Assessment of Reinforcement Learning for Macro Placement
This work addresses the need for transparency and reproducibility in AI-driven chip design tools for researchers and practitioners, but it is incremental as it focuses on assessment rather than introducing new methods.
The researchers tackled the problem of evaluating Google Brain's deep reinforcement learning approach for macro placement by providing an open-source implementation and assessment, revealing discrepancies and testing on new benchmarks, with results showing competitive performance but not achieving state-of-the-art across all cases.
We provide open, transparent implementation and assessment of Google Brain's deep reinforcement learning approach to macro placement and its Circuit Training (CT) implementation in GitHub. We implement in open source key "blackbox" elements of CT, and clarify discrepancies between CT and Nature paper. New testcases on open enablements are developed and released. We assess CT alongside multiple alternative macro placers, with all evaluation flows and related scripts public in GitHub. Our experiments also encompass academic mixed-size placement benchmarks, as well as ablation and stability studies. We comment on the impact of Nature and CT, as well as directions for future research.