Pasar al contenido principal

Resultados relacionados

Resultados totales (Incluyendo duplicados): 36
Encontrada(s) 1 página(s)

RUA. Repositorio Institucional de la Universidad de Alicante
Publicación Otros (Other)

Estrategias para el diseño de laboratorios orientados al aprendizaje continuo

  • Fernández, Agustín
  • Llosa, Josep
  • Sánchez-Carracedo, Fermín
Trabajar regularmente es beneficioso para el aprendizaje. Una de las herramientas que existen para fomentar el trabajo regular son las clases de laboratorio. En este artículo presentamos una estrategia para diseñar prácticas orientadas al aprendizaje continuo en asignaturas en las que la misión fundamental del laboratorio es reforzar los conocimientos adquiridos en las clases teóricas. Los puntos básicos de esta estrategia son: (a) las sesiones de laboratorio han de tener una periodicidad, como mínimo, semanal; (b) deben estar perfectamente sincronizadas con la teoría y (c) el alumno debe estar motivado para realizarlas. Los dos primeros puntos están relacionados con el diseño de la asignatura. El punto (c), además, está directamente relacionado con el método de evaluación y con la actitud del profesor durante las sesiones de laboratorio. Con la metodología propuesta, el seguimiento de las prácticas no requiere un gran esfuerzo por parte del profesor una vez han sido diseñadas. Los resultados obtenidos en las asignaturas en las que se ha aplicado la metodología descrita indican que existe una alta correlación entre las notas de laboratorio y las de teoría/problemas, lo que muestra que el método propuesto es beneficioso para el aprendizaje.
RUA. Repositorio Institucional de la Universidad de Alicante
Publicación Otros (Other)

Evaluación formativa con feedback rápido usando mandos interactivos

  • Álvarez, Carlos
  • Llosa, Josep
Numerosos estudios han demostrado que el uso del feedback como herramienta docente resulta beneficioso para el aprendizaje. Para que este feedback sea útil debe ser rápido, es decir, llegar al alumno poco después de haber realizado su tarea.En este artículo presentamos resultados de una prueba piloto en la que se ha usado una herramienta para proporcionar feedback basada en transparencias y mandos interactivos. Con esta herramienta los alumnos contestan todos a la vez a preguntas realizadas en clase. La herramienta recopila los datos y ofrece estadísticas inmediatas sobre las respuestas de los alumnos. Esto permite al profesor detectar los errores comunes y enfatizar aquellos aspectos más deficitarios para los alumnos, proporcionando feedback inmediato. En este trabajo explicamos las características de la herramienta, el entorno de trabajo, los beneficios aportados, las deficiencias observadas y las incidencias producidas. Finalmente proporcionamos una evaluación estadística de la prueba, tanto a través de encuestas a los estudiantes como analizando sus resultados académicos.
RUA. Repositorio Institucional de la Universidad de Alicante
Publicación Otros (Other)

Trabajo no presencial en colaboración: triple realimentación a coste razonable

  • Jiménez-González, Daniel
  • López, David
  • Álvarez, Carlos
  • Alonso, Javier
Ofrecer una realimentación (feedback) rápida a los estudiantes es un componente fundamental en el proceso de aprendizaje. En este artículo presentamos un método que permite ofrecer una triple realimentación de una manera rápida y con un coste razonable. Los estudiantes se agrupan en equipos para resolver problemas en colaboración (primera realimentación, de sus compañeros). Las soluciones resultantes son valoradas por el profesor, que las devuelve comentadas al grupo (segunda realimentación). Con las soluciones de todos los equipos, el profesor puede deducir que conceptos necesitan ser enfatizados en las próximas clases (tercera realimentación). En nuestra propuesta, cada estudiante es responsable de la discusión de algunos problemas, así que desarrolla capacidades como el trabajo en equipo o el liderazgo. Además, la cantidad de trabajo para el profesor es razonable ya que sólo corrige las soluciones consensuadas por cada equipo. Esta metodología ha sido probada en una asignatura de primer curso de Ingeniería Informática usando la plataforma de aprendizaje Moodle. Hemos recogido datos durante cuatro cuatrimestres para evaluar la efectividad de la propuesta. Los resultados muestran un mayor número de aprobados y mejores notas entre los estudiantes que siguen esta metodología, la cual ha tenido una buena aceptación por parte de los estudiantes.
RUA. Repositorio Institucional de la Universidad de Alicante
Publicación Otros (Other)

Estrategia para el diseño de asignaturas en el EEES

  • Sánchez-Carracedo, Fermín
  • Garcia, Jordi
  • Sancho, Maria-Ribera
Los planes de estudios del EEES deben diseñarse a partir de competencias profesionales, por lo que será necesario establecer una relación directa entre los objetivos formativos de las asignaturas y las competencias de la titulación. Los créditos ECTS, que cuantifican el esfuerzo requerido para el aprendizaje, miden el tiempo dedicado por el alumno, a diferencia de los actuales créditos que miden el tiempo presencial del profesor. Esto es consecuencia del cambio de modelo educativo, que pasa de estar centrado en la enseñanza (del profesor) a estar centrado en el aprendizaje (del alumno), y tiene implicaciones directas en las metodologías docentes. En este artículo se propone una estrategia para diseñar las asignaturas pertenecientes a los nuevos planes de estudios del EEES en seis etapas. Inicialmente, la asignatura tiene preasignados un número de créditos ECTS y un conjunto de competencias a las que debe contribuir con el nivel de competencia que le ha sido asignado por el plan de estudios. Al final del proceso, se obtiene una asignatura basada en objetivos formativos cuyas estrategias docentes y modelo de evaluación están adaptados al EEES.
RUA. Repositorio Institucional de la Universidad de Alicante
Publicación Otros (Other)

Montando el puzzle: visión global de un sistema operativo

  • Verdú, Javier
  • López, David
  • Pajuelo, Alex
A menudo, algunos estudiantes que han demostrado ser perfectamente capaces de resolver problemas de cualquier tema de un curso, fracasan ante problemas que combinan conocimientos de diversos temas. Esto se debe a que existe una tendencia a dividir el conocimiento en paquetes pequeños y no relacionados entre sí. Los alumnos estudian estos paquetes y los aprenden, pero no obtienen una visión de conjunto: conociendo cada árbol, roca y animal del bosque, no entienden las interacciones que lo convierten en un ecosistema. En este trabajo presentamos una metodología que permite obtener una visión de conjunto de una asignatura usando técnicas de aprendizaje activo. Aunque la experiencia ha sido realizada en una asignatura de Sistemas Operativos, pensamos que la metodología es adaptable a otros cursos.
RUA. Repositorio Institucional de la Universidad de Alicante
Publicación Otros (Other)

Cómo formar Ingenieros en Informática en la competencia Sostenibilidad y Compromiso Social

