|
|
Statická analýza v nástroji Facebook Infer zaměřená na detekci porušení atomičnosti
Harmim, Dominik ; Smrčka, Aleš (oponent) ; Vojnar, Tomáš (vedoucí práce)
Cílem této práce je navrhnout statický analyzátor, který bude sloužit pro detekci porušení atomicity. Navržený analyzátor Atomer je implementován jako modul pro Facebook Infer, což je volně šířený a snadno rozšířitelný nástroj, který umožňuje efektivní modulární a inkrementální analýzu. Analyzátor pracuje na úrovni sekvencí volání funkcí. Navržené řešení je založeno na předpokladu, že sekvence, které jsou zavolány atomicky jednou, by měly být pravděpodobně volány atomicky vždy. Implementovaný analyzátor byl úspěšně ověřen a vyhodnocen jak na malých programech, vytvořených pro testovací účely, tak na veřejně dostupných testovacích programech, které vznikly ze skutečných nízkoúrovňových programů.
|
|
|
Application of Genetic Algorithms and Data Mining in Noise-based Testing of Concurrent Software
Šimková, Hana ; Kofroň, Jan (oponent) ; Lourenco, Joao (oponent) ; Vojnar, Tomáš (vedoucí práce)
This thesis proposes an improvement of the efficiency of testing concurrent software by employing data mining techniques and genetic algorithms in the process of testing concurrent software. Concurrent, or multi-threaded, programming has become very popular over the last few years. However, as the concurrent programming is far more demanding the sequential programming, its increased use leads to a significant increase in the number of errors that appear in commercial software due to errors in synchronization. Finding such errors using traditional testing methods is difficult. Moreover, repeated test executions of traditional testing that are performed in the same environment will typically examine similar interleavings only. Hence, the noise-based injection approach is used for influencing the scheduling by injecting various kinds of noise (delays, context switches, and so on) into the common thread behaviour which stress the software and can to show some rare behaviour. However, for the noise injection to be efficient, one has to choose suitable noise injection heuristics from among the many existing ones as well as to suitably choose values of their various parameters, which is not easy. In this work, there are used data mining methods and genetic algorithms and their combinations to deal with the problem of choosing such noise injection heuristics and values of their parameters. Besides setting up of the goals of the thesis, this proposal also provides a brief summary of the state of the art in application of data mining techniques and genetic algorithms to program testing problems.
|
|
|
Advanced Static Analysis of Atomicity in Concurrent Programs through Facebook Infer
Harmim, Dominik ; Rogalewicz, Adam (oponent) ; Vojnar, Tomáš (vedoucí práce)
Atomer is a static analyser based on the idea that if some sequences of functions of a multi-threaded program are executed under locks in some runs, likely, they are always intended to execute atomically. Atomer thus strives to look for such sequences and then detects for which of them the atomicity may be broken in some other program runs. The author of this master's thesis proposed and implemented the first version of Atomer as a plugin of the Facebook Infer framework in his bachelor's thesis. In the master's thesis, a new and significantly improved version of Atomer is proposed. The improvements aim at both increasing scalability as well as precision. Moreover, support for several initially not supported programming features has been added (including, e.g., the possibility of analysing C++ and Java programs or support for re-entrant locks or lock guards). Through a number of experiments (including experiments with real-life code and real-life bugs), it is shown that the new version of Atomer is indeed much more general, scalable, and precise.
|
|
|
Statická analýza v nástroji Meta Infer zaměřená na detekci souběhu nad daty
Svobodová, Lucie ; Fiedor, Jan (oponent) ; Vojnar, Tomáš (vedoucí práce)
Modern software systems often use concurrent programs to improve performance and increase efficiency. However, ensuring the reliability and safety of such systems can be challenging due to the increased potential for bugs, including data races, to arise. In this thesis, we introduce a new static data race detector, DarC, designed for programs written in C using the Pthreads library. The proposed detector is implemented as an analyser plugin in Meta Infer, a static analysis framework with an emphasis on compositional, incremental, and consequently highly-scalable analysis. Our approach involves recording a set of accesses that occur in the analysed program along with information about the set of locks held during these accesses. The tool then identifies pairs of accesses that may lead to data races and reports them to the user. Our tool was successfully evaluated on a set of benchmarking programs as well as on real-life projects, showing its potential for effectively detecting data races in C programs.
