Invited: Toward Sustainable and Transparent Benchmarking for Academic Physical Design Research

TL;DR

This work addresses the need for sustainable, reproducible benchmarking in both 2D and emerging Pin-3D physical design by introducing RosettaStone 2.0, a backplane tightly integrated with OpenROAD-Research and ORFS-Research. It delivers an end-to-end RTL-to-GDS reference flow for 2D and Face-to-face (F2F) hybrid-bonding 3D designs, including 3D enablements, stage-wise checkpoints, and METRICS2.1-compatible reporting. The framework supports benchmark translation, synthetic netlist generation via ArtNet, and community governance through CI-regression and standardized evaluation contracts, enabling apples-to-apples comparisons across tool versions and PDK variants. The approach promises to accelerate credible, cumulative progress in both planar and 3D physical design research by providing a reusable, open, and extensible benchmarking backplane with transparent artifact management and evaluation semantics.

Abstract

This paper presents RosettaStone 2.0, an open benchmark translation and evaluation framework built on OpenROAD-Research. RosettaStone 2.0 provides complete RTL-to-GDS reference flows for both conventional 2D designs and Pin-3D-style face-to-face (F2F) hybrid-bonded 3D designs, enabling rigorous apples-to-apples comparison across planar and three-dimensional implementation settings. The framework is integrated within OpenROAD-flow-scripts (ORFS)-Research; it incorporates continuous integration (CI)-based regression testing and provides a standardized evaluation pipeline based on the METRICS2.1 convention, with structured logs and reports generated by ORFS-Research. To support transparent and reproducible research, RosettaStone 2.0 further provides a community-facing leaderboard, which is governed by verified pull requests and enforced through Developer Certificate of Origin (DCO) compliance.

Figures (11)

• Figure 1: RosettaStone 2.0 roadmap (vision) to sustain and expand the open-source benchmarking backplane. Not all planned components are fully validated in this paper.
• Figure 2: Overview of the Pin-3D flow built on OpenROAD-Research and ORFS-Research.
• Figure 3: Metal stack, PDN strategy, and tier strategy in the 3D ASAP7 PDK.
• Figure 4: Rebuilding of rows for heterogeneous legalization.
• Figure 5: Cross-tier net count vs. UBfactor based on 50 random seeds per point. The trace shows the Expected Minimum Cutsize (bootstrapped, $N=15$), representing the typical best result achievable.