  • Franquesa, David
  • Cruz, Josep-Llorenç
  • Álvarez, Carlos
  • Sánchez-Carracedo, Fermín
  • Fernández, Agustín
  • López, David
En los nuevos planes de estudios hay que desarrollar competencias que resultan novedosas: prácticamente no han sido trabajadas con anterioridad. Cómo enseñarlas y cómo evaluarlas es una preocupación para los diseñadores de los nuevos planes. Este artículo analiza la competencia “Sostenibilidad y Compromiso Social”, explicando técnicas para desarrollarla tanto a nivel de comprensión como al de aplicación, según la taxonomía de Bloom, y analiza las condiciones que deben darse en un centro para poder implementar estas técnicas en las asignaturas de su plan de estudios.
RUA. Repositorio Institucional de la Universidad de Alicante
Publicación Otros (Other)

Aprendizaje cooperativo en cursos multidisciplinares

  • Álvarez, Carlos
  • López, David
  • Jiménez-González, Daniel
  • Alonso, Javier
La enseñanza de contenidos técnicos en grupos multidisciplinares es una tarea compleja debido a la diversidad de los conocimientos iniciales de los alumnos implicados. Sin embargo esta dificultad puede llegar a convertirse en una poderosa herramienta. En este artículo se presenta la experiencia obtenida durante una asignatura de software libre impartida a un grupo multidisciplinar formado por alumnos de distintas carreras técnicas. La base de nuestra propuesta es obligar a los estudiantes a cooperar forzando grupos compuestos por alumnos de diferentes carreras. Este tipo de agrupación obliga a los alumnos a realizar trabajo cooperativo y aprendizaje entre iguales, lo que les permite desarrollar habilidades tanto técnicas como profesionales. Nuestros resultados muestran que con este enfoque se consiguen buenos resultados tanto en el aprendizaje como en la aceptación por parte de los alumnos de la asignatura y del método de enseñanza.
RUA. Repositorio Institucional de la Universidad de Alicante
Publicación Otros (Other)

Mapa Conceptual Global como herramienta para la visión de conjunto de un Sistema Operativo

  • Verdú, Javier
  • López, David
  • Pajuelo, Alex
Numerosas asignaturas están formadas por un temario que está totalmente interrelacionado. Al final del curso los estudiantes deberían haber adquirido los conocimientos de cada tema pero, más importante aún, deberían saber cómo interactúan los diferentes temas entre ellos para obtener una visión global de la asignatura. Sin embargo, a menudo los estudiantes se centran en los temas por separado, en parte porque no les ofrecemos herramientas que les ayuden a relacionar las distintas partes del curso. En este trabajo presentamos el uso de un Mapa Conceptual Global (MCG) de una asignatura como recurso docente que ayuda al estudiante a obtener una visión de conjunto de todo el temario. La experiencia ha sido realizada como complemento de una clase de aprendizaje activo en una asignatura de Sistemas Operativos, pero pensamos que puede ser fácilmente aplicable a otros cursos.
UPCommons. Portal del coneixement obert de la UPC
Publicación Contribución de congreso (ConferenceOutput)

Loop level speculation in a task based programming model

  • Gayatri, Rahulkumar
  • Badia Sala, Rosa Maria|||0000-0003-2941-5499
  • Ayguadé Parra, Eduard|||0000-0002-5146-103X
Uncountable loops (such as while loops in C) and if-conditions are some of the most common constructs in programming. While-loops are widely used to determine the convergence in linear algebra algorithms or goal finding problems from graph algorithms, to name a few. In general while-loops are used whenever the loop iteration space, the number of iterations a loop executes is unknown. Usually in while-loops, the execution of the next iteration is decided inside the current loop iteration (i.e. the execution of iteration i depends on the values computed in iteration i-1). This precludes their parallel execution in today's ubiquitous multi-core architectures. In this paper a technique to speculatively create parallel tasks from the next iterations before the current one completes is proposed. If consecutive loop-iterations are only control dependent, then multiple iterations can be executed simultaneously; later in the execution path, the runtime system will decide to either commit the results of such speculatively executed iterations or undo the changes made by them. Data dependences within or between non-speculative and speculative work are honored to guarantee correctness. The proposed technique is implemented in SMPSs, a task-based dataflow programming model for shared-memory multiprocessor architectures. The approach is evaluated on a set of applications from graph algorithms and linear algebra. Results are promising with an average increase in the speedup of 1.2x with 16 threads when compared to non speculative execution of the applications. The increase in the speedup is significant, since the performance gain is achieved over an already parallelized version of the benchmarks., We thankfully acknowledge the support of the European Commission through the TERAFLUX project (FP7-249013) and the HiPEAC-3 (contract FP7-ICT 287759) Network of Excellence (FP7/ICT 217068), the support of the Spanish Ministry of Education (TIN2007-60625, CSD2007-00050 and FI program), the Generalitat de Catalunya (2009- SGR-980) and the national project (TIN2012-34557)., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Artículo científico (JournalArticle)

Dynamic CPU provisioning for self-managed secure web applications in SMP hosting platforms

  • Guitart Fernández, Jordi|||0000-0003-0751-3100
  • Carrera Pérez, David|||0000-0003-4898-3424
  • Beltran Querol, Vicenç
  • Torres Viñals, Jordi|||0000-0003-1963-7418
  • Ayguadé Parra, Eduard|||0000-0002-5146-103X
Overload control mechanisms such as admission control and connection differentiation have proven effective for preventing overload of application servers running secure web applications. However, achieving optimal results in overload prevention is only possible when some kind of resource management is considered in addition to these mechanisms. In this paper we propose an overload control strategy for secure web applications that brings together dynamic provisioning of platform resources and admission control based on secure socket layer (SSL) connection differentiation. Dynamic provisioning enables additional resources to be allocated to an application on demand to handle workload increases, while the admission control mechanism avoids the server’s performance degradation by dynamically limiting the number of new SSL connections accepted and preferentially serving resumed SSL connections (to maximize performance on session-based environments) while additional resources are being provisioned. Our evaluation demonstrates the benefit of our proposal for efficiently managing the resources and preventing server overload on a 4-way multiprocessor Linux hosting platform, especially when the hosting platform is fully overloaded., This work is supported by the Ministry of Science and Technology of Spain and the European Union (FEDER funds) under Contracts TIN2004-07739-C02-01 and TIN2007-60625 and by the BSC (Barcelona Supercomputing Center)., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Artículo científico (JournalArticle)

Leveraging task-parallelism in message-passing dense matrix factorizations using SMPSs

