Optimalizace vědeckých algoritmů pro paralelní zpracování a vysoce-výkoné počítání
| Název práce v češtině: | Optimalizace vědeckých algoritmů pro paralelní zpracování a vysoce-výkoné počítání |
|---|---|
| Název v anglickém jazyce: | Optimizing scientific algorithms for parallel processing and high-performance computing |
| Klíčová slova: | paralelní|GPU|HPC|vědecké|výpočty |
| Klíčová slova anglicky: | parallel|GPU|HPC|scientific|computing |
| Akademický rok vypsání: | 2024/2025 |
| Typ práce: | disertační práce |
| Jazyk práce: | |
| Ústav: | Katedra distribuovaných a spolehlivých systémů (32-KDSS) |
| Vedoucí / školitel: | doc. RNDr. Martin Kruliš, Ph.D. |
| Řešitel: | skrytý - zadáno a potvrzeno stud. odd. |
| Datum přihlášení: | 15.02.2025 |
| Datum zadání: | 17.02.2025 |
| Datum potvrzení stud. oddělením: | 17.02.2025 |
| Zásady pro vypracování |
| In the past two decades, mainstream hardware has experienced a significant shift towards parallelism. Multicore CPUs, as well as manycore GPUs, are present in commodity PCs, servers, and HPC clusters. Unfortunately, most applications and algorithms of the day are not ready to fully utilize parallel hardware to its full potential.
In scientific computing, this problem is perhaps even more pronounced as the scientists from domains like physics, biology, or pharmacy focus have limited experience in computer science, especially in tasks as complex and advanced as code optimizations or HPC parallel computing. The main objective of this thesis is to study and improve algorithms and methods used in empirical sciences (such as iterative stencils, simulations, numeric computations, or rudimentary machine learning). It will focus mainly on accelerating these methods on mainstream hardware platforms, such as x86 CPUs and CUDA-enabled GPUs, but it will also include basic (serial) code optimizations, vectorization, or distributed computing (OpenMPI). In addition, it will investigate the possibilities of using emerging AI technologies (especially LLM models) to assist with the optimization and parallelization process. The outlined research will improve the knowledge of methods for designing parallel applications in the domain of scientific computing. This should help scientist in other domains design their HPC software more easily and with better optimizations. Furthermore, software artifacts produced during the research should be directly applicable to scientific research. |
| Seznam odborné literatury |
| - |