How fpga-research/fabulous works
Project type and purpose
• FABulous is an open-source Python-based FPGA fabric generation framework and toolchain.
• It provides a CLI for orchestrating a modular pipeline that:
• parses a CSV fabric definition and tile templates
• generates HDL/VHDL fabric models and project metadata
• invokes external CAD tools (Yosys, nextpnr or VPR) to produce bitstreams
• produces simulation test benches and a geometry file for FABulator.
File structure
• Top level:
• FABulous.py – CLI entry point
• fabric_generator/ – core code-generation modules (file_parser, code_generator, model_generation)
• geometry_generator/ – modules to extract fabric geometry
• fabric_cad/ – TCL and Python scripts for mapping primitives and driving synthesis flows
• fabric_files/ – HDL/VHDL templates & project skeletons for Verilog and VHDL
• docs/ – Sphinx documentation, figures, usage guides, build scripts
• Dockerfile & .github/workflows – CI for docs and automation
• Patterns:
• Orchestration layer (CLI) controls a sequence of pure-Python stages.
• Separation of “templates” vs “generator logic” vs “external tool adapters.”
Insights from the README
• Workflow steps in the shell:
1) load_fabric [fabric.csv]
2) run_FABulous_fabric → generates HDL, .FABulous metadata
3) run_FABulous_bitstream npnr/VPR → calls CAD toolchain to produce a bitstream
4) gen_geometry → emits a CSV/JSON geometry file for FABulator
5) exit
• Supports both interactive shell and TCL‐driven batch mode.
• Emulation path uses Icarus Verilog for test simulation.