|
|
|
Advanced Static Analysis of Atomicity in Concurrent Programs through Facebook Infer
Harmim, Dominik ; Rogalewicz, Adam (oponent) ; Vojnar, Tomáš (vedoucí práce)
Atomer is a static analyser based on the idea that if some sequences of functions of a multi-threaded program are executed under locks in some runs, likely, they are always intended to execute atomically. Atomer thus strives to look for such sequences and then detects for which of them the atomicity may be broken in some other program runs. The author of this master's thesis proposed and implemented the first version of Atomer as a plugin of the Facebook Infer framework in his bachelor's thesis. In the master's thesis, a new and significantly improved version of Atomer is proposed. The improvements aim at both increasing scalability as well as precision. Moreover, support for several initially not supported programming features has been added (including, e.g., the possibility of analysing C++ and Java programs or support for re-entrant locks or lock guards). Through a number of experiments (including experiments with real-life code and real-life bugs), it is shown that the new version of Atomer is indeed much more general, scalable, and precise.
|
|
|
Application of Genetic Algorithms and Data Mining in Noise-based Testing of Concurrent Software
Šimková, Hana ; Kofroň, Jan (oponent) ; Lourenco, Joao (oponent) ; Vojnar, Tomáš (vedoucí práce)
This thesis proposes an improvement of the efficiency of testing concurrent software by employing data mining techniques and genetic algorithms in the process of testing concurrent software. Concurrent, or multi-threaded, programming has become very popular over the last few years. However, as the concurrent programming is far more demanding the sequential programming, its increased use leads to a significant increase in the number of errors that appear in commercial software due to errors in synchronization. Finding such errors using traditional testing methods is difficult. Moreover, repeated test executions of traditional testing that are performed in the same environment will typically examine similar interleavings only. Hence, the noise-based injection approach is used for influencing the scheduling by injecting various kinds of noise (delays, context switches, and so on) into the common thread behaviour which stress the software and can to show some rare behaviour. However, for the noise injection to be efficient, one has to choose suitable noise injection heuristics from among the many existing ones as well as to suitably choose values of their various parameters, which is not easy. In this work, there are used data mining methods and genetic algorithms and their combinations to deal with the problem of choosing such noise injection heuristics and values of their parameters. Besides setting up of the goals of the thesis, this proposal also provides a brief summary of the state of the art in application of data mining techniques and genetic algorithms to program testing problems.
|
|
|
Statická analýza v nástroji Facebook Infer zaměřená na detekci porušení atomičnosti
Harmim, Dominik ; Smrčka, Aleš (oponent) ; Vojnar, Tomáš (vedoucí práce)
Cílem této práce je navrhnout statický analyzátor, který bude sloužit pro detekci porušení atomicity. Navržený analyzátor Atomer je implementován jako modul pro Facebook Infer, což je volně šířený a snadno rozšířitelný nástroj, který umožňuje efektivní modulární a inkrementální analýzu. Analyzátor pracuje na úrovni sekvencí volání funkcí. Navržené řešení je založeno na předpokladu, že sekvence, které jsou zavolány atomicky jednou, by měly být pravděpodobně volány atomicky vždy. Implementovaný analyzátor byl úspěšně ověřen a vyhodnocen jak na malých programech, vytvořených pro testovací účely, tak na veřejně dostupných testovacích programech, které vznikly ze skutečných nízkoúrovňových programů.
|
|
|
Vývoj vláknových aplikací v jazyce Java
ATTL, Karel
Tato diplomová práce pojednává o programování vícevláknových aplikací v jazyce Java. S verzí Java 5 se v API jazyka objevuje i knihovna java.util.concurrent, která významným způsobem ulehčuje a zefektivňuje návrh paralelních aplikací. Práce je pojatá jako úvod do programování vícevláknových aplikací, a zároveň ji lze využít jako studijní materiál. Teoretický úvod pojednává o procesech a technologickému pozadí multitaskingu jako analogie k vláknům, zároveň se dotýká technologie Java a pozadí práce s pamětí. Zbytek diplomové práce už se věnuje praktickému nastínění práce s vlákny. Pokrývá toto téma od úplných základů, jakými je vytvoření objektu typu Thread, přes pokročilejší úlohy, jakými je práce s balíčkem java.util.concurrent a na závěr věnuje prostor problémům, se kterými se programátor může setkat při vývoji vícevláknových aplikací.
|
host ::
RSS.