How safie-tarek/accelerating_digital_verification_using_ai-based_constrained_stimulus_generation works
The README describes an RL-guided verification loop: a UVM simulation produces coverage feedback, an agent selects the next stimulus, and results are compared with constrained-random verification. The checked-in tree does not contain the README’s described rl/, integration/, or UVM source directories; it contains iDMA RTL and otherwise sparse support directories. Accordingly, the workflow below is documented rather than source-verified, and README.md is the evidence for its relationships.
Main entry points: README.md RL agent · idma_generated.sv iDMA RTL
Components 12
RL control
- RL agent
README.md - State building
README.md - Action selection
README.md - Coverage reward
README.md - Stimulus mapping
README.md
Verification
- Coverage report
README.md - Applied stimulus
README.md - Run budget
README.md - UVM simulation
README.md
Design under test
- iDMA RTL
rtl/idma_generated.sv
Evaluation
- Constrained-random
README.md - Coverage and cost comparison
README.md
Connections 13
- UVM simulation → Coverage report: produces
README.md - UVM simulation → Applied stimulus: records
README.md - RL agent → Coverage report: reads
README.md - RL agent → Applied stimulus: reads
README.md - RL agent → Run budget: reads
README.md - RL agent → State building: builds
README.md - RL agent → Action selection: selects through
README.md - RL agent → Coverage reward: scores progress with
README.md - RL agent → Stimulus mapping: maps actions through
README.md - Stimulus mapping → UVM simulation: sets stimulus for
README.md - UVM simulation → iDMA RTL: verifies
README.md - Constrained-random → Coverage and cost comparison: provides baseline
README.md - UVM simulation → Coverage and cost comparison: provides RL results
README.md