Recommended RISC-V Software Ecosystem Expansion for Key Industry Markets

Date: August 2026
Scope: Proposed expansion of tracked open-source projects in project-reports/scope.yml to cover Virtualization, Machine Learning, IoT & Edge, Automotive, and Aerospace markets.

Executive Summary

The current tracking list of 147 open-source projects covers core Linux OS tools, container engines, language runtimes, and baseline ML libraries. To support key growth markets for RISC-V hardware—specifically Virtualization, Machine Learning, IoT & Edge, Automotive, and Aerospace—the software tracking scope should be expanded to include critical industry-standard libraries, real-time operating systems (RTOS), and safety-critical middleware.

1. Virtualization Market

While the current scope tracks heavy-weight hypervisors (QEMU, libvirt) and container engines (containerd, runc), modern cloud and edge infrastructure relies on Rust-based microVMs, user-space I/O drivers, and type-1 bare-metal hypervisors:

Cloud Hypervisor (cloud-hypervisor/cloud-hypervisor): A Rust-based Virtual Machine Monitor (VMM) built on KVM. It is the modern standard for lightweight cloud microVMs (offering faster boot times and lower memory overhead than QEMU).
Firecracker (firecracker-microvm/firecracker): AWS-developed Rust microVM monitor for serverless/FaaS environments. Tracking RISC-V porting is critical for multi-tenant cloud/edge function execution.
crosvm (google/crosvm): ChromeOS Virtual Machine Monitor (Rust), heavily used in Android Virtualization Framework (AVF) and automotive domain isolation.
DPDK (Data Plane Development Kit) (DPDK/dpdk): High-speed packet processing drivers and libraries running in user space. Essential for Network Functions Virtualization (NFV) and virtual switches on RISC-V servers.
SPDK (Storage Performance Development Kit) (spdk/spdk): User-space, polled-mode NVMe storage drivers. Crucial for high-throughput virtualized block storage backends.
Xen Hypervisor (xen-project/xen): Type-1 bare-metal hypervisor. Critical for partitioning and safety-critical isolation in virtualized automotive and aerospace ECUs.

2. Machine Learning & AI Market

The current scope tracks key frameworks (PyTorch, vLLM, llama.cpp, LiteRT, ONNX), but lacks ML compilers, classical scientific ML packages, and vector search databases:

OpenAI Triton (triton-lang/triton): A Python-like language and compiler for custom ML kernels. Triton is the primary backend engine for PyTorch 2.0 (torch.compile / Inductor); establishing RISC-V Triton support is a prerequisite for PyTorch compilation performance.
Apache TVM (apache/tvm): An end-to-end deep learning compiler stack targeting embedded hardware and RISC-V vector extensions (RVV).
IREE (iree-org/iree): An MLIR-based execution environment and compiler for embedded and CPU vector acceleration. A key target of the RISE AI/ML working group.
MNN (Alibaba MNN) (alibaba/MNN): A lightweight mobile/embedded deep learning engine with dedicated RVV 1.0 assembly kernels.
Tencent NCNN (Tencent/ncnn): High-performance neural network inference framework optimized for mobile and embedded devices.
scikit-learn (scikit-learn/scikit-learn) & SciPy (scipy/scipy): Core Python classical machine learning and scientific computing libraries.
Milvus / Knowhere (milvus-io/milvus / knowhere): Vector database engine and vector index library for production RAG and AI search workflows.

3. IoT & Edge Work Market

Edge and IoT devices operate under strict memory, power, and real-time constraints, requiring real-time operating systems (RTOS), TinyML frameworks, and embedded security/communication protocols:

