Options
April 2, 2025
Conference Paper not in Proceedings
Title
EMSA5: A RISC-V Processor System for Enhanced Functional Safety in Embedded Applications
Title Supplement
Paper presented at RISC-V in Space Workshop 2025, 2-3 April 2025, Gothenburg, Sweden
Abstract
The RISC-V RV32 processor system EMSA5 was originally designed to meet the stringent requirements of functional safety as specified by ISO26262. It consists of a redundant multi-core architecture that supports advanced safety mechanisms such as Dual-Mode-Redundancy with Lockstep (DMR-L) and Triple-Mode-Redundancy (TMR), ensuring high reliability in critical applications. However, functional safety is also highly relevant in industrial, avionics and space applications. A comprehensive RISC-V system requires more than just a core; the EMSA5 includes a complete ecosystem tailored for software development, as well as essential peripherals for both processing and communication. To facilitate different communication interface requirements, EMSA5 covers both classic CAN and modern Ethernet-based network architectures, demonstrating its versatility in various automotive and industrial applications.
Furthermore, the incorporation of RISC-V processor extension significantly improves performance in a trade-off with resource utilization, enabling more efficient data processing and computational capabilities.
This work shows the EMSA5 processor for embedded sensor signal processing in a network environment. The algorithms are based on neural networks, and it is evaluated in performance with and without Vector Extension Zve32x for quantized data and Floating Point Extension for floating point data. In addition to performance, the system is analyzed in terms of logic resources. The EMSA5 processor is compatible with leading FPGA platforms such as Microchip and AMD Xilinx, and its silicon has been proven by foundries like Global Foundries. This underlines the EMSA5's contribution to the development of safe, efficient, and versatile processing solutions in the ever-evolving landscape of embedded systems.
Furthermore, the incorporation of RISC-V processor extension significantly improves performance in a trade-off with resource utilization, enabling more efficient data processing and computational capabilities.
This work shows the EMSA5 processor for embedded sensor signal processing in a network environment. The algorithms are based on neural networks, and it is evaluated in performance with and without Vector Extension Zve32x for quantized data and Floating Point Extension for floating point data. In addition to performance, the system is analyzed in terms of logic resources. The EMSA5 processor is compatible with leading FPGA platforms such as Microchip and AMD Xilinx, and its silicon has been proven by foundries like Global Foundries. This underlines the EMSA5's contribution to the development of safe, efficient, and versatile processing solutions in the ever-evolving landscape of embedded systems.
Conference
File(s)
Rights
Use according to copyright law
Language
English