Search papers, labs, and topics across Lattice.
This paper introduces LEAP, a self-supervised model that enables efficient per-cycle toggle propagation prediction in VLSI design, overcoming the limitations of traditional methods that suffer from high runtimes. By leveraging a linear-complexity graph transformer and innovative self-supervised pre-training tasks, LEAP achieves a remarkable 7.6x speedup over existing EDA tools while maintaining a near-perfect PR-AUC of 0.99. Additionally, LEAP integrates seamlessly with other machine learning power models, resulting in a mean absolute percentage error of just 4.55% in layout power predictions, significantly enhancing the efficiency of power analysis workflows.
LEAP achieves a 7.6x speedup in toggle propagation prediction while maintaining near-perfect accuracy, revolutionizing power analysis in VLSI design.
Accurate power analysis is critical in VLSI design, as it directly impacts power optimization strategies. However, traditional approaches are often hindered by the substantial runtime required for per-cycle toggle propagation in the netlist, which propagates register toggle information through combinational logic. To address this, we propose LEAP, the first work to enable per-cycle toggle propagation prediction with both high accuracy and efficiency. This is achieved through a novel, linear-complexity graph transformer capable of simulating toggle propagation, along with specially designed self-supervised pre-training tasks that enable the model to capture circuit structure and functionality. LEAP achieves a 7.6x speedup over the EDA tool in toggle propagation, and attains a near-perfect area under the Precision-Recall curve (PR-AUC) of 0.99 for prediction results. Moreover, LEAP can be seamlessly integrated with other machine learning based power models into LEAP-Power. This integration enables precise per-cycle layout power prediction directly from post-synthesis netlists, achieving a mean absolute percentage error(MAPE) of only 4.55%. By bypassing toggle propagation in the netlist, LEAP-Power delivers substantial runtime gains, running 5.3x faster than the model without LEAP.
