dse-loop

Automating parameter tuning involves iteratively running a target program, parsing a numeric objective metric from its output, and automatically selecting new parameter configurations until a stopping condition is met. This approach handles architecture simulation and EDA sweeps.

Design space exploration is an autonomous iterative loop that executes a workflow, evaluates design points using an objective direction, and logs results to prevent duplicate runs. It works for RTL and synthesis parameter sweeps by checking against a baseline run.

To start a design space exploration loop when parameter ranges are missing, the tool reads the source codebase to infer sensible sweep options. It logs the inferred parameter space for review before continuing the optimization loop.

You can recover an optimization loop after a crash using DSE_STATE.json and dse_log.csv files. These checkpoint recovery files record every iteration, including failures, ensuring the design space exploration resumes without duplicate configurations.

The best way to parse metrics during compiler and formal verification parameter searches is through programmatic metric extraction. The optimization loop requires disciplined logging of this numeric objective metric into CSV files to guide configuration selection.

Yes, design space exploration requires constraint-aware logging to dse_results for memory-subsystem configuration searches. This disciplined logging captures target metrics and prevents duplicate runs during the iterative optimization loop.