Πλοήγηση ανά Συγγραφέας "Leivadaros, Svoronos"

Τώρα δείχνει 1 - 1 of 1
  • Μικρογραφία εικόνας
    Τεκμήριο
    Design and implementation of an adaptive high efficiency FPGA-accelerated system for multi-disciplinary application domains.
    (ΕΛ.ΜΕ.ΠΑ., Σχολή Μηχανικών (ΣΜΗΧ), ΠΜΣ Πληροφορική και Πολυμέσα, 2021-04-07) Leivadaros, Svoronos; Λειβαδάρος, Σβορώνος
    The scope of this thesis is the design and implementation of an FPGA platform that utilizes state-of-the-art techniques and methodologies to allow improved energy efficiency and performance in carrying out computationally intensive tasks. The goal is to develop a framework for FPGA-based architectures that can be used in environments that include but are not limited to Cloud Computing Clusters, High Performance Computational Clusters and Distributed Data Centers. A proof of concept implementation of this framework with 3 accelerated tasks (Black and White Image Thresholding, Image Convolution with a 3x3 Kernel, Genome 2-mer Distribution Analysis) is also presented and compared with implementation of the same tasks on an FPGA and ARMv7 architecture CPU. To our knowledge, the methodology in designing the Partial Reconfiguration platform employed in this work is novel and allows designing Dynamic Partial Reconfiguration-enabled hardware platforms on an FPGA without the need for wrapper logic or the need to register all inputs and outputs to every reconfigurable module, facilitating the establishment of compatibility across implemented reconfigurable modules in early development and future integration of new accelerated functions on the FPGA platform. Performance and energy efficiency metrics are also presented for the 2 different implementation platforms.