RISC-V, the open standard instruction set architecture (ISA) based on RISC principles, has revolutionised the tech landscape since its inception at the University of California, Berkeley in 2010. Free and open source,
RISC-V's modular and simple design supports a wide range of applications, from embedded systems to supercomputers, making it a highly scalable and efficient solution. Managed by the RISC-V Foundation, its growing community and ecosystem provide a flexible and cost-effective alternative to proprietary ISAs like ARM and x86. RISC-V is increasingly adopted for AI on the edge, thanks to its open and customisable nature. It is well-suited for edge AI applications where efficiency, low power consumption, and tailored hardware are crucial. Its modular architecture allows for custom extensions optimised for AI workloads, such as vector processing and machine learning accelerators.

This flexibility enables the design of specialised processors that handle AI inference and processing directly on edge devices, reducing the need for constant cloud connectivity and enabling faster, more efficient AI operations in applications like IoT devices, smart cameras, and autonomous vehicles. Klepsydra has benchmarked the execution of AI algorithms on the Microchip PolarFire ICICLE using ESA’s OBPMark-ML models and TensorFlowLite as a baseline. The results were impressive, with Klepsydra AI running up to 5x faster than the baseline!