Estudi de comparació de rendiment de Benchmark multicore i multithreading amb OpenMP 2.5 i OpenMP 3.0

Estudi comparatiu amb benchmark del rendiment en dues plataformes multicore multithreading de diferents modalitats de paral·lelització de multiplicacions de matrius de nombres enters i de nombres en coma flotant mitjançant el model de memòria compartida OpenMP versió 2.5 i OpenMP versió 3.0.

A User-Centric Perspective on Parallel Programming with Focus on OpenMP

The process of developing software that takes advantage of multiple processors is commonly referred to as parallel programming. For various reasons, this process is much harder than the sequential case. For decades, parallel programming has been a problem for a small niche only: engineers working on parallelizing mostly numerical applications in High Performance Computing. This has changed with the advent of multi-core processors in mainstream computer architectures. Parallel programming in our days becomes a problem for a much larger group of developers. The main objective of this thesis was to find ways to make parallel programming easier for them. Different aims were identified in order to reach the objective: research the state of the art of parallel programming today, improve the education of software developers about the topic, and provide programmers with powerful abstractions to make their work easier. To reach these aims, several key steps were taken. To start with, a survey was conducted among parallel programmers to find out about the state of the art. More than 250 people participated, yielding results about the parallel programming systems and languages in use, as well as about common problems with these systems. Furthermore, a study was conducted in university classes on parallel programming. It resulted in a list of frequently made mistakes that were analyzed and used to create a programmers' checklist to avoid them in the future. For programmers' education, an online resource was setup to collect experiences and knowledge in the field of parallel programming - called the Parawiki. Another key step in this direction was the creation of the Thinking Parallel weblog, where more than 50.000 readers to date have read essays on the topic. For the third aim (powerful abstractions), it was decided to concentrate on one parallel programming system: OpenMP. Its ease of use and high level of abstraction were the most important reasons for this decision. Two different research directions were pursued. The first one resulted in a parallel library called AthenaMP. It contains so-called generic components, derived from design patterns for parallel programming. These include functionality to enhance the locks provided by OpenMP, to perform operations on large amounts of data (data-parallel programming), and to enable the implementation of irregular algorithms using task pools. AthenaMP itself serves a triple role: the components are well-documented and can be used directly in programs, it enables developers to study the source code and learn from it, and it is possible for compiler writers to use it as a testing ground for their OpenMP compilers. The second research direction was targeted at changing the OpenMP specification to make the system more powerful. The main contributions here were a proposal to enable thread-cancellation and a proposal to avoid busy waiting. Both were implemented in a research compiler, shown to be useful in example applications, and proposed to the OpenMP Language Committee.

OpenMP task scheduling strategies to mitigate hardware variability in tightly-coupled shared memory clusters

In questa tesi sono stati apportati due importanti contributi nel campo degli acceleratori embedded many-core. Abbiamo implementato un runtime OpenMP ottimizzato per la gestione del tasking model per sistemi a processori strettamente accoppiati in cluster e poi interconnessi attraverso una network on chip. Ci siamo focalizzati sulla loro scalabilità e sul supporto di task di granularità fine, come è tipico nelle applicazioni embedded. Il secondo contributo di questa tesi è stata proporre una estensione del runtime di OpenMP che cerca di prevedere la manifestazione di errori dati da fenomeni di variability tramite una schedulazione efficiente del carico di lavoro.

Experiences with OpenMP in tmLQCD

An overview is given of the lessons learned from the introduction of multi-threading using OpenMP in tmLQCD. In particular, programming style, performance measurements, cache misses, scaling, thread distribution for hybrid codes, race conditions, the overlapping of communication and computation and the measurement and reduction of certain overheads are discussed. Performance measurements and sampling profiles are given for different implementations of the hopping matrix computational kernel.

Diseño e implementación de funcionalidades OpenMP, basadas en modelos de concurrencia desarrollados en lenguaje Go

