National Technical University of Athens (NTUA)
Σχεδιασμός αρχιτεκτονικών και απεικόνιση εφαρμογών σε επαναδιαμορφούμενες πλατφόρμες με εργαλεία λογισμικού
Abstract
dc:descriptionIn recent years, reconfigurable architectures and more specifically Field Programmable Gate Arrays (FPGAs) have become efficient alternatives to Application Specific Integrated Circuits (ASICs). The key to FPGAs’ popularity is their feature to support application implementation by appropriately (re-)configuring the functionality of hardware resources. This allows FPGAs to provide higher flexibility, rapid product prototyping and significantly reduced non-recurring engineering (NRE) costs, as compared to ASIC (Application-Specific Integrated Circuit) devices.The features and capabilities of these reconfigurable architectures have changed and improved significantly over the last two decades. From simple arrays of Look-Up tables (LUT), we have now reached heterogeneous devices incorporating a series of hardware components (e.g. LUTs with different sizes, microprocessors, DSP and RAM blocks etc.). The logical structure of an FPGA has changed gradually from a homogeneous and regular architecture towards a heterogeneous System on Chip (SoC) device. Furthermore the complexity of current applications usually introduces restrictions on the architectural organization of the FPGA. Even if the demand for additional logical resources is met with platforms consisting of more complex logic blocks, or CLBs, (e.g. with more LUTs), the problem persists with communication intensive applications (e.g. telecommunications, encryption and image processing, video) since their performance is typically dependent on the availability of I/O bandwidth.This Doctoral thesis investigates the challenges and presents novel solutions on the field of application mapping onto Field Programmable Reconfigurable Arrays. The goal is to analyze and profile the obstacles and the constrains that limit the efficiency of application mapping and propose novel solutions to alleviate them. For this reason a novel software-supported methodology is introduced for enabling rapid architecture-level exploration for heterogeneous FPGAs that consist of different memory organizations and/ or hierarchies. Additionally a new tool framework was developed that enables application mapping onto these heterogeneous FPGAs. The proposed framework enables among others simultaneous handling of heterogeneous blocks with different hierarchies. In this thesis, in order to alleviate the communication bottleneck of complex applications and towards higher performance a novel architecture template was proposed for designing heterogeneous 3-D FPGAs with layers that contain blocks of different type. Added to the known challenges in the physical layer created by transistor shrinking, is the increased complexity of the target circuits and FPGAs, which makes the effectiveness and efficiency of the employed CAD tools become even more important. Techniques that accelerate core CAD algorithms can bring about important changes in product design times for these applications, whereas many designers may be willingvi to trade-off some quality of the solution for an improved run-time of the CAD tools. In order to integrate efficiently on this new landscape, FPGAs need to support fast application development and implementation. Industry has taken steps towards faster application development, exploring diverse solutions. FPGAs have been explored as a viable platform for various high performance applications and various embedded computing domains mainly due to the inherent parallelism and reprogrammability feature which can be applied either at design or run time.In order to address these limitations in this thesis, a novel methodology is introduced that targets to perform fast application’s implementation onto FPGAs. The goal of this approach is to reduce considerably the run-time overhead with the minimum possible performance degradation. Towards the same goals a cloud-inspired methodology and the supporting framework are proposed in order to enable the efficient mapping of multiple designs at runtime onto a single or multiple FPGAs. The introduced solution alleviates the previously mentioned constrains by offering fast execution times and enabling portability and scalability.In order to dynamically utilize FPGAs on run time we propose a novel methodology and the supporting toolflow that enable efficient mapping of multiple applications onto heterogeneous FPGAs. With the use of dynamic virtual kernels, memory optimizations and custom memory allocators, we alleviate the constrains imposed by CAD tools, and provide a proof of concept that application mapping onto FPGAs can be done on run time even on embedded systems.
Degree
thesis:*- Grantor dc:publisher
- National Technical University of Athens (NTUA)
- Year dc:date
- 2016
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Sidiropoulos, Charalampos
- Σιδηρόπουλος, Χαράλαμπος
Subjects
dc:subject × 13- FPGAs
- Επαναδιαμορφούμενες αρχιτεκτονικές
- CAD Εργαλεία
- Aλγόριθμοι
- Τρισδιάστατες αρχιτέκτονες
- Reconfigurable Architectures
- Algorithms
- CAD Tools
- 3D Architectures
- Επιστήμες Μηχανικού και Τεχνολογία
- Επιστήμη Ηλεκτρολόγου Μηχανικού, Ηλεκτρονικού Μηχανικού, Μηχανικού Η/Υ
- Engineering and Technology
- Electrical Engineering, Electronic Engineering, Information Engineering
Rights
dc:rights- Statement dc:rights
-
- BY_SA
- Language dc:language
- gre
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/38607