HULK-V: a Heterogeneous Ultra-low-power Linux capable RISC-V SoC

TL;DR

HULK-V tackles the IoT need for Linux-capable, ultra-low-power computing by fusing a 64-bit CVA6 host with an 8-core PMCA accelerator and a fully digital HyperRAM-based memory subsystem. It delivers up to $13.8~\text{GOps}$ and $157~\text{GOps/W}$ within a $250~\text{mW}$ envelope, enabled by a 128 kB Last Level Cache and a digital memory controller that supports flexible memory interleaving and high efficiency. The software stack integrates Linux with an OpenMP 5 offload model to transparently offload kernels to the PMCA, using a shared memory region and a Clang/LLVM-based toolchain for heterogeneous code. Benchmark results from FPGA-based emulation demonstrate meaningful PMCA speedups and energy efficiency gains, while the fully digital HyperRAM-based memory hierarchy achieves DDR-like performance at substantially lower area and power. The work is open-source, offering a practical, scalable platform for ultra-low-power Linux IoT with DSP/ML acceleration and a productive heterogeneous programming workflow.

Abstract

IoT applications span a wide range in performance and memory footprint, under tight cost and power constraints. High-end applications rely on power-hungry Systems-on-Chip (SoCs) featuring powerful processors, large LPDDR/DDR3/4/5 memories, and supporting full-fledged Operating Systems (OS). On the contrary, low-end applications typically rely on Ultra-Low-Power ucontrollers with a "close to metal" software environment and simple micro-kernel-based runtimes. Emerging applications and trends of IoT require the "best of both worlds": cheap and low-power SoC systems with a well-known and agile software environment based on full-fledged OS (e.g., Linux), coupled with extreme energy efficiency and parallel digital signal processing capabilities. We present HULK-V: an open-source Heterogeneous Linux-capable RISC-V-based SoC coupling a 64-bit RISC-V processor with an 8-core Programmable Multi-Core Accelerator (PMCA), delivering up to 13.8 GOps, up to 157 GOps/W and accelerating the execution of complex DSP and ML tasks by up to 112x over the host processor. HULK-V leverages a lightweight, fully digital memory hierarchy based on HyperRAM IoT DRAM that exposes up to 512 MB of DRAM memory to the host CPU. Featuring HyperRAMs, HULK-V doubles the energy efficiency without significant performance loss compared to featuring power-hungry LPDDR memories, requiring expensive and large mixed-signal PHYs. HULK-V, implemented in Global Foundries 22nm FDX technology, is a fully digital ultra-low-cost SoC running a 64-bit Linux software stack with OpenMP host-to-PMCA offload within a power envelope of just 250 mW.

Figures (9)

• Figure 1: Overall HULK-V SoC Architecture.
• Figure 2: Last Level Cache Architecture.
• Figure 3: HyperRAM Memory Controller Architecture.
• Figure 4: HULK-V SW stack (from HW to user-space)
• Figure 5: HULK-V floorplan.