University of Patras
Μεθοδολογίες μεταγλώττισης σε επαναπροσδιοριζόμενα συστήματα αρχιτεκτονικών πίνακα.
Abstract
dc:descriptionThe object of this PhD thesis focuses on developing efficient mapping techniquesfor coarse grain reconfigurable build arrays. Data intensive applications were used toevaluate the proposed methodologies. The aim is to optimize the applications’performance on characteristics targeting reconfigurable characteristics such asperformance, instructions per cycle, area of integration and processing resourceutilization. This is achieved by introducing novel mapping techniques and findingoptimal architectures.In the first part of the thesis research, development and automation of mappingtechniques was carried out targeting coarse grain reconfigurable arrays. The mainfeature of these architectures is the presence of a large number of processing elementsworking in parallel thus speeding up the execution of applications featuring paralleloperations. The function of these processing elements in embedded systems resemblesthat of a coprocessor.The research on reconfigurable array architectures has gained considerable interestbecause of their flexibility, scalability and performance, particularly in data intensiveapplications. Nevertheless, compiling these applications on reconfigurable architecturesis characterized by high degree of complexity. Appropriate tools and special mappingmethodologies are needed to exploit the characteristics of these architectures. Bearingthis in mind, we proposed a novel reconfigurable methodology for mappingapplications, which has also been automated with the use of a prototype compiler toolaiming at a parametric architectural model. The result was finding the best architectureson the basis of performance, the instructions per cycle term and the tool execution timefor a sample set of applications.It is difficult to evaluate the efficiency of a reconfigurable array architecture table interms of speed and area of integration, so there have been few cases studying the effectof architectural parameters on factors such as surface integration and the number ofinstructions per clock cycle. Moreover, no work has examined the multipliers’ impact embedded in reconfigurable architectures processing elements. Using the existingreconfigurable mapping methodology and a parametric implementation of thearchitecture in hardware description language, we examine the effect of multipliers onthe part of the mapping phase and architecture.We also describe an original mapping methodology introduced for the purpose ofefficiently mapping the Fast Fourier Transform (FFT) algorithm on reconfigurable arrayarchitectures. The FFT algorithm is characterized by a large number of operationsprimarily multiplications that slow the performance of a reconfigurable architecture.Exploiting the existence of an internal structure inside the FFT algorithm and by the useof a reconfigurable architecture template of 16 processing elements, we developed anovel mapping technique. Additionally, our technique takes into account the memoryhierarchy between main memory and reconfigurable architecture in order to furtheraccelerate the implementation of the FFT algorithm. Using the proposed mappingtechnique results in processing elements utilization of over 90% value which is at least37% better than the best value of the related literature.
Degree
thesis:*- Grantor dc:publisher
- University of Patras
- Year dc:date
- 2012
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Georgiopoulos, Stavros
- Γεωργιόπουλος, Σταύρος
Subjects
dc:subject × 26- Επαναπροσδιοριζόμενος
- Ενσωματωμένος
- Χαρτογράφηση
- Συνεπεξεργαστής
- Πολλαπλασιαστής
- Ιεραρχία μνήμης
- Επαναχρησιμοποίηση
- Αρχιτεκτονική
- ΠΙΝΑΚΑΣ
- Μεταγλωττιστής
- Reconfigurable
- Embedded
- Mapping
- Coprocessor
- Multipliers
- Memory hierarchy
- Reuse
- Architecture
- Array
- Compilers
- Επιστήμη Ηλεκτρολόγου Μηχανικού, Ηλεκτρονικού Μηχανικού, Μηχανικού Η/Υ
- Επιστήμες Μηχανικού και Τεχνολογία
- Υπολογιστές, Υλικό (hardware) και Αρχιτεκτονική
- Electrical Engineering, Electronic Engineering, Information Engineering
- Engineering and Technology
- Computer science, Hardware and Architecture
Rights
- Language dc:language
- gre
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/30331