The Speedup Test

نویسندگان

  • Sebastien BRIAIS
  • Sebastien Briais
چکیده

Numerous code optimisation methods are usually experimented by doing multiple observations of the initial and the optimised executions times in order to declare a speedup. Even with fixed input and execution environment, programs executions times vary in general. So hence different kinds of speedups may be reported: the speedup of the average execution time, the speedup of the minimal execution time, the speedup of the median, etc. Many published speedups in the literature are observations of a set of experiments. In order to improve the reproducibility of the experimental results, this technical report presents a rigorous statistical methodology regarding program performance analysis. We rely on well known statistical tests (Shapiro-wilk’s test, Fisher’s F-test, Student’s t-test, Kolmogorov-Smirnov’s test, Wilcoxon-Mann-Whitney’s test) to study if the observed speedups are statistically significant or not. By fixing 0 < α < 1 a desired risk level, we are able to analyse the statistical significance of the average execution time as well as the median. We can also check if P [X > Y ] > 12 , the probability that an individual execution of the optimised code is faster than the individual execution of the initial code. Our methodology defines a consistent improvement compared to the usual performance analysis method in high performance computing as in [15, 11]. We explain in each situation what are the hypothesis that must be checked to declare a correct risk level for the statistics. The Speedup-Test protocol certifying the observed speedups with rigorous statistics is implemented and distributed as an open source tool based on R software. Key-words: Code optimisation, program performance evaluation and analysis, statistics ∗ INRIA-Saclay. [email protected] † Laboratoire de mathématiques de Versailles. [email protected] ‡ Laboratoire PRiSM. [email protected] in ria -0 04 43 83 9, v er si on 2 26 J an 2 01 0 Un protocole statistique pour l’analyse de l’accélération des programmes Résumé : De nombreuses techniqus d’optimisations de programmes sont expérimentées en mesurant plusieurs fois les temps d’exécutions du code initial et du code transformé. Même en fixant les données d’entrés et l’environment d’exécution, les temps observés pour les exécutions des programmes sont variables en général. Ainsi, plusieurs facteurs d’accélérations possibles peuvent être observés: accélération du temps minimum, accélération du temps moyen et accélération du temps médian. Ces observations ne sont pas toujours significatives statistiquement. Afin d’améliorer la reproducibilité des performances des programmes, nous présentons dans ce document une méthodologie statistique rigoureuse basée sur plusieurs tests connus (test de Shapiro-wilk, test F de Fisher, test de Student, test de Kolmogorov-Smirnov, test de Wilcoxon-Mann-Whitney’s). En se fixant un niveau de risque α souhaité, nous somme capabls de comparer entre deux moyennes ou deux médianes variables. Notre méthodologie définit une amélioration par rapport aux protocols usuels décrits dans [11, 15]. Par ailleurs, nous expliquons dans chaque situation d’observation d’accélération quelles sont les hypothèses à vérifier pour déclarer un niveau de risque correct. Le protocole statistique, appelé le Speedup-Test, certifiant que les accélérations observées sont statistiquement valides est distribué sous forme de logiciel libre basé sur R. Mots-clés : Optimisation de code, analyse et évaluation des performances des programmes, statistiques in ria -0 04 43 83 9, v er si on 2 26 J an 2 01 0 The Speedup-Test 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

Sign Effect, Speedup – Delay Asymmetry and Gender Effect In the Tehran Stock Exchange

The present study first investigates the asymmetry of speedup - delay (gain and loss) and Sign Effect in Tehran Stock Exchange, then examine the effect of gender on the discount rate and explain it with the loss aversion. The sample is 403 investors in Tehran Stock Exchange, and the nonparametric Tests are used to test the research hypotheses. The research results show that there is an asymmetr...

متن کامل

An approach to Improve Particle Swarm Optimization Algorithm Using CUDA

The time consumption in solving computationally heavy problems has always been a concern for computer programmers. Due to simplicity of its implementation, the PSO (Particle Swarm Optimization) is a suitable meta-heuristic algorithm for solving computationally heavy problems. However, despite the simplicity, the algorithm is inefficient for solving real computationally heavy problems but the pr...

متن کامل

Speeding up the Stress Analysis of Hollow Circular FGM Cylinders by Parallel Finite Element Method

In this article, a parallel computer program is implemented, based on Finite Element Method, to speed up the analysis of hollow circular cylinders, made from Functionally Graded Materials (FGMs). FGMs are inhomogeneous materials, which their composition gradually varies over volume. In parallel processing, an algorithm is first divided to independent tasks, which may use individual or shared da...

متن کامل

The Speedup-Test: a statistical methodology for programme speedup analysis and computation

In the area of high performance computing and embedded systems, numerous code optimisation methods exist to accelerate the speed of the computation (or optimise another performance criteria). They are usually experimented by doing multiple observations of the initial and the optimised execution times of a program in order to declare a speedup. Even with fixed input and execution environment, pr...

متن کامل

A generalized implicit enumeration algorithm for a class of integer nonlinear programming problems

Presented here is a generalization of the implicit enumeration algorithm that can be applied when the objec-tive function is being maximized and can be rewritten as the difference of two non-decreasing functions. Also developed is a computational algorithm, named linear speedup, to use whatever explicit linear constraints are present to speedup the search for a solution. The method is easy to u...

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 2010