WYWenjian Yu

Wenjian Yu

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Wenjian Yu has a total of 151 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • A Reduction-Based Fast Algorithm for Modeling Memristive Crossbar Array With IR-Drop Effect

    by Yibin Zhang, Wenjian Yu

    The in-memory computing (IMC) based on memristive crossbar array has attracted great attention due to its great potential to enable fast and energy-efficient matrix-vector multiplication (MVM). However, the IR-drop effect caused by wire resistance largely compromises the computational accuracy. Efficient modeling of IR-drop effect in large-scale memristive crossbar arrays is of great significance for their practical application. In this paper, a highly-efficient approach for accurately modeling the memristive array with IR-drop effect is proposed, which computes an equivalent conductance matrix to capture the relationship between the array’s input and output. It is based on a novel recursive network reduction approach which exploits the special topological structure of memristive array. For an m×n array, the proposed approach has O(max(m, n)3) time complexity and O(max(m, n)2) space complexity, thus achieving significant advantage over existing approaches. Numerical experiments with the memristive arrays on various sizes and wire resistances have validated the efficiency of the proposed approach. It is shown that the proposed approach runs over 55X faster than the existing approach based on advanced sparse direct solver on a 2048×2048 array, without loss of accuracy.

  • Open Access

    APEC-Route: Guiding Analog Routing with On-the-Fly Parasitic Extraction Based on a Neural Capacitance Model

    by Haoyi Zhang, Jiechen Huang, Bingyang Liu, Wenjian Yu, Yibo Lin, Runsheng Wang

    Analog routing must balance design rules and layout constraints while managing interconnect resistance and capacitance. Many analog-routing frameworks combine wirelength objectives with constraint handling and evaluate detailed parasitic effects after routing. This separation motivates investigating whether low-latency parasitic estimates can guide candidate expansions during routing. We present APEC-Route, an analog-routing framework that combines interval-tree-based wire segmentation, a multilayer perceptron (MLP)-based neighbor-aware total-capacitance model, and analytical wire-resistance evaluation. The resulting estimates guide the search in parasitic-minimization and parasitic-matching modes. On four finalized 65 nm layouts, the extractor attains a geometric-mean relative error of 5.54–6.77% and a mean absolute percentage error of 7.41–9.93% against the field-solver reference; each in-search query takes 251–345 μ s. For each of three 65 nm block-level routed-circuit cases—an operational transconductance amplifier (OTA), a comparator, and a capacitive digital-to-analog converter (CDAC)—SAGERoute and the w/o-R/C, R-only, CC-only, and full APEC-Route variants use the same fixed placement. Across both controlled capacitance-term comparisons, enabling the capacitance term improves eight of nine independent circuit metrics, with OTA phase margin as the sole exception. The derived ENOB increases consistently with SINAD. Relative to w/o R/C, the full objective increases OTA gain and CMRR by 2.7 and 1.0 dB and UGB by 0.01 MHz, reduces comparator delay, hysteresis voltage, and offset voltage by 20.5%, 16.0%, and 26.6%, and reduces CDAC delay by 10.4% while increasing SINAD by 16.4 dB; OTA phase margin decreases by 1.1°.

  • Open Access

    CapBench: A Multi-PDK Dataset for Machine-Learning-Based Post-Layout Capacitance Extraction

    by Hector R. Rodriguez, Jiechen Huang, Wenjian Yu

    We present CapBench, a fully reproducible, multi-PDK dataset for capacitance extraction. The dataset is derived from open-source designs, including single-core CPUs, Systems-on-Chip, and media accelerators. All designs are fully placed and routed using 14 independent OpenROAD flow runs spanning three technology nodes: ASAP7, NanGate45, and Sky130HD. From these layouts, we extract 61,855 3D windows across three size tiers to enable transfer learning and scalability studies. High-fidelity capacitance labels are generated using RWCap, a state-of-the-art random-walk solver, and validated against the industry-standard Raphael, achieving a mean absolute error of 0.64% for total capacitance. Each window is pre-processed into density maps, graph representations, and point clouds. We evaluate 10 machine learning architectures that illustrate dataset usage and serve as baselines, including convolutional neural networks (CNNs), point cloud transformers, and graph neural networks (GNNs). CNNs demonstrate the lowest errors (1.75%), while GNNs are up to 41.4X faster but exhibit the larger errors (10.2%), illustrating a clear accuracy-speed trade-off. Code and dataset are available at https://github.com/THU-numbda/CapBench.

