Towards a Base-Station-on-Chip: RISC-V Hardware Acceleration for wireless communication

TL;DR

This paper explores implementing wireless communication signal processing tasks on RISC-V vector processors to create a compact, efficient Base-Station-on-Chip for next-generation 5G/6G networks, integrating AI/ML capabilities.

Contribution

It demonstrates how RISC-V vector extensions and custom instructions can optimize wireless signal processing kernels for a BSoC platform.

Findings

Efficient implementation of CE, mMIMO, and beamforming on RISC-V DSPs.

RISC-V Vector Extensions improve throughput and latency for PHY kernels.

Hardware/software co-design achieves scalable, low-power wireless processing.

Abstract

The evolution of 5G and the emergence of 6G wireless communication systems impose higher demands for computing capabilities and lower power consumption in the front-end and processing circuitry. Furthermore, the incorporation of Artificial Intelligence (AI)/Machine Learning (ML) in the Radio Access Network (RAN) introduces heightened computational needs and stringent low-latency requirements for both training and inference. The concept of a Base Station on Chip (BSoC) addresses those demands by consolidating of the signal processing, neural network computations and network management functions into a single chip. This new computing platform relies on a sophisticated hardware/software co-design to optimize performance, power efficiency, and scalability, enabling a compact, yet adaptable and intelligent base station solution for next-generation wireless networks. This research investigates the efficient implementation of conventional Channel Estimation (CE), massive Multiple Input Multiple Output (mMIMO), and beamforming kernels on a state-of-the-art RISC-V vector Digital Signal Processors (DSP) to capitalize on Data Level Parallelism (DLP). Moreover, it explores how RISC-V Vector Extensions (RVV) combined with custom instructions can effectively address the throughput and latency demands of LOW Physical Layer (PHY) kernels.