Institutional Repository
Technical University of Crete
Technical University of Crete
EN
|
EL
| URI: | http://purl.tuc.gr/dl/dias/9541CD3E-E6F4-48D1-A325-8B380EAD9C4B | ||
| Year | 2024 | ||
| Type of Item | Diploma Work | ||
| License |
|
||
| Bibliographic Citation | Alexandros-Andreas Stavropoulos, "Implementation of complex signal processing applications using high level synthesis tools on a VERSAL platform", Diploma Work, School of Electrical and Computer Engineering, Technical University of Crete, Chania, Greece, 2024 https://doi.org/10.26233/heallink.tuc.101129 | ||
| Appears in Collections |
The growing demand for efficient and scalable hardware architectures to support complex machine learning (ML) and signal processing applications has led to the exploration of versatile platforms like the AMD/Xilinx Versal Adaptive Compute Acceleration Platform (ACAP). This thesis investigates the performance of the Versal VCK190, focusing on its Network-on-Chip (NoC) architecture and its potential for accelerating convolutional neural networks (CNNs) and matrix multiplication tasks. By leveraging the Vitis and Vitis AI development environments, benchmarks for CNN models and matrix operations were deployed to evaluate the throughput and efficiency of the NoC. The results demonstrate the strengths of the Versal platform in managing highperformance AI workloads, while also revealing areas where optimizations could improve performance, particularly in real-time applications. Key contributions of this work include the development of custom benchmarks for CNNs and matrix multiplication, a thorough evaluation of NoC throughput and performance analysis based on real-time AI inference. The findings offer valuable insights into the capabilities and limitations of the Versal architecture for heterogeneous computing applications.
Copyright © DIAS 2013