  • Martín Huertas, Alberto Francisco|||0000-0001-5751-4561
  • Reyes, Ruyman
  • Badia Sala, Rosa Maria|||0000-0003-2941-5499
  • Quintana Ortí, Enrique Salvador
In this paper, we investigate how to exploit task-parallelism during the execution of the Cholesky factorization on clusters of multicore processors with the SMPSs programming model. Our analysis reveals that the major difficulties in adapting the code for this operation in ScaLAPACK to SMPSs lie in algorithmic restrictions and the semantics of the SMPSs programming model, but also that they both can be overcome with a limited programming effort. The experimental results report considerable gains in performance and scalability of the routine parallelized with SMPSs when compared with conventional approaches to execute the original ScaLAPACK implementation in parallel as well as two recent message-passing routines for this operation. In summary, our study opens the door to the possibility of reusing message-passing legacy codes/libraries for linear algebra, by introducing up-to-date techniques like dynamic out-of-order scheduling that significantly upgrade their performance, while avoiding a costly rewrite/reimplementation., This research was supported by Project EU INFRA-2010-1.2.2 \TEXT:Towards EXa op applicaTions". The researcher at BSC-CNS was supported by the HiPEAC-2 Network of Excellence (FP7/ICT 217068), the Spanish Ministry of Education (CICYT TIN2011-23283, TIN2007-60625 and CSD2007- 00050), and the Generalitat de Catalunya (2009-SGR-980). The researcher at CIMNE was partially funded by the UPC postdoctoral grants under the programme \BKC5-Atracció i Fidelització de talent al BKC". The researcher at UJI was supported by project CICYT TIN2008-06570-C04-01 and FEDER. We thank Jesus Labarta, from BSC-CNS, for helpful discussions on SMPSs and his help with the performance analysis of the codes with Paraver. We thank Vladimir Marjanovic, also from BSC-CNS, for his help in the set-up and tuning of the MPI/SMPSs tools on JuRoPa. Finally, we thank Rafael Mayo, from UJI, for his support in the preliminary stages of this work. The authors gratefully acknowledge the computing time granted on the supercomputer JuRoPa at Jülich Supercomputing Centrer., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Artículo científico (JournalArticle)

Business-driven management of infrastructure-level risks in Cloud providers

  • Fitó Comellas, Josep Oriol
  • Guitart Fernández, Jordi|||0000-0003-0751-3100
Cloud computing is an innovative and promising paradigm that is leading to remarkable changes in the way in which hardware and software are designed and purchased, as well as how IT systems are managed. However, the Cloud is a risky paradigm. For instance, the use of Cloud services, which usually are external assets to their consumers, implies unprecedented risks that must be taken into account.; In this paper, we propose the involvement of the risk management discipline into the Cloud computing realm. We present a risk management approach led by business-level objectives (BLOs) of Cloud organizations. Its main goal is to assist in business-driven self-managed Cloud providers, by facing uncertainties always present in their internal decision-making processes. Our Cloud-aware risk management method includes a SEmi-quantitative BLO-driven Cloud Risk Assessment (SEBCRA) as the core subprocess. Its aim is to constantly rank and prioritize risks affecting the governing business-level goals.; In addition, we present, as a use case, a PaaS provider that incorporates our risk management approach to enhance the achievement of two BLOs, i.e. maximization of profit and customer satisfaction. In particular, it can manage - identify, assess, and treat - the most critical Cloud infrastructure-level risks, i.e. provisioning its private Cloud, either under- or over-provisioning, as well as resource failures. We present some risk treatment responses to face these risks and we evaluate their impact on the above-mentioned BLOs. Our results show that the best responses to address risks may change over time depending on the current provider's status. As a result, an adaptive management of risks should be considered as a mandatory process for Cloud providers to ensure their success in the ever-growing worldwide ecosystem of Clouds. (C) 2012 Elsevier B.V. All rights reserved., This work is supported by the Ministry of Science and Technology of Spain and the European Union (FEDER funds) under contract TIN2007-60625, by the Generalitat de Catalunya under contract 2009-SGR-980, and by the European Commission under FP7-ICT-2009-5 contract 257115 (OPTIMIS)., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Artículo científico (JournalArticle)

DaSH: a benchmark suite for hybrid dataflow and shared memory programming models

  • Gajinov, Vladimir
  • Stipic, Srdjan
  • Eric, Igor
  • Unsal, Osman Sabri
  • Ayguadé Parra, Eduard|||0000-0002-5146-103X
  • Cristal Kestelman, Adrián|||0000-0003-1277-9296
The current trend in development of parallel programming models is to combine different well established models into a single programming, model in order to support efficient implementation of a wide range of real world applications. The dataflow model has particularly managed to recapture the interest of the research community due to its ability to express parallelism efficiently. Thus, a number of recently proposed hybrid parallel programming models combine dataflow and traditional shared memory models. Their findings have influenced the introduction of task dependency in the OpenMP 4.0 standard.; This article presents DaSH - the first comprehensive benchmark suite for hybrid dataflow and shared memory programming models. DaSH features 11 benchmarks, each representing one of the Berkeley dwarfs that capture patterns of communication and computation common to a wide range of emerging applications. DaSH also includes sequential and shared-memory implementations based on OpenMP and Intel TBB to facilitate easy comparison between hybrid dataflow implementations and traditional shared memory implementations based on work-sharing and/or tasks. Finally, we use DaSH to evaluate three different hybrid dataflow models, identify their advantages and shortcomings, and motivate further research on their characteristics., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Contribución de congreso (ConferenceOutput)

Analysis of the Task Superscalar architecture hardware design

  • Yazdanpanah Ahmadabadi, Fahimeh
  • Jiménez González, Daniel|||0000-0001-6064-7883
  • Álvarez Martínez, Carlos|||0000-0003-0536-5183
  • Etsion, Yoav
  • Badia Sala, Rosa Maria|||0000-0003-2941-5499
In this paper, we analyze the operational flow of two hardware implementations of the Task Superscalar architecture. The Task Superscalar is an experimental task based dataflow scheduler that dynamically detects inter-task data dependencies, identifies task-level parallelism, and executes tasks in the out-of-order manner. In this paper, we present a base implementation of the Task Superscalar architecture, as well as a new design with improved performance. We study the behavior of processing some dependent and non-dependent tasks with both base and improved hardware designs and present the simulation results compared with the results of the runtime implementation., This work is supported by the Ministry of Science and Technology of Spain and the European Union (FEDER funds) under contract TIN2007-60625, by the Generalitat de Catalunya (contract 2009-SGR-980), and by the European FP7 project TERAFLUX id. 249013, http://www.tera ux.eu. We would also like to thank the Xilinx University Program for its hardware and software donations.
UPCommons. Portal del coneixement obert de la UPC
Publicación Contribución de congreso (ConferenceOutput)