A raíz de la aparición de los procesadores dotados de varios “cores”, la programación paralela, un concepto que, por otra parte no era nada nuevo y se conocía desde hace décadas, sufrió un nuevo impulso, pues se creía que se podía superar el techo tecnológico que había estado limitando el rendimiento de esta programación durante años. Este impulso se ha ido manteniendo hasta la actualidad, movido por la necesidad de sistemas cada vez más potentes y gracias al abaratamiento de los costes de fabricación. Esta tendencia ha motivado la aparición de nuevo software y lenguajes con componentes orientados precisamente al campo de la programación paralela. Este es el caso del lenguaje Go, desarrollado por Google y lanzado en 2009. Este lenguaje se basa en modelos de concurrencia que lo hacen muy adecuados para abordar desarrollos de naturaleza paralela. Sin embargo, la programación paralela es un campo complejo y heterogéneo, y los programadores son reticentes a utilizar herramientas nuevas, en beneficio de aquellas que ya conocen y les son familiares. Un buen ejemplo son aquellas implementaciones de lenguajes conocidos, pero orientadas a programación paralela, y que siguen las directrices de un estándar ampliamente reconocido y aceptado. Este es el caso del estándar OpenMP, un Interfaz de Programación de Aplicaciones (API) flexible, portable y escalable, orientado a la programación paralela multiproceso en arquitecturas multi-core o multinucleo. Dicho estándar posee actualmente implementaciones en los lenguajes C, C++ y Fortran. Este proyecto nace como un intento de aunar ambos conceptos: un lenguaje emergente con interesantes posibilidades en el campo de la programación paralela, y un estándar reputado y ampliamente extendido, con el que los programadores se encuentran familiarizados. El objetivo principal es el desarrollo de un conjunto de librerías del sistema (que engloben directivas de compilación o pragmas, librerías de ejecución y variables de entorno), soportadas por las características y los modelos de concurrencia propios de Go; y que añadan funcionalidades propias del estándar OpenMP. La idea es añadir funcionalidades que permitan programar en lenguaje Go utilizando la sintaxis que OpenMP proporciona para otros lenguajes, como Fortan y C/C++ (concretamente, similar a esta última), y, de esta forma, dotar al usuario de Go de herramientas para programar estructuras de procesamiento paralelo de forma sencilla y transparente, de la misma manera que lo haría utilizando C/C++.---ABSTRACT---As a result of the appearance of processors equipped with multiple "cores ", parallel programming, a concept which, moreover, it was not new and it was known for decades, suffered a new impulse, because it was believed they could overcome the technological ceiling had been limiting the performance of this program for years. This impulse has been maintained until today, driven by the need for ever more powerful systems and thanks to the decrease in manufacturing costs. This trend has led to the emergence of new software and languages with components guided specifically to the field of parallel programming. This is the case of Go language, developed by Google and released in 2009. This language is based on concurrency models that make it well suited to tackle developments in parallel nature. However, parallel programming is a complex and heterogeneous field, and developers are reluctant to use new tools to benefit those who already know and are familiar. A good example are those implementations from well-known languages, but parallel programming oriented, and witch follow the guidelines of a standard widely recognized and accepted. This is the case of the OpenMP standard, an application programming interface (API), flexible, portable and scalable, parallel programming oriented, and designed for multi-core architectures. This standard currently has implementations in C, C ++ and Fortran. This project was born as an attempt to combine two concepts: an emerging language, with interesting possibilities in the field of parallel programming, and a reputed and widespread standard, with which programmers are familiar with. The main objective is to develop a set of system libraries (which includes compiler directives or pragmas, runtime libraries and environment variables), supported by the characteristics and concurrency patterns of Go; and that add custom features from the OpenMP standard. The idea is to add features that allow programming in Go language using the syntax OpenMP provides for other languages, like Fortran and C / C ++ (specifically, similar to the latter ), and, in this way, provide Go users with tools for programming parallel structures easily and, in the same way they would using C / C ++.

Theory manual to OctVCE – a cartesian cell CFD code with special application to blast wave problems, Department of Mechanical Engineering Report 2007/12

OctVCE is a cartesian cell CFD code produced especially for numerical simulations of shock and blast wave interactions with complex geometries, in particular, from explosions. Virtual Cell Embedding (VCE) was chosen as its cartesian cell kernel for its simplicity and sufficiency for practical engineering design problems. The code uses a finite-volume formulation of the unsteady Euler equations with a second order explicit Runge-Kutta Godonov (MUSCL) scheme. Gradients are calculated using a least-squares method with a minmod limiter. Flux solvers used are AUSM, AUSMDV and EFM. No fluid-structure coupling or chemical reactions are allowed, but gas models can be perfect gas and JWL or JWLB for the explosive products. This report also describes the code’s ‘octree’ mesh adaptive capability and point-inclusion query procedures for the VCE geometry engine. Finally, some space will also be devoted to describing code parallelization using the shared-memory OpenMP paradigm. The user manual to the code is to be found in the companion report 2007/13.

A framework for the development of parallel and distributed real-time embedded systems

Embedded real-time applications increasingly present high computation requirements, which need to be completed within specific deadlines, but that present highly variable patterns, depending on the set of data available in a determined instant. The current trend to provide parallel processing in the embedded domain allows providing higher processing power; however, it does not address the variability in the processing pattern. Dimensioning each device for its worst-case scenario implies lower average utilization, and increased available, but unusable, processing in the overall system. A solution for this problem is to extend the parallel execution of the applications, allowing networked nodes to distribute the workload, on peak situations, to neighbour nodes. In this context, this report proposes a framework to develop parallel and distributed real-time embedded applications, transparently using OpenMP and Message Passing Interface (MPI), within a programming model based on OpenMP. The technical report also devises an integrated timing model, which enables the structured reasoning on the timing behaviour of these hybrid architectures.

