Thesis supervisor: Zoltán Juhász
Location of studies (in Hungarian): University of Pannonia H-8200 Veszprém, Egyetem u. 10. Abbreviation of location of studies: PE
Description of the research topic:
Programmable graphics processors have changed today’s high-performance computing landscape. These chips contain thousands of cores and deliver several Tflops computational power but require a degree of parallelism in the order of tens of thousands. With careful algorithm design that takes hardware execution characteristics into account, the execution times can be drastically reduced. The challenge is to design algorithms that match these architectures ideally and implement them in way that produces maximum execution efficiency. The goal of this research is to study a selected set of important and hard problems in key application areas (such as simulation, numerical methods, advanced signal processing, imaging, medical data processing) in order to develop best-in-class implementations. The work will involve algorithm theory and design, GPU architecture research, GPU programming, numerical algorithm analysis and the study of potential application areas that could benefit from the results.
Preliminary results can be found in the following publications:
1. Couturier, Raphaël, ed. Designing scientific applications on gpus. CRC Press, 2013.
2. Micikevicius, Paulius. "GPU performance analysis and optimization." In GPU technology conference, vol. 84. 2012.
Number of students who can be accepted: 1
Deadline for application: 2020-09-30
2024. IV. 17. ODT ülés Az ODT következő ülésére 2024. június 14-én, pénteken 10.00 órakor kerül sor a Semmelweis Egyetem Szenátusi termében (Bp. Üllői út 26. I. emelet).
Login