HARP: Adaptive abort recurrence prediction for Hardware Transactional Memory

  • Armejach Sanosa, Adrià|||0000-0003-2869-668X
  • Negi, Anurag
  • Cristal Kestelman, Adrián|||0000-0003-1277-9296
  • Unsal, Osman Sabri
  • Stenström, Per
  • Harris, Tim
Hardware Transactional Memory (HTM) exposes parallelism by allowing possibly conflicting sections of code, called transactions, to execute concurrently in multithreaded applications. However, conflicts among concurrent transactions result in wasted computation and expensive rollbacks. Under high contention HTM protocol overheads can, in many cases, amount to several times the useful work done. Blindly scheduling transactions in the presence of contention is therefore clearly suboptimal from a resource utilization standpoint, especially in situations where several scheduling options exist. This paper presents HARP (Hardware Abort Recurrence Predictor), a hardware-only mechanism to avoid speculation when it is likely to fail. Inspired by branch prediction strategies and prior work on contention management and scheduling in HTM, HARP uses past behavior of transactions and locality in conflicting memory references to accurately predict conflicts. The prediction mechanism adapts to variations in workload characteristics and enables better utilization of computational resources. We show that an HTM protocol that integrates HARP exhibits reductions in both wasted execution time and serialization overheads when compared to prior work, leading to a significant increase in throughput (~30%) in both single-application and multi-application scenarios., Adria Armejach's work is supported by the Ministry of Science and Technology of Spain and the European Union under contracts TIN2007-60625 and TIN2008-02055-E. Anurag Negi's work at Chalmers is supported by grants from the Swedish Foundation for Strategic Research (SSF) under the SCHEME project., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Contribución de congreso (ConferenceOutput)

Self-adaptive OmpSs tasks in heterogeneous environments

  • Planas Carbonell, Judit|||0000-0002-8221-7988
  • Badia Sala, Rosa Maria|||0000-0003-2941-5499
  • Ayguadé Parra, Eduard|||0000-0002-5146-103X
  • Labarta Mancho, Jesús José|||0000-0002-7489-4727
As new heterogeneous systems and hardware accelerators appear, high performance computers can reach a higher level of computational power. Nevertheless, this does not come for free: the more heterogeneity the system presents, the more complex becomes the programming task in terms of resource management. OmpSs is a task-based programming model and framework focused on the runtime exploitation of parallelism from annotated sequential applications. This paper presents a set of extensions to this framework: we show how the application programmer can expose different specialized versions of tasks (i.e. pieces of specific code targeted and optimized for a particular architecture) and how the system can choose between these versions at runtime to obtain the best performance achievable for the given application. From the results obtained in a multi-GPU system, we prove that our proposal gives flexibility to application's source code and can potentially increase application's performance., This work has been supported by the European Commission through the ENCORE project (FP7-248647), the TERAFLUX project (FP7-249013), the TEXT project (FP7-261580), the HiPEAC-3 Network of Excellence (FP7-ICT 287759), the Intel-BSC Exascale Lab collaboration project, the support of the Spanish Ministry of Education (CSD2007- 00050 and FPU program), the projects of Computación de Altas Prestaciones V and VI (TIN2007-60625, TIN2012-34557) and the Generalitat de Catalunya (2009-SGR-980)., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Artículo científico (JournalArticle)

A service framework for energy-aware monitoring and VM management in clouds

  • Katsaros, Gregory
  • Subirats, Josep
  • Fitó Comellas, Josep Oriol
  • Guitart Fernández, Jordi|||0000-0003-0751-3100
  • Gilet, Pierre
  • Espling, Daniel
The monitoring of QoS parameters in Services Computing as well as in Clouds has been a functionality provided by all contemporary systems. As the optimization of energy consumption becomes a major concern for system designers and administrators, it can be considered as another QoS metric to be monitored. In this paper, we present a service framework that allows us to monitor the energy consumption of a Cloud infrastructure, calculate its energy efficiency, and evaluate the gathered data in order to put in place an effective virtual machine (VM) management. In that context, a simulation scenario of an eco-driven VM placement policy resulted in a 14% improvement of the infrastructure’s energy efficiency. In total, the proposed approaches and implementations have been validated against a testbed, producing very promising results regarding the prospect of energy efficiency as an important quality factor in Clouds., This work has been supported by the OPTIMIS project (http://www.optimis-project.eu/)[33], [21] and has been partly funded by the European Commission’s IST activity of the 7th Framework Program under contract number 257115. This work expresses the opinions of the authors and not necessarily those of the European Commission. The European Commission is not liable for any use that may be made of the information contained in this work. It is also supported by the Ministry of Science and Technology of Spain and the European Union (FEDER funds) under contract TIN2007-60625, by the Generalitat de Catalunya under contract 2009-SGR-980., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Contribución de congreso (ConferenceOutput)

Scalability and parallel execution of OmpSs-OpenCL tasks on heterogeneous CPU-GPU environment

  • Elangovan, Vinoth Krishnan
  • Badia Sala, Rosa Maria|||0000-0003-2941-5499
  • Ayguadé Parra, Eduard|||0000-0002-5146-103X
With heterogeneous computing becoming mainstream, researchers and software vendors have been trying to exploit the best of the underlying architectures like GPUs or CPUs to enhance performance. Parallel programming models play a crucial role in achieving this enhancement. One such model is OpenCL, a parallel computing API for cross platform computations targeting heterogeneous architectures. However, OpenCL is a low-level programming language, therefore it can be time consuming to directly develop OpenCL code. To address this shortcoming, OpenCL has been integrated with OmpSs, a task-based programming model to provide abstraction to the user thereby reducing programmer effort. OmpSs-OpenCL programming model deals with a single OpenCL device either a CPU or a GPU. In this paper, we upgrade OmpSs-OpenCL programming model by supporting parallel execution of tasks across multiple CPU-GPU heterogeneous platforms. We discuss the design of the programming model along with its asynchronous runtime system. We investigated scalability of four OmpSs-OpenCL benchmarks across 4 GPUs gaining speedup of up to 4x. Further, in order to achieve effective utilization of the computing resources, we present static and work-stealing scheduling techniques. We show results of parallel execution of applications using OmpSs-OpenCL model and use heterogeneous workloads to evaluate our scheduling techniques on a heterogeneous CPU-GPU platform., We thankfully acknowledge the support of the European Commission through the TERAFLUX project (FP7-249013) and the HiPEAC-2 Network of Excellence (FP7/ICT 217068),the support of the Spanish Ministry of Education (TIN-2007-60625, TIN-2012-34557, CSD2007-00050 and FI program) and the Generalitat de Catalunya (2009-SGR-980), Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Contribución de congreso (ConferenceOutput)

