AI-Driven Timing-Aware Power Optimization Uses Learned Ranking With Baseline-Preserving Repair
Amit Sharma · hf · 2026-10-16
IEEE member Amit Sharma proposes an AI-driven framework for post-route power recovery, where cell choices must change under tightly coupled timing, electrical, and physical constraints, and each edit's value depends on the cost of repairing its side effects.
- Framework: A learned model drives both phases—ranking power-recovery edits to try and the repair actions that follow—while deterministic EDA analyses govern every acceptance, record baseline constraint margins, identify damage introduced by each recovery transaction, hold pre-existing violations at baseline, and cap repair spending as a fraction of measured savings.
- Key finding: A companion study on public designs shows conventional whole-design timing-closure repair folds in pre-existing violators and falters as the ratio of pre-existing violators to recovery damage rises, while a command aimed only at the damage retains nearly all the saving—motivating explicit baseline preservation: returning every constraint to its pre-recovery margin, separated from the stronger requirement that the whole design become timing-clean.
More from Infra
- Meme: driving to the chip factory in 2028 captures AI's compute anxiety — beffjezos · 2026-10-16
- One volunteer maintains the time zone file behind 4 billion phones — danbri · 2026-10-16
- DeepSeek V4.1 Flash on 4x DGX Spark: weights alone are 511GB, the 494 tok/s claim doesn't fit — Aroochacha · 2026-10-16
- Open-sourced Intel GPU Vulkan inference upgrades: 2x ingestion, 60%+ faster inference — Fun-Aardvark-1143 · 2026-10-16
- ELSFP pain points: >3dB insertion loss, costly PMF, harsh packaging requirements — jwt0625 · 2026-10-16
- MLCommons launches MLPerf Endpoints, a benchmark that measures inference endpoints the way buyers actually purchase them — TheKanter · 2026-10-16