Empirical study of Amdahl’s law on multicore processors

Carsten Bruns 1, 2, 3 Sid Touati 1, 3, 2
1 KAIROS - Logical Time for Formal Embedded System Design
CRISAM - Inria Sophia Antipolis - Méditerranée , Laboratoire I3S - COMRED - COMmunications, Réseaux, systèmes Embarqués et Distribués
Abstract : Since many years, we observe a shift from classical multiprocessor systems tomulticores, which tightly integrate multiple CPU cores on a single die or package. This shift doesnot modify the fundamentals of parallel programming, but makes harder the understanding andthe tuning of the performances of parallel applications. Multicores technology leads to sharing ofmicroarchitectural resources between the individual cores, which Abel et al. [1] classified in storageand bandwidth resources. In this work, we empirically analyze the effects of such sharing onprogram performance, through repeatable experiments. We show that they can dominate scalingbehavior, besides the effects described by Amdahl’s law and synchronization or communicationconsiderations. In addition to the classification of [1], we view the physical temperature and powerbudget also as a shared resource. It is a very important factor for performance nowadays, sinceDVFS over a wide range is needed to meet these constraints in multicores. Furthermore, wedemonstrate that resource sharing not just leads a flat speedup curve with increasing thread countbut can even cause slowdowns. Last, we propose a formal modeling of the performances to allowdeeper analysis. Our work aims to gain a better understanding of performance limiting factors inhigh performance multicores, it shall serve as basis to avoid them and to find solutions to tune theparallel applications.
Complete list of metadatas

Cited literature [73 references]  Display  Hide  Download

https://hal.inria.fr/hal-02404346
Contributor : Sid Touati <>
Submitted on : Wednesday, December 11, 2019 - 12:09:59 PM
Last modification on : Monday, January 13, 2020 - 1:18:49 AM

File

RR-9311 (1).pdf
Files produced by the author(s)

Identifiers

  • HAL Id : hal-02404346, version 1

Citation

Carsten Bruns, Sid Touati. Empirical study of Amdahl’s law on multicore processors. [Research Report] RR-9311, INRIA Sophia-Antipolis Méditerranée; Université Côte d'Azur, CNRS, I3S, France. 2019. ⟨hal-02404346⟩

Share

Metrics

Record views

60

Files downloads

173