Programmable and scalable reductions on clusters

  • Ciesko, Jan
  • Bueno Hedo, Javier
  • Puzovic, Nikola
  • Ramírez Bellido, Alejandro
  • Badia Sala, Rosa Maria|||0000-0003-2941-5499
  • Labarta Mancho, Jesús José|||0000-0002-7489-4727
Reductions matter and they are here to stay. Wide adoption of parallel processing hardware in a broad range of computer applications has encouraged recent research efforts on their efficient parallelization. Furthermore, trends towards high productivity languages in mainstream computing increases the demand for efficient programming support. In this paper we present a new approach on parallel reductions for distributed memory systems that provides both scalability and programmability. Using OmpSs, a task-based parallel programming model, the developer has the ability to express scalable reductions through a single pragma annotation. This pragma annotation is applicable for tasks as well as for work-sharing constructs (with implicit tasking) and instructs the compiler to generate the required runtime calls. The supporting runtime handles data and task distribution, parallel execution and data reduction. Scalability is achieved through a software cache that maximizes local and temporal data reuse and allows overlapped computation and communication. Results confirm scalability for up to 32 12-core cluster nodes., We thankfully acknowledge the support of the European Commission through the ENCORE project (FP7-248647), the TERAFLUX project (FP7-249013), the TEXT project (FP7- 261580), and the HiPEAC-3 Network of Excellence (FP7/ICT217068), further the support of Intel-BSC Exascale Center, the Spanish Ministry of Education (TIN2007-60625, CSD2007-00050 and FPU program) and the Generalitat de Catalunya (2009-SGR-980).
UPCommons. Portal del coneixement obert de la UPC
Publicación Artículo científico (JournalArticle)

Thread assignment of multithreaded network applications in multicore/multithreaded processors

  • Radojković, Petar
  • Cakarevic, Vladimir
  • Verdú Mulà, Javier|||0000-0003-4485-2419
  • Pajuelo González, Manuel Alejandro|||0000-0002-5510-6860
  • Cazorla Almeida, Francisco Javier
  • Nemirovsky, Mario
  • Valero Cortés, Mateo|||0000-0003-2917-2482
© 2013 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works., The introduction of multithreaded processors comprised of a large number of cores with many shared resources makes thread scheduling, and in particular optimal assignment of running threads to processor hardware contexts to become one of the most promising ways to improve the system performance. However, finding optimal thread assignments for workloads running in state-of-the-art multicore/multithreaded processors is an NP-complete problem. In this paper, we propose BlackBox scheduler, a systematic method for thread assignment of multithreaded network applications running on multicore/multithreaded processors. The method requires minimum information about the target processor architecture and no data about the hardware requirements of the applications under study. The proposed method is evaluated with an industrial case study for a set of multithreaded network applications running on the UltraSPARC T2 processor. In most of the experiments, the proposed thread assignment method detected the best actual thread assignment in the evaluation sample. The method improved the system performance from 5 to 48 percent with respect to load balancing algorithms used in state-of-the-art OSs, and up to 60 percent with respect to a naive thread assignment., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Contribución de congreso (ConferenceOutput)

Implementing OmpSs support for regions of data in architectures with multiple address spaces

  • Bueno Hedo, Javier
  • Martorell Bofill, Xavier|||0000-0002-0417-3430
  • Badia Sala, Rosa Maria|||0000-0003-2941-5499
  • Ayguadé Parra, Eduard|||0000-0002-5146-103X
  • Labarta Mancho, Jesús José|||0000-0002-7489-4727
The need for features for managing complex data accesses in modern programming models has increased due to the emerging hardware architectures. HPC hardware has moved towards clusters of accelerators and/or multicores, architectures with a complex memory hierarchy exposed to the programmer. We present the implementation of data regions on the OmpSs programming model, a high-productivity annotation-based programming model derived from OpenMP. This enables the programmer to specify regions of strided and/or overlapped data used by the parallel tasks of the application. The data will be automatically managed by the underlying run-time environment, which could transparently apply optimization techniques to improve performance. This approach based on a high-productivity programming model contrasts with more direct approaches like MPI, where the programmer has to explicitly deal with the data management. It is generally believed that these are capable of achieving the best possible performance, so we also compare the performance of several OmpSs applications against well-known counterparts MPI implementations obtaining comparable or better results., We thankfully acknowledge the support of the European Commission through the ENCORE project (FP7-248647), the TERAFLUX project (FP7-249013), the TEXT project (FP7-261580), and the HiPEAC-3 Network of Excellence (ICT FP7 NoE 287759), the support of the Spanish Ministry of Science and Technology (TIN2007-60625, TIN2012-34557, CSD2007-00050 and FPU program) and the Generalitat de Catalunya (2009-SGR-980, Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Artículo científico (JournalArticle)

Hints to improve automatic load balancing with LeWI for hybrid applications

  • Garcia Gasulla, Marta|||0000-0003-3682-9905
  • Labarta Mancho, Jesús José|||0000-0002-7489-4727
  • Corbalán González, Julita|||0000-0002-3926-5634
The DLB (Dynamic Load Balancing) library and LeWl (LEnd When Idle) algorithm provide a runtime solution to deal with the load imbalance of parallel applications independently of the source of imbalance. DLB relies on the usage of hybrid programming models and exploits the malleability of the second level of parallelism to redistribute computation power across processes.; When executing real applications with LeWl, although application's performance is significantly improved, we have observed in some cases efficiency values between 60% and 70%, far from our theoretical limit.; This work is a deep analysis of the sources of efficiency loss correlated with application characteristics, parallelization schemes and programming models. We have based our analysis in fine grain monitoring tools and metrics and validated our conclusions by reproducing them in synthetic experiments. As a result, this work teaches us some lessons that can be seen as hints to programmers to help LeWI make an efficient use of computational resources and obtain the maximum performance. (C) 2014 Elsevier Inc. All rights reserved., We thankfully acknowledge the support of the Spanish Ministry of Education (TIN2007-60625, and CSD2007-00050) and the Generalitat de Catalunya (2009-SGR-980)., Peer Reviewed
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

Programmable and scalable reductions on clusters

  • Ciesko, Jan
  • Bueno Hedo, Javier
  • Puzovic, Nikola
  • Ramírez Bellido, Alejandro
  • Badia Sala, Rosa Maria
  • Labarta Mancho, Jesús José
Reductions matter and they are here to stay. Wide adoption of parallel processing hardware in a broad range of computer applications has encouraged recent research efforts on their efficient parallelization. Furthermore, trends towards high productivity languages in mainstream computing increases the demand for efficient programming support. In this paper we present a new approach on parallel reductions for distributed memory systems that provides both scalability and programmability. Using OmpSs, a task-based parallel programming model, the developer has the ability to express scalable reductions through a single pragma annotation. This pragma annotation is applicable for tasks as well as for work-sharing constructs (with implicit tasking) and instructs the compiler to generate the required runtime calls. The supporting runtime handles data and task distribution, parallel execution and data reduction. Scalability is achieved through a software cache that maximizes local and temporal data reuse and allows overlapped computation and communication. Results confirm scalability for up to 32 12-core cluster nodes., We thankfully acknowledge the support of the European Commission through the ENCORE project (FP7-248647), the TERAFLUX project (FP7-249013), the TEXT project (FP7- 261580), and the HiPEAC-3 Network of Excellence (FP7/ICT217068), further the support of Intel-BSC Exascale Center, the Spanish Ministry of Education (TIN2007-60625, CSD2007-00050 and FPU program) and the Generalitat de Catalunya (2009-SGR-980)., Postprint (published version)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Artículo científico (JournalArticle)

