Lorenzo Marino
Exploration on a dynamically reconfigurable coprocessor engine.
Rel. Maurizio Martina, Luigi Giuffrida, Guido Masera. Politecnico di Torino, Master of science program in Computer Engineering, 2026
|
|
PDF (Tesi_di_laurea)
- Thesis
Restricted to: Only staff users fino al 27 March 2029 (data di embargo). Licence: Creative Commons Attribution Non-commercial No Derivatives. Download (4MB) |
Abstract
The transition of Artificial Intelligence from cloud architectures to edge computing is accelerating, driven by the needs for data privacy, reduced latency, and operational cost-efficiency. Enabling complex workloads at the edge requires specialized hardware capable of intensive computation within a constrained energy environment. While existing solutions such as TPUs, FPGAs, and specialized microcontrollers address these needs, this master thesis proposes an alternative open-source approach: the integration of a Coarse-Grained Reconfigurable Architecture (CGRA) named R.A.C.E. (Reconfigurable Arithmetic Coarse-grain Engine) into a RISC-V ecosystem. CGRAs offer a compelling trade-off between the flexibility of general-purpose processors and the efficiency of dedicated ASICs. However, the communication overhead between the processor core and the accelerator remains a critical performance bottleneck.
This work focuses on enhancing the CORE-V eXtension InterFace (CV-X-IF) to overcome standard bandwidth limitations and enable SIMD (Single Instruction, Multiple Data) capabilities for the R.A.C.E
Relators
Academic year
Publication type
Number of Pages
Course of studies
Classe di laurea
URI
![]() |
Modify record (reserved for operators) |
