OpenTitan Benchmark for svlens

Measures svlens accuracy and performance against real SoC RTL.

Quick Start

make bench

Manual Steps

cd bench/opentitan
bash fetch.sh                    # Clone OpenTitan (one-time, ~2GB)
python3 gen_filelist.py          # Generate .f filelists from .core files
bash run.sh                      # Run svlens on all targets
python3 evaluate.py              # Compute recall/precision

Results appear in bench/opentitan/results/bench_report.md.

Targets

Level	IP	Top Module	Purpose
L1	aes	aes	Multi-clock IP
L2	hmac	hmac	Simple baseline
L3	uart	uart	External interface
L4	top_earlgrey	top_earlgrey	Full SoC scale

Requirements

• svlens built (make build)
• Python 3.8+ with PyYAML (pip install pyyaml)
• Git
• /usr/bin/time (for memory measurement)
• ~2GB disk for OpenTitan clone