A service framework for energy-aware monitoring and VM management in clouds

  • Katsaros, Gregory
  • Subirats, Josep
  • Fitó Comellas, Josep Oriol
  • Guitart Fernández, Jordi
  • Gilet, Pierre
  • Espling, Daniel
The monitoring of QoS parameters in Services Computing as well as in Clouds has been a functionality provided by all contemporary systems. As the optimization of energy consumption becomes a major concern for system designers and administrators, it can be considered as another QoS metric to be monitored. In this paper, we present a service framework that allows us to monitor the energy consumption of a Cloud infrastructure, calculate its energy efficiency, and evaluate the gathered data in order to put in place an effective virtual machine (VM) management. In that context, a simulation scenario of an eco-driven VM placement policy resulted in a 14% improvement of the infrastructure’s energy efficiency. In total, the proposed approaches and implementations have been validated against a testbed, producing very promising results regarding the prospect of energy efficiency as an important quality factor in Clouds., This work has been supported by the OPTIMIS project (http://www.optimis-project.eu/)[33], [21] and has been partly funded by the European Commission’s IST activity of the 7th Framework Program under contract number 257115. This work expresses the opinions of the authors and not necessarily those of the European Commission. The European Commission is not liable for any use that may be made of the information contained in this work. It is also supported by the Ministry of Science and Technology of Spain and the European Union (FEDER funds) under contract TIN2007-60625, by the Generalitat de Catalunya under contract 2009-SGR-980., Peer Reviewed, Postprint (author's final draft)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Artículo científico (JournalArticle)

DaSH: a benchmark suite for hybrid dataflow and shared memory programming models

  • Gajinov, Vladimir
  • Stipic, Srdjan
  • Eric, Igor
  • Unsal, Osman Sabri
  • Ayguadé Parra, Eduard
  • Cristal Kestelman, Adrián
The current trend in development of parallel programming models is to combine different well established models into a single programming, model in order to support efficient implementation of a wide range of real world applications. The dataflow model has particularly managed to recapture the interest of the research community due to its ability to express parallelism efficiently. Thus, a number of recently proposed hybrid parallel programming models combine dataflow and traditional shared memory models. Their findings have influenced the introduction of task dependency in the OpenMP 4.0 standard.; This article presents DaSH - the first comprehensive benchmark suite for hybrid dataflow and shared memory programming models. DaSH features 11 benchmarks, each representing one of the Berkeley dwarfs that capture patterns of communication and computation common to a wide range of emerging applications. DaSH also includes sequential and shared-memory implementations based on OpenMP and Intel TBB to facilitate easy comparison between hybrid dataflow implementations and traditional shared memory implementations based on work-sharing and/or tasks. Finally, we use DaSH to evaluate three different hybrid dataflow models, identify their advantages and shortcomings, and motivate further research on their characteristics., Peer Reviewed, Postprint (published version)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Artículo científico (JournalArticle)

Dynamic CPU provisioning for self-managed secure web applications in SMP hosting platforms

  • Guitart Fernández, Jordi
  • Carrera Pérez, David
  • Beltran Querol, Vicenç
  • Torres Viñals, Jordi
  • Ayguadé Parra, Eduard
Overload control mechanisms such as admission control and connection differentiation have proven effective for preventing overload of application servers running secure web applications. However, achieving optimal results in overload prevention is only possible when some kind of resource management is considered in addition to these mechanisms. In this paper we propose an overload control strategy for secure web applications that brings together dynamic provisioning of platform resources and admission control based on secure socket layer (SSL) connection differentiation. Dynamic provisioning enables additional resources to be allocated to an application on demand to handle workload increases, while the admission control mechanism avoids the server’s performance degradation by dynamically limiting the number of new SSL connections accepted and preferentially serving resumed SSL connections (to maximize performance on session-based environments) while additional resources are being provisioned. Our evaluation demonstrates the benefit of our proposal for efficiently managing the resources and preventing server overload on a 4-way multiprocessor Linux hosting platform, especially when the hosting platform is fully overloaded., This work is supported by the Ministry of Science and Technology of Spain and the European Union (FEDER funds) under Contracts TIN2004-07739-C02-01 and TIN2007-60625 and by the BSC (Barcelona Supercomputing Center)., Peer Reviewed, Postprint (author's final draft)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

HARP: Adaptive abort recurrence prediction for Hardware Transactional Memory

  • Armejach Sanosa, Adrià
  • Negi, Anurag
  • Cristal Kestelman, Adrián
  • Unsal, Osman Sabri
  • Stenström, Per
  • Harris, Tim
Hardware Transactional Memory (HTM) exposes parallelism by allowing possibly conflicting sections of code, called transactions, to execute concurrently in multithreaded applications. However, conflicts among concurrent transactions result in wasted computation and expensive rollbacks. Under high contention HTM protocol overheads can, in many cases, amount to several times the useful work done. Blindly scheduling transactions in the presence of contention is therefore clearly suboptimal from a resource utilization standpoint, especially in situations where several scheduling options exist. This paper presents HARP (Hardware Abort Recurrence Predictor), a hardware-only mechanism to avoid speculation when it is likely to fail. Inspired by branch prediction strategies and prior work on contention management and scheduling in HTM, HARP uses past behavior of transactions and locality in conflicting memory references to accurately predict conflicts. The prediction mechanism adapts to variations in workload characteristics and enables better utilization of computational resources. We show that an HTM protocol that integrates HARP exhibits reductions in both wasted execution time and serialization overheads when compared to prior work, leading to a significant increase in throughput (~30%) in both single-application and multi-application scenarios., Adria Armejach's work is supported by the Ministry of Science and Technology of Spain and the European Union under contracts TIN2007-60625 and TIN2008-02055-E. Anurag Negi's work at Chalmers is supported by grants from the Swedish Foundation for Strategic Research (SSF) under the SCHEME project., Peer Reviewed, Postprint (author's final draft)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

Implementing OmpSs support for regions of data in architectures with multiple address spaces

  • Bueno Hedo, Javier
  • Martorell Bofill, Xavier
  • Badia Sala, Rosa Maria
  • Ayguadé Parra, Eduard
  • Labarta Mancho, Jesús José
The need for features for managing complex data accesses in modern programming models has increased due to the emerging hardware architectures. HPC hardware has moved towards clusters of accelerators and/or multicores, architectures with a complex memory hierarchy exposed to the programmer. We present the implementation of data regions on the OmpSs programming model, a high-productivity annotation-based programming model derived from OpenMP. This enables the programmer to specify regions of strided and/or overlapped data used by the parallel tasks of the application. The data will be automatically managed by the underlying run-time environment, which could transparently apply optimization techniques to improve performance. This approach based on a high-productivity programming model contrasts with more direct approaches like MPI, where the programmer has to explicitly deal with the data management. It is generally believed that these are capable of achieving the best possible performance, so we also compare the performance of several OmpSs applications against well-known counterparts MPI implementations obtaining comparable or better results., We thankfully acknowledge the support of the European Commission through the ENCORE project (FP7-248647), the TERAFLUX project (FP7-249013), the TEXT project (FP7-261580), and the HiPEAC-3 Network of Excellence (ICT FP7 NoE 287759), the support of the Spanish Ministry of Science and Technology (TIN2007-60625, TIN2012-34557, CSD2007-00050 and FPU program) and the Generalitat de Catalunya (2009-SGR-980, Peer Reviewed, Postprint (published version)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Artículo científico (JournalArticle)

