KRS Unleashed speeds up FPGA prototyping 80x for ROS 2
Build times cut by 80x with modular FPGA acceleration toolchain.
KRS Unleashed revives and heavily refactors the Kria Robotic Stack (KRS), originally created by Xilinx/AMD for FPGA acceleration on their Kria platform. The new toolchain, developed over a year as part of a PhD, tackles the original’s painful rebuild times and fragile dependency chains. By splitting the workflow into three isolated workspaces—OS (firmware and sysroot), vitis (bitstream automation via Python CLI), and krs (lightweight ROS 2 colcon integration)—the tool reduces build times by up to 80x. Users can reuse firmware across projects, swap Vitis versions without breaking dependencies, and even use vitis+OS standalone for rapid host code iteration. The first public release includes a five‑part Hackster tutorial and pending support for the Jazzy ROS 2 distro.
The refactor explicitly addresses real developer pain points: long clean builds, opaque package purposes, and fragile cross‑references. With logical separation, teams can develop firmware, accelerators, and ROS code in parallel, and the colcon mixin integration remains familiar to ROS 2 users. A new tracetools-kernels package extends ros‑tracing to arbitrary code regions. The project is open‑source on GitHub (TUD‑ADS/KRS‑Unleashed), and a paper is published at IEEE. For FPGA‑accelerated robotics, this lowers the barrier to high‑performance, real‑time prototyping.
- Refactors original KRS into three isolated workspaces: OS (firmware/sysroot), vitis (bitstream automation), and krs (ROS 2 colcon integration).
- Build times are up to 80x faster by caching OS and sysroot across projects and avoiding unnecessary re-synthesis.
- Includes a Python CLI for Vitis automation and a tracetools-kernels package for arbitrary ROS 2 code region tracing.
Why It Matters
Makes FPGA acceleration in ROS 2 practical for rapid prototyping, slashing iteration time from hours to minutes.