SAURIA-CHESHIRE integration demonstrator

This repository contains a demonstration of the integration of the SAURIA convolutional accelerator and CHESHIRE MCU. The project includes source code, hardware design files, and supporting scripts for simulation and FPGA deployment.

Project Structure

• hw/: Hardware design files, including SystemVerilog modules and packages.

  • axi_intfc_bridge.sv, axi_lite_intfc_bridge.sv, sauria_demo_soc.sv: Core hardware modules.
  • include/: Contains reusable hardware packages like sauria_demo_pkg.sv.

• sw/: Software components for the SoC.

  • data/: Input, weight, and output tensor data files.
  • lib/: Utility libraries and headers for software development.
  • src/: Source files for software applications.

• target/: Target-specific files for FPGA and simulation.

  • fpga/: FPGA-related files.
  • sim/: Simulation files, including source and ModelSim configurations.

• vendor/: Dependencies and submodules.

  • cheshire/: Cheshire-related files.
  • SAURIA/: SAURIA-related files.

• Root Files:

  • .gitignore: Git ignore rules.
  • Bender.yml: Dependency management configuration.
  • Makefile: Build automation script.
  • README.md: Project documentation.
  • sauria_demo.mk: Makefile for the SAURIA demo.

Getting Started

1. Set-up the project, allowing Bender to update the project's dependecies:

   make set-up

2. Build the hardware model:

   make hw-all

3. Build the software:

   make sw

4. Simulate SAURIA-CHESIRE demo, in two ways:

• Using make sim-gui BIN_SEL=$BIN_SEL, with $BIN_SEL equal to 0 for the Hello world example, 1 for a register & SRAM write & read test, 2 for a complete single convolutional layer test [WORK IN PROGRESS]
• Sourcing start.sauria_demo.tcl using QuestaSim opened in target/sim/vsim, and setting the BIN_SEL before calling the tcl script

License

This project is licensed under the terms specified in the repository. Please refer to the license file for details.

Acknowledgments

This project has been developed as part of the project NODES which has received funding from the MI_JQ—M4C215 of PNRR funded by the European Union - NextGenerationEU (Grant agreement no. ECSOOOOOOB6)