Hints to improve automatic load balancing with LeWI for hybrid applications

  • Garcia Gasulla, Marta
  • Labarta Mancho, Jesús José
  • Corbalán González, Julita
The DLB (Dynamic Load Balancing) library and LeWl (LEnd When Idle) algorithm provide a runtime solution to deal with the load imbalance of parallel applications independently of the source of imbalance. DLB relies on the usage of hybrid programming models and exploits the malleability of the second level of parallelism to redistribute computation power across processes.; When executing real applications with LeWl, although application's performance is significantly improved, we have observed in some cases efficiency values between 60% and 70%, far from our theoretical limit.; This work is a deep analysis of the sources of efficiency loss correlated with application characteristics, parallelization schemes and programming models. We have based our analysis in fine grain monitoring tools and metrics and validated our conclusions by reproducing them in synthetic experiments. As a result, this work teaches us some lessons that can be seen as hints to programmers to help LeWI make an efficient use of computational resources and obtain the maximum performance. (C) 2014 Elsevier Inc. All rights reserved., We thankfully acknowledge the support of the Spanish Ministry of Education (TIN2007-60625, and CSD2007-00050) and the Generalitat de Catalunya (2009-SGR-980)., Peer Reviewed, Postprint (author's final draft)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Artículo científico (JournalArticle)

Thread assignment of multithreaded network applications in multicore/multithreaded processors

  • Radojković, Petar
  • Cakarevic, Vladimir
  • Verdú Mulà, Javier
  • Pajuelo González, Manuel Alejandro
  • Cazorla Almeida, Francisco Javier
  • Nemirovsky, Mario
  • Valero Cortés, Mateo
© 2013 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works., The introduction of multithreaded processors comprised of a large number of cores with many shared resources makes thread scheduling, and in particular optimal assignment of running threads to processor hardware contexts to become one of the most promising ways to improve the system performance. However, finding optimal thread assignments for workloads running in state-of-the-art multicore/multithreaded processors is an NP-complete problem. In this paper, we propose BlackBox scheduler, a systematic method for thread assignment of multithreaded network applications running on multicore/multithreaded processors. The method requires minimum information about the target processor architecture and no data about the hardware requirements of the applications under study. The proposed method is evaluated with an industrial case study for a set of multithreaded network applications running on the UltraSPARC T2 processor. In most of the experiments, the proposed thread assignment method detected the best actual thread assignment in the evaluation sample. The method improved the system performance from 5 to 48 percent with respect to load balancing algorithms used in state-of-the-art OSs, and up to 60 percent with respect to a naive thread assignment., Peer Reviewed, Postprint (author's final draft)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Artículo científico (JournalArticle)

Leveraging task-parallelism in message-passing dense matrix factorizations using SMPSs

  • Martín Huertas, Alberto Francisco
  • Reyes, Ruyman
  • Badia Sala, Rosa Maria
  • Quintana Ortí, Enrique Salvador
In this paper, we investigate how to exploit task-parallelism during the execution of the Cholesky factorization on clusters of multicore processors with the SMPSs programming model. Our analysis reveals that the major difficulties in adapting the code for this operation in ScaLAPACK to SMPSs lie in algorithmic restrictions and the semantics of the SMPSs programming model, but also that they both can be overcome with a limited programming effort. The experimental results report considerable gains in performance and scalability of the routine parallelized with SMPSs when compared with conventional approaches to execute the original ScaLAPACK implementation in parallel as well as two recent message-passing routines for this operation. In summary, our study opens the door to the possibility of reusing message-passing legacy codes/libraries for linear algebra, by introducing up-to-date techniques like dynamic out-of-order scheduling that significantly upgrade their performance, while avoiding a costly rewrite/reimplementation., This research was supported by Project EU INFRA-2010-1.2.2 \TEXT:Towards EXa op applicaTions". The researcher at BSC-CNS was supported by the HiPEAC-2 Network of Excellence (FP7/ICT 217068), the Spanish Ministry of Education (CICYT TIN2011-23283, TIN2007-60625 and CSD2007- 00050), and the Generalitat de Catalunya (2009-SGR-980). The researcher at CIMNE was partially funded by the UPC postdoctoral grants under the programme \BKC5-Atracció i Fidelització de talent al BKC". The researcher at UJI was supported by project CICYT TIN2008-06570-C04-01 and FEDER. We thank Jesus Labarta, from BSC-CNS, for helpful discussions on SMPSs and his help with the performance analysis of the codes with Paraver. We thank Vladimir Marjanovic, also from BSC-CNS, for his help in the set-up and tuning of the MPI/SMPSs tools on JuRoPa. Finally, we thank Rafael Mayo, from UJI, for his support in the preliminary stages of this work. The authors gratefully acknowledge the computing time granted on the supercomputer JuRoPa at Jülich Supercomputing Centrer., Peer Reviewed, Preprint
Recercat. Dipósit de la Recerca de Catalunya
Publicación Artículo científico (JournalArticle)

Business-driven management of infrastructure-level risks in Cloud providers

  • Fitó Comellas, Josep Oriol
  • Guitart Fernández, Jordi
Cloud computing is an innovative and promising paradigm that is leading to remarkable changes in the way in which hardware and software are designed and purchased, as well as how IT systems are managed. However, the Cloud is a risky paradigm. For instance, the use of Cloud services, which usually are external assets to their consumers, implies unprecedented risks that must be taken into account.; In this paper, we propose the involvement of the risk management discipline into the Cloud computing realm. We present a risk management approach led by business-level objectives (BLOs) of Cloud organizations. Its main goal is to assist in business-driven self-managed Cloud providers, by facing uncertainties always present in their internal decision-making processes. Our Cloud-aware risk management method includes a SEmi-quantitative BLO-driven Cloud Risk Assessment (SEBCRA) as the core subprocess. Its aim is to constantly rank and prioritize risks affecting the governing business-level goals.; In addition, we present, as a use case, a PaaS provider that incorporates our risk management approach to enhance the achievement of two BLOs, i.e. maximization of profit and customer satisfaction. In particular, it can manage - identify, assess, and treat - the most critical Cloud infrastructure-level risks, i.e. provisioning its private Cloud, either under- or over-provisioning, as well as resource failures. We present some risk treatment responses to face these risks and we evaluate their impact on the above-mentioned BLOs. Our results show that the best responses to address risks may change over time depending on the current provider's status. As a result, an adaptive management of risks should be considered as a mandatory process for Cloud providers to ensure their success in the ever-growing worldwide ecosystem of Clouds. (C) 2012 Elsevier B.V. All rights reserved., This work is supported by the Ministry of Science and Technology of Spain and the European Union (FEDER funds) under contract TIN2007-60625, by the Generalitat de Catalunya under contract 2009-SGR-980, and by the European Commission under FP7-ICT-2009-5 contract 257115 (OPTIMIS)., Peer Reviewed, Postprint (author's final draft)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

