Assessing the Performance of the SRR Loop Scheduler with Irregular Workloads

نویسندگان

  • Pedro Henrique Penna
  • Eduardo C. Inacio
  • Márcio Bastos Castro
  • Patricia Della Mea Plentz
  • Henrique C. Freitas
  • François Broquedis
  • Jean-François Méhaut
چکیده

The input workload of an irregular application must be evenly distributed among its threads to enable cutting-edge performance. To address this need in OpenMP, several loop scheduling strategies were proposed. While having this ever-increasing number of strategies at disposal is helpful, it has become a non-trivial task to select the best one for a particular application. Nevertheless, this challenge becomes easier to be tackled when existing scheduling strategies are extensively evaluated. Therefore, in this paper, we present a performance and scalability evaluation of the recently-proposed loop scheduling strategy named Smart Round-Robin (SRR). To deliver a comprehensive analysis, we coupled a kernel benchmarking technique with several rigorous statistical tools, and considered OpenMP’s Static and Dynamic loop schedulers as our baselines. Our results unveiled that SRR performs better on irregular applications with symmetric workloads and coarse-grained parallelization, achieving up to 1.9x and 1.5x speedup over OpenMP’s Static and Dynamic schedulers on synthetic kernels, respectively. On a N-Body Simulations application kernel, SRR delivered 2.48x better performance in contrast to OpenMP’s Dynamic scheduler. Key-words: Loop Scheduling, Irregular Workloads, Kernel Benchmarking, OpenMP, SRR. ∗ Federal University of Santa Cararina (UFSC) † Pontifical Catholic University of Minas Gerais (PUC Minas) ‡ Grenoble Institute of Technology (Grenoble INP) § University of Grenoble Alpes (UGA) Assessing the Performance of the SRR Loop Scheduler with Irregular Workloads Résumé : La charge de calcul d’une application irrégulière doit être distribuée équitablement entre les différents flots d’exécution (threads) pour obtenir un haut niveau de performance. Pour atteindre cet objectif avec OpenMP, plusieurs stratégies d’ordonnancement des boucles ont déjà été proposées dans la littérature. Pouvoir disposer de nombreuses stratégies d’ordonnancement est vraiment utile pour aider le programmeur à choisir la meilleure stratégie par rapport aux spécificités et caractéristiques de son application. Il est donc important de pouvoir évaluer les stratégies d’ordonnancement précisément.Dans ce rapport, nous présentons une analyse détaillée et rigoureuse de la stratégie SRR (Smart Round Robin) qui avait été proposée dans un précédent article. Cette analyse complète est basée sur le couplage de d’analyse comparative de noyaux avec des outils statistiques. La stratégie SRR a ainsi été comparée aux stratégies statique et dynamique qu’on retrouve dans OpenMP. Les résultats montrent, que sur des applications irrégulières (noyaux synthéqtiques) avec des charges de calcul symétriques et parallélisation grossière, SRR permet d’améliorer entre 1.5 et 1.9 les performances. Sur le noyau d’une application réelle (Simulation N-Body), SRR améliore les performances d’un facteur 2.48 par rapport à l’ordonnanceur dynamique OpenMP. Mots-clés : Ordonnancement de Boucle, Charges de Travail Irrégulières, Kernel Benchmarking, OpenMP, SRR. Assessing the Performance of the SRR Loop Scheduler 3

برای دانلود متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

منابع مشابه

Hierarchical smoothed round robin scheduling in high-speed networks

The hierarchical smoothed round robin (HSRR) packet scheduler that improves the performance of the smoothed round robin (SRR) for real-world packet scheduling scenarios is presented. The SRR suffers from an inability to fairly distribute service in highly skewed weight distributions where such distributions often occur in practice. As a result, the transmission behaviour of the SRR can be burst...

متن کامل

Near Optimal Work-Stealing Tree Scheduler for Highly Irregular Data-Parallel Workloads

We present a work-stealing algorithm for runtime scheduling of dataparallel operations in the context of shared-memory architectures on data sets with highly-irregular workloads that are not known a priori to the scheduler. This scheduler can parallelize loops and operations expressible with a parallel reduce or a parallel scan. The scheduler is based on the work-stealing tree data structure, w...

متن کامل

A Distributed QoS Scheduler for Smoothing Output Traffic of Input Buffered Switches

Abstract. To provide stringent service guarantees such as latency and backlog bounds for input-buffered switches, a set of scheduling algorithm and admission control strategy is proposed. This set of traffic control strategy is primarily based on a single-server scheduling algorithm called Smoothed Round Robin (SRR). SRR possesses a number of advantages which are very attractive to the implemen...

متن کامل

A reliability-based maintenance technicians’ workloads optimisation model with stochastic consideration

The growing interest in technicians’ workloads research is probably associated with the recent surge in competition. This was prompted by unprecedented technological development that triggers changes in customer tastes and preferences for industrial goods. In a quest for business improvement, this worldwide intense competition in industries has stimulated theories and practical frameworks that ...

متن کامل

Coexisting Scheduling Policies Boosting I/O Virtual Machines

Deploying multiple Virtual Machines (VMs) running various types of workloads on current many-core cloud computing infrastructures raises an important issue: The Virtual Machine Monitor (VMM) has to efficiently multiplex VM accesses to the hardware. We argue that altering the scheduling concept can optimize the system’s overall performance. Currently, the Xen VMM achieves near native performance...

متن کامل

ذخیره در منابع من


  با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید

برای دانلود متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

عنوان ژورنال:

دوره   شماره 

صفحات  -

تاریخ انتشار 2017