  • Open Access

    AttentionCap: Transformer Based Capacitance Matrix Learning Toward Full-Chip Extraction

    by Jiechen Huang, Hector R. Rodriguez, Dingcheng Yang, Zuochang Ye, Yibo Lin, Wenjian Yu

  • Open Access

    MCMC-Escape: Multi-Capacity Ordered Escape Routing Based on Monte-Carlo Tree Search

    by Jianxuan Yu, Zhenyi Gao, Sheqin Dong, Zuochang Ye, Wenjian Yu

    Ordered escape routing (OER), which seeks the routing paths from some signal pins to the boundary of a pin array in a given order, is an important research topic for PCB design. Although reinforcement learning based methods for OER have been proposed, the routing capacity between two adjacent pins is assumed to be just one. In this work, we propose MCMC-Escape, a Monte-Carlo tree search (MCTS) based multi-capacity ordered escape router, which includes, in turn, the initial solving approach, the improved Monte-Carlo tree search (improved MCTS) process, and the last routing attempt approach based on wires removing and re-routing. In the improved MCTS, the prior knowledge based pruning strategies and the fine-tuning strategy are proposed to enhance the efficiency of solving multi-capacity OER (MC-OER) problems, while the weight adjusting strategy is proposed to address the path occupancy issues arising from multiple capacity. Experimental results demonstrate that MCMC-Escape can effectively solve large-scale MC-OER problems and outperform existing methods in terms of routing success rate, runtime and wire length. For a set of problems with 50×50 pin array, MCMC-Escape achieves 4X higher success rate of routing with 50% less solving time than MCMCF-Router [ 1 ], while reducing the average total wire length.

  • Open Access

    Efficient Parallel ILU Factorization and Forward/Backward Substitution with Application to Large-Scale Nonlinear Circuit Simulation

    by Jiawen Cheng, Lingjie Li, Zhiqiang Liu, Kan Liu, Shan Shen, Zhenya Zhou, Wenjian Yu

    Efficient techniques are proposed for parallel incomplete LU (ILU) factorization and forward/backward substitution, for the sparse matrices with the same sparsity pattern. These parallel algorithms are then used as a preconditioner for the generalized minimal residual (GMRES) algorithm to obtain an ILU-GMRES solver for large-scale circuit simulation. The novelty of the parallel ILU and substitution algorithms includes a subtree-based task scheduling scheme, a nested dissection-based approach for generating task queues, and the task packing and reverse-order execution techniques for forward/backward substitution. Experiments on 43 matrices dumped from circuit simulation show that the 8-thread parallel ILU with threshold (ILUT) factorization and forward/backward substitution with the proposed techniques achieve 4.2 \(\times\) and 3.3 \(\times\) parallel speedups on average, respectively. The proposed parallel ILUT-GMRES solver runs 5.3 \(\times\) , on average, faster than PARDISO on these benchmarks. When integrated into Ngspice, it enables up to 2.3 \(\times\) and 1.7 \(\times\) faster execution of a step of Newton-Raphson iteration than the commercial parallel HSPICE, for the DC analysis and time integration stages, respectively.

  • Easyn: Resynthesis and Legalization Joint NIMCH Design Placement

    by Jiahao Xu, Yuan Pu, Yu Zhang, Fangzhou Liu, Wenjian Yu, Kai-Yuan Chao, Bei Yu

    Non-integer multiple cell height (NIMCH) libraries have been adopted by advanced manufacturers in the latest technology node, to offer better power efficiency, performance optimization, and area occupancy. However, in order to meet manufacturing requirements, such as power supply, NIMCH placement must comply with several additional layout restrictions. Currently, the state-of-the-art method applies resynthesis and legalization to the initial placement result to achieve a minimum displacement. However, the two decoupled phases lead to a reduction in the solution space and an insensitivity to timing. Therefore, we propose an iterative algorithm with corresponding data structure, which comprehensively considers displacement and timing information. Experiments demonstrate that our method can generate legalized NIMCH designs with a lower maximum delay and a reduced area.


  • A Dimensional Order Search Based Mapping Algorithm for Deploying DNNs onto Many-Core Neuromorphic IC

    by Song Wang, Rui Lu, Bo Xiao, Zhaohui Wang, Yunpeng Guo, Xin Sun, Wenjian Yu

    Many-core neuromorphic integrated circuits (ICs) have the potential advantages of low power consumption, high parallelism, etc. for the edge computing of deep learning. A key problem in the application of many-core neuromorphic IC is how to efficiently deploy the deep neural networks onto the core clusters. Notice that mapping the neural network clusters to physical cores is NP-hard due to the super-exponential growth of the search space. In this work, we propose a novel approach termed the Dimensional Order Search (DOS) based mapping algorithm, which integrates a branch-and-bound strategy to reduce the size of search space to no more than 4!·2!·8!. The DOS further employs a time-sharing protocol that eliminates deadlocks via priority-based channel arbitration. A key strength of DOS lies in its ability to preserve the optimality of mappings with increasing core allocations, backed by asymptotic convergence guarantees. DOS achieves a probability asymptotically approaching unity as core allocation scales. Furthermore, we provide a mathematical proof of the optimality-preserving property of the proposed mapping algorithm. Experimental results on a 160-core TianjicX neuromorphic platform demonstrate that DOS achieves significant improvements in routing efficiency compared to conventional methods. Specifically, DOS reduces mapping time for ResNet-50 by 3.9×, 5.55×, and 10.99× versus genetic algorithms (GA), simulated annealing (SA), and reinforcement learning (RL) approaches respectively. Similar gains are observed for VGG-16 (3.32×/ 5.96×/ 3.44×) and Inception-V3 (3.94×/ 7.37×/ 5.11×) when compared to GA, SA, and RL baselines. Our work bridges the gap between the complexity of theoretical mapping and practical implementations of many cores.