Scalability and parallel execution of OmpSs-OpenCL tasks on heterogeneous CPU-GPU environment

  • Elangovan, Vinoth Krishnan
  • Badia Sala, Rosa Maria
  • Ayguadé Parra, Eduard
With heterogeneous computing becoming mainstream, researchers and software vendors have been trying to exploit the best of the underlying architectures like GPUs or CPUs to enhance performance. Parallel programming models play a crucial role in achieving this enhancement. One such model is OpenCL, a parallel computing API for cross platform computations targeting heterogeneous architectures. However, OpenCL is a low-level programming language, therefore it can be time consuming to directly develop OpenCL code. To address this shortcoming, OpenCL has been integrated with OmpSs, a task-based programming model to provide abstraction to the user thereby reducing programmer effort. OmpSs-OpenCL programming model deals with a single OpenCL device either a CPU or a GPU. In this paper, we upgrade OmpSs-OpenCL programming model by supporting parallel execution of tasks across multiple CPU-GPU heterogeneous platforms. We discuss the design of the programming model along with its asynchronous runtime system. We investigated scalability of four OmpSs-OpenCL benchmarks across 4 GPUs gaining speedup of up to 4x. Further, in order to achieve effective utilization of the computing resources, we present static and work-stealing scheduling techniques. We show results of parallel execution of applications using OmpSs-OpenCL model and use heterogeneous workloads to evaluate our scheduling techniques on a heterogeneous CPU-GPU platform., We thankfully acknowledge the support of the European Commission through the TERAFLUX project (FP7-249013) and the HiPEAC-2 Network of Excellence (FP7/ICT 217068),the support of the Spanish Ministry of Education (TIN-2007-60625, TIN-2012-34557, CSD2007-00050 and FI program) and the Generalitat de Catalunya (2009-SGR-980), Peer Reviewed, Postprint (author’s final draft)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

Self-adaptive OmpSs tasks in heterogeneous environments

  • Planas Carbonell, Judit
  • Badia Sala, Rosa Maria
  • Ayguadé Parra, Eduard
  • Labarta Mancho, Jesús José
As new heterogeneous systems and hardware accelerators appear, high performance computers can reach a higher level of computational power. Nevertheless, this does not come for free: the more heterogeneity the system presents, the more complex becomes the programming task in terms of resource management. OmpSs is a task-based programming model and framework focused on the runtime exploitation of parallelism from annotated sequential applications. This paper presents a set of extensions to this framework: we show how the application programmer can expose different specialized versions of tasks (i.e. pieces of specific code targeted and optimized for a particular architecture) and how the system can choose between these versions at runtime to obtain the best performance achievable for the given application. From the results obtained in a multi-GPU system, we prove that our proposal gives flexibility to application's source code and can potentially increase application's performance., This work has been supported by the European Commission through the ENCORE project (FP7-248647), the TERAFLUX project (FP7-249013), the TEXT project (FP7-261580), the HiPEAC-3 Network of Excellence (FP7-ICT 287759), the Intel-BSC Exascale Lab collaboration project, the support of the Spanish Ministry of Education (CSD2007- 00050 and FPU program), the projects of Computación de Altas Prestaciones V and VI (TIN2007-60625, TIN2012-34557) and the Generalitat de Catalunya (2009-SGR-980)., Peer Reviewed, Postprint (author’s final draft)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

Analysis of the Task Superscalar architecture hardware design

  • Yazdanpanah Ahmadabadi, Fahimeh
  • Jiménez González, Daniel
  • Álvarez Martínez, Carlos
  • Etsion, Yoav
  • Badia Sala, Rosa Maria
In this paper, we analyze the operational flow of two hardware implementations of the Task Superscalar architecture. The Task Superscalar is an experimental task based dataflow scheduler that dynamically detects inter-task data dependencies, identifies task-level parallelism, and executes tasks in the out-of-order manner. In this paper, we present a base implementation of the Task Superscalar architecture, as well as a new design with improved performance. We study the behavior of processing some dependent and non-dependent tasks with both base and improved hardware designs and present the simulation results compared with the results of the runtime implementation., This work is supported by the Ministry of Science and Technology of Spain and the European Union (FEDER funds) under contract TIN2007-60625, by the Generalitat de Catalunya (contract 2009-SGR-980), and by the European FP7 project TERAFLUX id. 249013, http://www.tera ux.eu. We would also like to thank the Xilinx University Program for its hardware and software donations., Postprint (author’s final draft)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

Loop level speculation in a task based programming model

  • Gayatri, Rahulkumar
  • Badia Sala, Rosa Maria
  • Ayguadé Parra, Eduard
Uncountable loops (such as while loops in C) and if-conditions are some of the most common constructs in programming. While-loops are widely used to determine the convergence in linear algebra algorithms or goal finding problems from graph algorithms, to name a few. In general while-loops are used whenever the loop iteration space, the number of iterations a loop executes is unknown. Usually in while-loops, the execution of the next iteration is decided inside the current loop iteration (i.e. the execution of iteration i depends on the values computed in iteration i-1). This precludes their parallel execution in today's ubiquitous multi-core architectures. In this paper a technique to speculatively create parallel tasks from the next iterations before the current one completes is proposed. If consecutive loop-iterations are only control dependent, then multiple iterations can be executed simultaneously; later in the execution path, the runtime system will decide to either commit the results of such speculatively executed iterations or undo the changes made by them. Data dependences within or between non-speculative and speculative work are honored to guarantee correctness. The proposed technique is implemented in SMPSs, a task-based dataflow programming model for shared-memory multiprocessor architectures. The approach is evaluated on a set of applications from graph algorithms and linear algebra. Results are promising with an average increase in the speedup of 1.2x with 16 threads when compared to non speculative execution of the applications. The increase in the speedup is significant, since the performance gain is achieved over an already parallelized version of the benchmarks., We thankfully acknowledge the support of the European Commission through the TERAFLUX project (FP7-249013) and the HiPEAC-3 (contract FP7-ICT 287759) Network of Excellence (FP7/ICT 217068), the support of the Spanish Ministry of Education (TIN2007-60625, CSD2007-00050 and FI program), the Generalitat de Catalunya (2009- SGR-980) and the national project (TIN2012-34557)., Peer Reviewed, Postprint (author’s final draft)