SymRTLO is a state-of-the-art neuron-symbolic RTL optimizer that combines LLM-based rewriting, symbolic reasoning, and fast verification to achieve up to 48% lower power, 91% faster timing, and 47% reduced area.

We introduce ArchPower, the first open-source dataset for ML-based CPU architecture-level power modeling. It provides 200 realistic CPU data samples based on 25 CPU configurations and 8 workloads.