Logo attribution: The GenIE logo was created with the assistance of generative AI.
Code for NorCAS Paper: GenIE: Reuse-Oriented Generation of Domain-Specific Instruction Extensions
The full code (including automation and synthesis scripts) will be mage availabel later this year.
Some parts of the implementation are already available via: https://github.com/tum-ei-eda/isaac-toolkit
Support for:
- 64-bit application cores (multi-issue pipelines)
- custom memory & branch instructions
- additional simulators and compilers (e.g. RISCV-GCC)
- alternative retargeting/synthesis tools
- different extension interfaces (i.e. OpenHWGroup CV-X-IF)
The HLS tool used in the paper is currently not available at this point in time. We are working on the integration of alternative HLS solutions and extension interfaces.
This research is partially funded by the German Federal Ministry of Education and Research (BMBF) within the projects Scale4Edge (grant number 16ME0465) and MANNHEIM-FlexKI (grant number 01IS22086L).
Launch experiments from a browser form or the CLI, then inspect stages, live logs, and metrics in the same lightweight web app. See the monitor guide for startup and remote access.
Local LLVM retargeting uses the ISAAC port of current Seal5 main, including the typed CoreDSL parser.
GenIE tracks the audited ISAAC demo sources and supports Spike plugin and CFU flows with the current retargeting job API. See the migration guide for configuration and validation, and the Seal5 guide for service image rebuilding and repository branches to push.