Disclaimer: This thesis was mainly supervised by Ing. Jiří Pavela; I have served as additional technical consultant. The goal of this thesis was to experimentally try to measure the consumption of the energy of profiled programs.
Disclaimer: This thesis was mainly supervised by Ing. Jiří Pavela; I have served as additional technical consultant. The goal of this thesis was to extend our profilers with profiler for programs written in C#, potentially extending our support for Windows (sub)systems. The results were interpreted using tree-like views similar to outputs of kcachegrind.
Disclaimer: This thesis was mainly supervised by Ing. Jiří Pavela; I have served as additional technical consultant. The goal of this thesis was to extend our trace profiler with support for PIN framework. This would allow us to overcome some of the limitations of our current underlying instrumentation frameworks (eBPF and SystemTap).
The goal of this thesis was to extend the capabilities of Perun with injecting noise into program to either simulate potential code optimisations or to make performance issues manifest during profiling.
Disclaimer: While this thesis was officially supervised by Adam Rogalewicz (due some internal restrictions), I was the main supervisor of this whole thesis. The goal of this work was to optimize the profiling time and space based on several aspects of projects or profiling process: based on structure of the profiled program, based on complexity of the analysed functions, or recency of the changes. This is without doubt, my most favourite work I have supervised.
Disclaimer: While this thesis was officially supervised by Adam Rogalewicz (due some internal restrictions), I was the main supervisor of this whole thesis. The goal of this thesis was to extend the capabilities of Perun with new models (besides so-far supported regression analysis) as well as new approaches to detect performance degradations in software. In particular, we come up with non-parametric models (models not dependent on other variables) such as regressograms, moving averages or kernel regressions; as well as methods for detecting performance issues based on integral computation or local statistics.
Disclaimer: This thesis was supervised by Tomáš Vojnar, I was the technical consultant. The goal of this thesis was build upon our work on automatic complexity analysis; this was initial work that only had to reproduce the results of Loopus tools invented by our colleagues in TU Vienna (Moritz Sinn and Florian Zuleger) in the Facebook Infer framework.
Disclaimer: While this thesis was officially supervised by Adam Rogalewicz (due some internal restrictions), I was the main supervisor of this whole thesis. The goal of this thesis was to extend the notion of performance fuzztesting to more fine-tuned (custom) rules and more performance-based evaluation. The resulting fuzzer is part of Perun and provides new ideas (sadly unpublished).
Disclaimer: This thesis was supervised together with RedHat company; I served mainly as a formal supervisor; technical consultants were Zdeněk Kraus and Otavio Rodolfo Piske. This thesis goal was to extend the capabilities of performance testing conducted in Red Hat company.
The goal of this thesis was to create a graphical user interface for my Perun project. The GUI was realized based on modern approaches to website development using vue.js framework. The GUI is, however, currently outdated and not merged in our upstream.
Disclaimer: This thesis was supervised together with RedHat company; I served mainly as a formal superviser; technical consultant were Stanislav Židek and Hubert Kario. The goal of the thesis was to extend the capabilities of testing attacks based on fuzztesting solved within the RedHat company.
The goal of this thesis was to create a memory profiler of C/C++. The results were interpretered using several kinds of visualizations: (1) heat maps (both in GUI and in CLI using ncurses library), and (2) flamegraphs.
The goal of this thesis was to create a time profiler of C/C++ programs together with tracking sizes of advanced data structures (such as linked lists or skip lists). The collected data was postprocessed using regression analysis to infer mathematical models (which modeled runtime of functions based on sizes of underlying structures). The results were visualized using scatter plots.