Programmer Productivity and Performance on Embedded RISC-V CPUs
Talkat RISC-V Forum: Developer Tools and Tool Chains 2021 · on Wednesday, 2021-06-02· at 17:05· Virtual·
summary
ePython implements a 24KB interpreter for tiny RISC‑V embedded cores and uses the Olympus compiler with a novel dynamic-loading technique to run effectively unlimited Python code while needing only ≈10KB memory at boot. Functions and data live as segmented placeholders and are fetched on demand (with a simple prefetcher), producing performance near native C on a 16‑core PicoRV softcore; the original interpreter is about 400× slower, while the compiled/dynamic approach is close to native in the Jacobi/Laplace diffusion benchmark.
AI-written summary · about
chapters (8)
- 0:00Introduction
- 0:30IoT Edge Computing
- 1:28Epyson
- 2:59Compilation
- 4:36Load procedure
- 5:41Data
- 6:33Benchmarks
- 8:25Conclusion
related talks
56mScripting the Internet of Things
01h 01mThe Genius of RISC-V Microprocessors
10mRISC-V Virtual Platforms for Early RISC-V Embedded Software Development
10mEnabling Production Program Software Development for RISC-V
15mRISC-V CPUs for the New Era of Heterogeneous Computing
24mUniversal Tools for Acceleration, Timing, Integration & Machine Enhancement
11mGetting Started with RISC-V Custom Instructions
25mCo-operative JIT Compilation for Resource-Constrained Low-Power Coprocessors