Fusion SoC Workshop User Guide

Demo Access Options

CNB Codespace

1. Login to CNB repo.
2. Fork the repo.
3. Click Workspace.
4. Connect to the workspace (VS Code recommended).
5. Jump to the section “How to run Fusion Framework with Docker Image”.

Gitpod

1. Login to Gitpod.
2. Create a new project.
3. Create a new environment.
4. Open the workspace in VS Code.
5. Jump to the section “How to run Fusion Framework with Docker Image”.

Local Docker Environment

1. Load the docker image.

   docker load -i chipyard.tar

2. Run the docker image.

   docker run -it docker.cnb.cool/fudaneda/docker/chipyard /bin/bash

How to Run Fusion Framework with Docker Image

1. Change the workspace folder to /root/chipyard/.

   cd /root/chipyard

2. Build the front-end compiler and run the Taylor example compile.

   cd /root/chipyard/generators/fgra/app-compiler
   ./build.sh
   cd /root/chipyard/generators/fgra/benchmarks/taylor
   ../compile.sh taylor kernel
   cd ../ && ./dot2json.sh

3. Activate the chipyard conda environment and build the hardware.

   In CNB, you can change VERILATOR_THREADS in build-verilator.sh to 64 for faster simulation.

   cd /root/chipyard/generators/fgra
   source /root/chipyard/env.sh
   ./build-verilog.sh && ./build-verilator.sh

4. Build and run the mapper to generate cgra_execute.c and cgra_call.txt.

   conda deactivate && conda deactivate
   cd /root/chipyard/generators/fgra/fgra-mapper
   ./build.sh
   ./run.sh

cgra_execute.c:

cgra_call.txt:

5. Update taylor.c according to the generated artifacts and build the binary.

   

   cd /root/chipyard/generators/fgra/software/tests/bareMetalC
   ./build.sh

6. Generate the simulation executable and run it from the FGRA root directory.

   cd /root/chipyard/generators/fgra
   ./run-verilator.sh taylor

Success!