Search papers, labs, and topics across Lattice.
This paper introduces DOPP, a 3D-IC netlist partitioning framework that directly optimizes for Power, Performance, and Area (PPA) metrics instead of relying solely on proxy objectives. DOPP uses a D-optimal design to efficiently select a subset of partitioning candidates for full PPA evaluation, building surrogate models to predict PPA for unevaluated candidates. Experiments on eight 3D-IC designs show DOPP achieves significant PPA improvements (up to 21.85% TNS reduction) with comparable runtime to proxy-driven methods by evaluating only a fraction of the candidate space.
Stop relying on inaccurate proxies: directly optimizing 3D-IC partitioning for PPA metrics is now practical, yielding significant improvements without increasing runtime.
3D-IC netlist partitioning is commonly optimized using proxy objectives, while final PPA is treated as a costly evaluation rather than an optimization signal. This proxy-driven paradigm makes it difficult to reliably translate additional PPA evaluations into better PPA outcomes. To bridge this gap, we present DOPP (D-Optimal PPA-driven partitioning selection), an approach that bridges the gap between proxies and true PPA metrics. Across eight 3D-IC designs, our framework improves PPA over Open3DBench (average relative improvements of 9.99% congestion, 7.87% routed wirelength, 7.75% WNS, 21.85% TNS, and 1.18% power). Compared with exhaustive evaluation over the full candidate set, DOPP achieves comparable best-found PPA while evaluating only a small fraction of candidates, substantially reducing evaluation cost. By parallelizing evaluations, our method delivers these gains while maintaining wall-clock runtime comparable to traditional baselines.
