Paper

Neural networks empowered by custom instructions – Case study

Embedded AI on L31 core

Whitepaper Cover - Neural networks empowered by custom instructions – Case study

Embedded devices are typically resource-constrained, making it difficult to run AI algorithms on embedded platforms. The Codasip Application Engineering team looked at what could make it easier from a software and hardware point of view. They used the Codasip L31 RISC-V core and Codasip Studio to explore and customize the design.

The Codasip Application Engineering team used TensorFlow Lite for Microcontrollers (TFLite-Micro) as a dedicated AI framework and compared the performance of the Codasip L31 processor core with both standard and custom extensions. This case study highlighted the benefits of custom instructions for neural networks.

Other papers & case studies

23 Jul 2024papers

CHERI is not just a hardware extension – Poster

23 Jul 2024papers

Commercializing CHERI on a Codasip A730 RISC-V application core – Poster

23 Jul 2024papers

Let’s make a standard for CHERI-RISC-V – Poster

23 Jul 2024papers

Optimizing neural network classification on resource constrained processors through customization – Poster

09 Feb 2024papers

3 months to accelerate AI on a RISC-V core

14 Aug 2023papers

Re-targetable C/C++ LLVM compiler for RISC-V – Poster

31 Jul 2023papers

Effortless DSP extensions design for embedded RISC-V CPUs – Poster

19 Jul 2023papers

FIR and Median Filter accelerators in CodAL – Technical paper

27 Jun 2023papers

Compact CORDIC accelerator implementation for embedded RISC-V core – Poster

27 Jun 2023papers

RISC-V as an enabler of heterogeneous compute – Poster