Fast Detection of Concurrency Errors by State Space Traversal with Randomization and Early Backtracking

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F00216208%3A11320%2F19%3A10387411" target="_blank" >RIV/00216208:11320/19:10387411 - isvavai.cz</a>

Výsledek na webu

<a href="https://verso.is.cuni.cz/pub/verso.fpl?fname=obd_publikace_handle&handle=YtjLTyM86p" target="_blank" >https://verso.is.cuni.cz/pub/verso.fpl?fname=obd_publikace_handle&handle=YtjLTyM86p</a>

DOI - Digital Object Identifier

<a href="http://dx.doi.org/10.1007/s10009-018-0484-7" target="_blank" >10.1007/s10009-018-0484-7</a>

Jazyk výsledku

angličtina

Název v původním jazyce

Fast Detection of Concurrency Errors by State Space Traversal with Randomization and Early Backtracking

Popis výsledku v původním jazyce

State space traversal is a very popular approach to detect concurrency errors and test concurrent programs. However, it is not practically feasible for complex programs with many thread interleavings and a large state space. Many techniques explore only a part of the state space in order to find errors quickly --- building upon the observation that errors can often be found in a particular small part of the state space. Great improvements of performance have been achieved also through randomization. In the context of this research direction, we present the DFS-RB algorithm that augments the standard algorithm for depth-first traversal with early backtracking. Specifically, it is possible to backtrack early from a state before all outgoing transitions have been explored. The DFS-RB algorithm is non-deterministic --- it uses random numbers, together with values of several parameters, to determine when and how early backtracking takes place in the search. To evaluate DFS-RB, we performed a large experimental study with our prototype implementation in Java Pathfinder on several Java programs. The results show that DFS-RB achieves better performance in terms of speed and error detection than many state-of-the-art techniques for many benchmarks in our set. Nevertheless, it is difficult to find a single configuration of DFS-RB that works well for many different benchmarks. We designed a ranking algorithm whose purpose is to identify configurations that yield overall consistently good performance with a small variation.

Název v anglickém jazyce

Fast Detection of Concurrency Errors by State Space Traversal with Randomization and Early Backtracking

Popis výsledku anglicky

State space traversal is a very popular approach to detect concurrency errors and test concurrent programs. However, it is not practically feasible for complex programs with many thread interleavings and a large state space. Many techniques explore only a part of the state space in order to find errors quickly --- building upon the observation that errors can often be found in a particular small part of the state space. Great improvements of performance have been achieved also through randomization. In the context of this research direction, we present the DFS-RB algorithm that augments the standard algorithm for depth-first traversal with early backtracking. Specifically, it is possible to backtrack early from a state before all outgoing transitions have been explored. The DFS-RB algorithm is non-deterministic --- it uses random numbers, together with values of several parameters, to determine when and how early backtracking takes place in the search. To evaluate DFS-RB, we performed a large experimental study with our prototype implementation in Java Pathfinder on several Java programs. The results show that DFS-RB achieves better performance in terms of speed and error detection than many state-of-the-art techniques for many benchmarks in our set. Nevertheless, it is difficult to find a single configuration of DFS-RB that works well for many different benchmarks. We designed a ranking algorithm whose purpose is to identify configurations that yield overall consistently good performance with a small variation.

Druh

J_imp - Článek v periodiku v databázi Web of Science

CEP obor

—

OECD FORD obor

10201 - Computer sciences, information science, bioinformathics (hardware development to be 2.2, social aspect to be 5.8)

Projekt

<a href="/cs/project/GA18-17403S" target="_blank" >GA18-17403S: Automatická inkrementální verifikace a odstraňování chyb pro souběžné systémy</a><br>

Návaznosti

P - Projekt vyzkumu a vyvoje financovany z verejnych zdroju (s odkazem do CEP)

Rok uplatnění

2019

Kód důvěrnosti údajů

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Název periodika

International Journal on Software Tools for Technology Transfer

ISSN

1433-2779

e-ISSN

—

Svazek periodika

21

Číslo periodika v rámci svazku

4

Stát vydavatele periodika

DE - Spolková republika Německo

Počet stran výsledku

36

Strana od-do

365-400

Kód UT WoS článku

000476539900001

EID výsledku v databázi Scopus

2-s2.0-85041223636