Ocena zmogljivosti RISC-V vektorskega procesorja za robno obdelavo senzorskih podatkov

Kučina, Bernard (author), Pilipović, Ratko (mentor), Machidon, Octavian Mihai (comentor)

V okviru diplomskega dela smo se želeli bolje spoznati z odprtokodno ukazno arhitekturo RISC-V ter raziskati njene možnosti za pospeševanje računskih operacij. Poseben poudarek smo namenili vektorski razširitvi RVV, ki omogoča izvajanje operacij v slogu SIMD in tako odpira nove možnosti za učinkovito obdelavo podatkov. Kot primer uporabe razširitve smo izbrali implementacijo algoritma za izračun matrike GLCM in z njo povezanih značilk. Algoritem smo prilagodili za delovanje na procesorju s podporo za razširitev RVV ter ga eksperimentalno ovrednotili. Pokazali smo, da vektorska razširitev omogoča opazno pohitritev izračuna v primerjavi z osnovnim zaporednim izvajanjem, pri tem pa ohranja natančnost rezultatov in ustreznost za nadaljnjo uporabo v analizi slik.

• Language: Slovenian
• Keywords: GLCM, SIMD, RISC-V vektorski procesor
• Work type: Bachelor thesis/paper
• Typology: 2.11 - Undergraduate Thesis
• Organization: FRI - Faculty of Computer and Information Science
• Year: 2025
• PID: 20.500.12556/RUL-172692
• COBISS.SI-ID: 249559811
• Publication date in RUL: 10.09.2025

Secondary language

• Language: English
• Title: Performance evaluation of a RISC-V vector processor for edge processing of Sensor Data

Abstract

In this thesis, we explored the open-source RISC-V Vector Extension (RVV), which enables SIMD-style operations and opens up new opportunities for more efficient data processing. As a case study, we implemented an algorithm for calculating the GLCM matrix and its associated features. The algorithm was adapted to run on a processor with RVV support and experimentally evaluated. The results demonstrated that the use of the vector extension provides a significant acceleration of computation compared to the basic sequential approach, while maintaining both the accuracy of the results and their applicability in further image analysis.

• Keywords: GLCM, SIMD, RISC-V vector processor