERSIDAD CENTRAL DE VENEZUELA
Prepara tus exámenes
Prepara tus exámenes y mejora tus resultados gracias a la gran cantidad de recursos disponibles en Docsity
Consigue puntos base para descargar
Gana puntos ayudando a otros estudiantes o consíguelos activando un Plan Premium
Orientación Universidad
Prepara tus exámenes
Prepara tus exámenes y mejora tus resultados gracias a la gran cantidad de recursos disponibles en Docsity
Busca documentos
Prepara tus exámenes con los documentos que comparten otros estudiantes como tú en Docsity
Busca tu universidad
Encuentra los documentos específicos para los exámenes de tu universidad
Video Cursos
Estudia con lecciones y exámenes resueltos basados en los programas académicos de las mejores universidades
Quiz
Responde a preguntas de exámenes reales y pon a prueba tu preparación
Consigue puntos base para descargar
Gana puntos ayudando a otros estudiantes o consíguelos activando un Plan Premium
Comunidad
Pregúntale a la comunidad
Pide ayuda a la comunidad y resuelve tus dudas de estudio
Ebooks gratuitos
¡Nuestros e-books salva-estudiantes!
Descarga nuestras guías gratuitas sobre técnicas de estudio, métodos para controlar la ansiedad y consejos para la tesis preparadas por los tutores de Docsity
tutorial de open mp con lenguaje c, Diapositivas de Computación Distribuida
este es un tutorial en el que se encuentra conceptos relacionados a open mp
Tipo: Diapositivas
2022/2023
1 / 89
Esta página no es visible en la vista previa
¡No te pierdas las partes importantes!
Documentos relacionados
Vista previa parcial del texto
¡Descarga tutorial de open mp con lenguaje c y más Diapositivas en PDF de Computación Distribuida solo en Docsity!
OpenMP Tutorial
Robinson Rivas Suárez Universidad Central de Venezuela INTRODUCTION TO OpenMP Prof. Robinson Rivas-Suárez UNIVERSIDAD CENTRAL DE VENEZUELA
BsC in Computer Science (Maracaibo, Venezuela) Master in Computer Science (Caracas, Venezuela) Network Engineer (OIC, Japan)
Robinson
- That’s really necessary?
- Why to think about shared memory
- OpenMP and new models of programming
- Everyday platforms
- New trends
Overview
- Often, we professors teach our students that algorithms must be designed, planned and then programmed as much abstract as possible, i.e. only thinking on the problem to be solved rather than the computer that actually solves it.
- This way of thinking is so-called abstract design
Back to school
People who are more than casually interested in computers should have at least some idea of what the underlying hardware is like. Otherwise the programs they write will be pretty weird ― Donald Knuth, The Art of Computer Programming, Volume 1: Fundamental Algorithms
An opinion
- More than two processors using the same memory system, at least parts of the same memory system
- It happens when there are many cores per processor. It was rare in the very early years, but it is really common today
Shared Memory
- Think in a standard smartphone today. Tipically, ARM processors have 4 cores (even more)
- So, computers we have in our pockets needs to be programmed in an efficient yet reliable way
- HPC in terms of shared memory is not just for top 500 list supercomputers … it matters everywhere!!
Shared Memory
- OpenMP is an standard set of instructions available for the most important HPC programming languages
- It was originally proposed as an open specification in 1996
- First official draft of the OpenMP consortium was released in 1997
OpenMP
OpenMP is managed by the nonprofit technology consortium OpenMP Architecture Review Board (or OpenMP ARB ), jointly defined by a group of major computer hardware and software vendors, including AMD, IBM, Intel, Cray, HP, Fujitsu, Nvidia, NEC, Red Hat, Texas Instruments, Oracle Corporation, and more. source: Wikipedia
- Portability: code runs in any platform, since thread creation is encapsulated inside the actual O.S.
- Programmer has not to control the thread behavior. It releases him/her from the heaviest part of thread management
- Because is intended to work in shared memory, OpenMP does not deal with message passing. It eliminates the most common source of errors in parallel programming
Advantages of OpenMP
- Higly scalable: performance usually goes better when code is executed in more cores
- By design, OpenMP permits to keep the sequential code as it is. It allows to use the programs in different environments without need to recompile (on the same architecture)
Advantages of OpenMP
- Debugging process is not that easy. OpenMP doesn’t have an efficient native error handling mechanism
- By now, internal details of threads are hidden for programmer. It is of course an advantage for readyness and portability, but doesn’t allow programmers to do fine control to improve performance on specific architectures
OpenMP issues
- Many tools has being designed to help with this. But are not part of the standard itself.
- Some examples are profilers like Valgrind, VAMPIR, SCORE-P, etc
- BSC (Barcelona-Spain)has developed a few useful tools to help better understanding of OpenMP parallel code
OpenMP issues
- OpenMP doesn’t have memory-related instructions. You can’t deal with cache, memory layers or new architecture models. Everything is seem as plain shared memory (*)
- Startup time for threads is high compared to sequential time for execution. So, if the problem is really small, OpenMP parallel version could be worst than sequential version (*) we’ll be back about this later
OpenMP issues
OpenMP is designed as a series of compiler directives plus an small suite of library functions. So:
- OpenMP is NOT a library nor a set of pre-compiled code
- OpenMP is NOT designed to run well in a distributed-memory environment