Zephyr RTOS (zephyrproject-rtos/zephyr): The Linux Foundation’s flagship real-time operating system for IoT devices. Has active RISC-V support, but requires continuous tracking for RVV and POSIX API compliance.
FreeRTOS (FreeRTOS/FreeRTOS): The de-facto open-source microcontroller RTOS for low-power edge hardware.
TFLite for Microcontrollers (tflite-micro) (tensorflow/tflite-micro): TinyML inference engine designed to run neural networks on microcontrollers with KB-level RAM (includes RVV microkernels).
mbedTLS (Mbed-TLS/mbedtls) / wolfSSL (wolfSSL/wolfssl): Embedded TLS/crypto libraries designed for small-footprint hardware. Essential for validating hardware crypto ISA extensions (Zkn/Zks).
Eclipse Mosquitto (eclipse/mosquitto): Lightweight MQTT message broker for IoT edge messaging.
micro-ROS (micro-ROS/micro_ros_setup): ROS 2 runtime optimized for microcontrollers, bridging edge sensors to autonomous systems.
cJSON (DaveGamble/cJSON): Ultra-lightweight C JSON parser used across millions of embedded IoT firmware builds.

4. Automotive Market

Automotive software demands real-time responsiveness, CAN-bus networking, safety-critical domain isolation, and autonomous driving middleware:

ROS 2 (Robot Operating System 2) (ros2/ros2): The foundational middleware framework for autonomous driving, robotics, and ADAS (Advanced Driver Assistance Systems).
eProsima Fast DDS (eProsima/Fast-DDS) / Eclipse CycloneDDS (eclipse-cyclonedds/cyclonedds): High-performance Data Distribution Service (DDS) implementations used as the real-time transport engine for ROS 2 in automotive ECUs.
SocketCAN / can-utils (linux-can/can-utils): Linux kernel CAN-bus subsystem utilities and user-space libraries for vehicle network communication.
Jailhouse Hypervisor (siemens/jailhouse): A Linux-based partitioning hypervisor designed specifically for real-time safety-critical automotive domain isolation.
openpilot (commaai/openpilot): Open-source driver assistance system (ADAS) software stack.

5. Aerospace Market

Aerospace and avionics systems require fault-tolerant flight software, satellite/UAV autopilot engines, space-grade RTOSs, and safety-critical language runtimes:

cFS (NASA Core Flight System) (nasa/cFS): NASA’s reusable flight software framework used in satellite missions, cubesats, and deep-space instruments.
NASA FPrime / F’ (nasa/fprime): NASA JPL’s component-driven flight software framework designed for small-scale space systems and CubeSats.
RTEMS (Real-Time Executive for Multiprocessor Systems) (RTEMS/rtems): Open-source real-time OS designed for space exploration, avionics, and flight hardware (long history on SPARC and RISC-V harts).
PX4 Autopilot (PX4/PX4-Autopilot) & ArduPilot (ArduPilot/ardupilot): Leading open-source autopilot software stacks for uncrewed aerial vehicles (UAVs) and drones.
MAVLink (mavlink/mavlink): Micro Air Vehicle communication protocol for drone and spacecraft telemetry.
GNAT / Ada & SPARK Runtimes (AdaCore/gnat-community): Formal verification and safety-critical language runtime required for DO-178C avionics certification.

Summary Matrix & Strategic Roadmap

Market Segment	Proposed Additions	Strategic Benefit to RISC-V Ecosystem
Virtualization	`cloud-hypervisor`, `firecracker`, `crosvm`, `DPDK`, `SPDK`, `Xen`	Unblocks modern Rust microVMs, high-speed NVMe/networking, and Type-1 bare-metal isolation.
Machine Learning	`Triton`, `Apache TVM`, `IREE`, `MNN`, `scikit-learn`, `SciPy`, `Milvus`	Enables PyTorch 2.0 `torch.compile` codegen, deep learning compilation, and vector search.
IoT & Edge Work	`Zephyr RTOS`, `FreeRTOS`, `tflite-micro`, `mbedTLS`, `wolfSSL`, `Mosquitto`	Secures embedded microcontroller RTOS support, TinyML execution, and hardware crypto ISA validation.
Automotive	`ROS 2`, `Fast DDS`, `CycloneDDS`, `can-utils`, `Jailhouse`	Provides the software foundation for autonomous driving (ADAS), vehicle CAN-bus, and ECU isolation.
Aerospace	`NASA cFS`, `NASA FPrime`, `RTEMS`, `PX4 Autopilot`, `ArduPilot`, `MAVLink`, `GNAT (Ada)`	Unblocks flight software for satellites/drones and enables safety-critical DO-178C avionics certification.