The extended unsymmetric frontal solution for multiple-point constraints

The purpose of this paper is to discuss the linear solution of equality constrained problems by using the Frontal solution method without explicit assembling. Design/methodology/approach - Re-written frontal solution method with a priori pivot and front sequence. OpenMP parallelization, nearly linear (in elimination and substitution) up to 40 threads. Constraints enforced at the local assembling stage. Findings - When compared with both standard sparse solvers and classical frontal implementations, memory requirements and code size are significantly reduced. Research limitations/implications - Large, non-linear problems with constraints typically make use of the Newton method with Lagrange multipliers. In the context of the solution of problems with large number of constraints, the matrix transformation methods (MTM) are often more cost-effective. The paper presents a complete solution, with topological ordering, for this problem. Practical implications - A complete software package in Fortran 2003 is described. Examples of clique-based problems are shown with large systems solved in core. Social implications - More realistic non-linear problems can be solved with this Frontal code at the core of the Newton method. Originality/value - Use of topological ordering of constraints. A-priori pivot and front sequences. No need for symbolic assembling. Constraints treated at the core of the Frontal solver. Use of OpenMP in the main Frontal loop, now quantified. Availability of Software.

Análise de imagens tomográficas: visualização e paralelização de processamento

Dissertação de Mestrado em Engenharia Informática

Programação paralela estruturada aplicada a um problema de reconstrução de imagem

A Digital Breast Tomosynthesis (DBT) é uma técnica que permite obter imagens mamárias 3D de alta qualidade, que só podem ser obtidas através de métodos de re-construção. Os métodos de reconstrução mais rápidos são os iterativos, sendo no en-tanto computacionalmente exigentes, necessitando de sofrer muitas optimizações. Exis-tem optimizações que usam computação paralela através da implementação em GPUs usando CUDA. Como é sabido, o desenvolvimento de programas eficientes que usam GPUs é ainda uma tarefa demorada, dado que os modelos de programação disponíveis são de baixo nível, e a portabilidade do código para outras arquitecturas não é imedia-ta. É uma mais valia poder criar programas paralelos de forma rápida, com possibili-dade de serem usados em diferentes arquitecturas, sem exigir muitos conhecimentos sobre a arquitectura subjacente e sobre os modelos de programação de baixo nível. Para resolver este problema, propomos a utilização de soluções existentes que reduzam o esforço de paralelização, permitindo a sua portabilidade, garantindo ao mesmo tempo um desempenho aceitável. Para tal, vamos utilizar um framework (FastFlow) com suporte para Algorithmic Skeletons, que tiram partido da programação paralela estruturada, capturando esquemas/padrões recorrentes que são comuns na programação paralela. O trabalho realizado centrou-se na paralelização de uma das fases de reconstru-ção da imagem 3D – geração da matriz de sistema – que é uma das mais demoradas do processo de reconstrução; esse trabalho incluiu um método de ordenação modificado em relação ao existente. Foram realizadas diferentes implementações em CPU e GPU (usando OpenMP, CUDA e FastFlow) o que permitiu comparar estes ambientes de programação em termos de facilidade de desenvolvimento e eficiência da solução. A comparação feita permite concluir que o desempenho das soluções baseadas no FastFlow não é muito diferente das tradicionais o que sugere que ferramentas deste tipo podem simplificar e agilizar a implementação de um algoritmos na área de recons-trução de imagens 3D, mantendo um bom desempenho.

Análisis de rendimiento de aplicaciones paralelas de memoria compartida: problema N-body

Este trabajo analiza el rendimiento de cuatro nodos de cómputo multiprocesador de memoria compartida para resolver el problema N-body. Se paraleliza el algoritmo serie, y se codifica usando el lenguaje C extendido con OpenMP. El resultado son dos variantes que obedecen a dos criterios de optimización diferentes: minimizar los requisitos de memoria y minimizar el volumen de cómputo. Posteriormente, se realiza un proceso de análisis de las prestaciones del programa sobre los nodos de cómputo. Se modela el rendimiento de las variantes secuenciales y paralelas de la aplicación, y de los nodos de cómputo; se instrumentan y ejecutan los programas para obtener resultados en forma de varias métricas; finalmente se muestran e interpretan los resultados, proporcionando claves que explican ineficiencias y cuellos de botella en el rendimiento y posibles líneas de mejora. La experiencia de este estudio concreto ha permitido esbozar una incipiente metodología de análisis de rendimiento, identificación de problemas y sintonización de algoritmos a nodos de cómputo multiprocesador de memoria compartida.

