Barone, Salvatore (2021) Designing Efficient Computing Systems: the Approximate-Computing Breakthrough. [Tesi di dottorato]
Anteprima |
Testo
Barone_Salvatore.pdf Download (4MB) | Anteprima |
| Tipologia del documento: | Tesi di dottorato |
|---|---|
| Lingua: | English |
| Titolo: | Designing Efficient Computing Systems: the Approximate-Computing Breakthrough |
| Autori: | Autore Email Barone, Salvatore salvatore.barone@unina.it |
| Data: | 2021 |
| Istituzione: | Università degli Studi di Napoli Federico II |
| Dipartimento: | Ingegneria Elettrica e delle Tecnologie dell'Informazione |
| Dottorato: | Information technology and electrical engineering |
| Ciclo di dottorato: | 34 |
| Coordinatore del Corso di dottorato: | nome email Riccio, Daniele daniele.riccio@unina.it |
| Tutor: | nome email Mazzeo, Antonino [non definito] |
| Data: | 2021 |
| Parole chiave: | Computing Systems Design Method, Approximate Computing, Evolutionary Algorithm, Design Space Exploration, Code Mutation. |
| Settori scientifico-disciplinari del MIUR: | Area 09 - Ingegneria industriale e dell'informazione > ING-INF/05 - Sistemi di elaborazione delle informazioni |
| Depositato il: | 31 Gen 2022 09:39 |
| Ultima modifica: | 28 Feb 2024 11:32 |
| URI: | http://www.fedoa.unina.it/id/eprint/14348 |
Abstract
Approximate Computing (AxC) paradigm aims at designing computing systems that can satisfy the rising performance demands and improve the energy efficiency. AxC exploits the gap between the level of accuracy required by a given application, and the actual precision provided by the computing system, for achieving diverse optimizations. Various AxC techniques have been proposed so far in the literature at different abstraction levels from hardware to software. These techniques have been successfully utilized and combined to realize approximate implementations of applications in various domains (e.g., data analytic, scientific computing, multimedia and signal processing, and machine learning). Unfortunately, state-of-the-art approximation methodologies focus on a single abstraction level, such as combining elementary components, e.g., arithmetic operations, and usually optimize hardware-requirements under error constraints, resulting in suboptimal solutions. This hinders the possibility for designers to explore different approximation opportunities, optimized for different applications and implementation targets. Therefore, we propose a methodology for the design of approximate applications which is based on multi-objective optimization and does not depend on either applications or techniques. We discuss each phase and steps the methodology breaks into, while devoting the needed relevance to their automation. In order to validate and evaluate our method, we resort to a vast plethora of applications, including generic combinational logic, arithmetic circuits, image-processing and machine-learning applications. For each of them, we report several case studies and experiments that empirically prove the validity and effectiveness of the methodology, which allow achieving significant savings both in terms of area and power required by hardware accelerators, at the cost of very low introduced error.
Downloads
Downloads per month over past year
Actions (login required)
 |
Modifica documento |