Performance analysis and tuning in multicore environments

Performance analysis is the task of monitor the behavior of a program execution. The main goal is to find out the possible adjustments that might be done in order improve the performance. To be able to get that improvement it is necessary to find the different causes of overhead. Nowadays we are already in the multicore era, but there is a gap between the level of development of the two main divisions of multicore technology (hardware and software). When we talk about multicore we are also speaking of shared memory systems, on this master thesis we talk about the issues involved on the performance analysis and tuning of applications running specifically in a shared Memory system. We move one step ahead to take the performance analysis to another level by analyzing the applications structure and patterns. We also present some tools specifically addressed to the performance analysis of OpenMP multithread application. At the end we present the results of some experiments performed with a set of OpenMP scientific application.

Alineamiento de secuencias genéticas en procesadores multicore

Este trabajo analiza el rendimiento del algoritmo de alineamiento de secuencias conocido como Needleman-Wunsch, sobre 3 sistemas de cómputo multiprocesador diferentes. Se analiza y se codifica el algoritmo serie usando el lenguaje de programación C y se plantean una serie de optimizaciones con la finalidad de minimizar el volumen y el tiempo de cómputo. Posteriormente, se realiza un análisis de las prestaciones del programa sobre los diferentes sistemas de cómputo. En la segunda parte del trabajo, se paraleliza el algoritmo serie y se codifica ayudándonos de OpenMP. El resultado son dos variantes del programa que difieren en la relación entre la cantidad de cómputo y la de comunicación. En la primera variante, la comunicación entre procesadores es poco frecuente y se realiza tras largos periodos de ejecución (granularidad gruesa). En cambio, en la segunda variante las tareas individuales son relativamente pequeñas en término de tiempo de ejecución y la comunicación entre los procesadores es frecuente (granularidad fina). Ambas variantes se ejecutan y analizan en arquitecturas multicore que explotan el paralelismo a nivel de thread. Los resultados obtenidos muestran la importancia de entender y saber analizar el efecto del multicore y multithreading en el rendimiento.

Factores de rendimiento en entornos multicore

Este documento refleja el estudio de investigación para la detección de factores que afectan al rendimiento en entornos multicore. Debido a la gran diversidad de arquitecturas multicore se ha definido un marco de trabajo, que consiste en la adopción de una arquitectura específica, un modelo de programación basado en paralelismo de datos, y aplicaciones del tipo Single Program Multiple Data. Una vez definido el marco de trabajo, se han evaluado los factores de rendimiento con especial atención al modelo de programación. Por este motivo, se ha analizado la librería de threads y la API OpenMP para detectar aquellas funciones sensibles de ser sintonizadas al permitir un comportamiento adaptativo de la aplicación al entorno, y que dependiendo de su adecuada utilización han de mejorar el rendimiento de la aplicación.

Factores de rendimiento en aplicaciones híbridas

En el entorno actual, diversas ramas de las ciencias, tienen la necesidad de auxiliarse de la computación de altas prestaciones para la obtención de resultados a relativamente corto plazo. Ello es debido fundamentalmente, al alto volumen de información que necesita ser procesada y también al costo computacional que demandan dichos cálculos. El beneficio al realizar este procesamiento de manera distribuida y paralela, logra acortar los tiempos de espera en la obtención de los resultados y de esta forma posibilita una toma decisiones con mayor anticipación. Para soportar ello, existen fundamentalmente dos modelos de programación ampliamente extendidos: el modelo de paso de mensajes a través de librerías basadas en el estándar MPI, y el de memoria compartida con la utilización de OpenMP. Las aplicaciones híbridas son aquellas que combinan ambos modelos con el fin de aprovechar en cada caso, las potencialidades específicas del paralelismo en cada uno. Lamentablemente, la práctica ha demostrado que la utilización de esta combinación de modelos, no garantiza necesariamente una mejoría en el comportamiento de las aplicaciones. Por lo tanto, un análisis de los factores que influyen en el rendimiento de las mismas, nos beneficiaría a la hora de implementarlas pero también, sería un primer paso con el fin de llegar a predecir su comportamiento. Adicionalmente, supondría una vía para determinar que parámetros de la aplicación modificar con el fin de mejorar su rendimiento. En el trabajo actual nos proponemos definir una metodología para la identificación de factores de rendimiento en aplicaciones híbridas y en congruencia, la identificación de algunos factores que influyen en el rendimiento de las mismas.