Timing is a critical issue in electronic design automation (EDA). To reduce the delay of a net, an important strategy is to minimize the path lengths from the source to the sinks. However, minimizing the path lengths will inevitably sacrifice the total wirelength. To balance the two objectives, researchers use shallow-light tree (SLT) to model and optimize the problem. In this article, we introduce MALT, a novel approach that uses a neural network to guide the construction of Steiner shallow-light trees. The constructed trees are further refined by a dynamic programming-based branch merging algorithm, which improves the wirelength without sacrificing the path lengths of any sinks. Our experimental results demonstrate that the proposed framework achieves significant improvements over both state-of-the-art traditional SLT generation algorithms and existing machine learning enhanced methods.

The analysis of IR-drop stands as a fundamental step in optimizing the power distribution network (PDN), and subsequently influences the design performance. However, traditional IR-drop analysis using commercial tools proves to be exceedingly time-consuming. Fast and accurate IR-drop analysis is desperately in demand to achieve high performance on timing and power. Recently, machine learning approaches have garnered attention owing to their remarkable speed and extensibility in IC designs. However, prior works for dynamic IR-drop prediction presented limited performance since they did not exploit the time-varying activities. In this paper, we proposed a dual-path model with spatial-temporal transformers to extract the static spatial features and dynamic time-variant activities for dynamic IR drop prediction. Experimental results on the large-scale advanced dataset CircuitNet show that our model significantly outperforms the state-of-the-art works.

Routing is an important but complicated step in physical synthesis. Considering the potential of leveraging AI to seek higher efficiency and better quality in solving routing problems, we study in this work the methodology of AI-assisted routing in a systematic way. Decoupling the functionalities of different routing components will give a high flexibility in determining where and how AI can be used in an effective manner, while maintaining a high degree of interpretability. Two applications along this direction are presented, aiming at tackling the difficulties in routing with AI assistance. These provide examples of how to implement the methodology in practice, while revealing its effectiveness and potential.

Macro placement significantly influences the performance of the FPGA placement. However, constraints in modern designs like relative placement constraint (RPC) and regional constraint (RC) are often overlooked in existing routability-driven FPGA placers during macro placement. These constraints introduce challenges in optimizing routability during global placement and macro legalization stages. In this paper, we propose a novel macro placer that specifically addresses these constraints while optimizing routability. Our macro placer integrates macro size-aware pseudo nets, RC guided spreading, and multi-stage look-ahead legalization techniques to enhance routability with specified design constraints. Experimental results show that compared with DreamplaceFPGA-MP and the macro placer in Vivado, our proposed approach achieves 6% and 8% total routing score reduction on the MLCAD2023 contest benchmark. Moreover, the place and route time is reduced by 3.5% on average and up to 43% after our macro placer is integrated into Vivado. These compelling results demonstrate the efficiency gains and superior routability optimization achieved through our approach.

With the ever-increasing complexity of microprocessors, microarchitectural design becomes over-challenging. Design space exploration (DSE) of microarchitecture configurations to obtain high-quality designs with different PPA trade-offs is time-consuming, due to the huge configuration space and inefficient VLSI verification flow. Many DSE frameworks proposed in previous works failed to systematically analyze the contribution of each algorithmic component to the full flow. This paper provides a novel methodology for designing DSE frameworks by separating DSE flow into stages, and discussing algorithmic instantiations in each stage with theoretical and experimental analyses. Newly formulated DSE frameworks guided by this methodology achieve state-of-the-art results in ICCAD’22 DSE contest evaluation environments.

Due to the lengthy design cycle, generating legal, diverse and valid layout patterns artificially to expand VLSI layout pattern libraries has become an important problem to solve in order to facilitate modern design-for-manufacturability (DFM) studies. Considering the more realistic demands and to enhance functionality, this work proposes a style-controllable and violation-fixable layout pattern generation framework based on conditional diffusion models named ControLayout, which treats pattern category and complexity as conditions to control the style of generated patterns, and leverages the idea of image masking-inpainting to fix violations adaptively. Experiments reveal its promising performance in controllability and different metrics compared with the state-of-the-art methods.

With minimal compromises on other metrics, eliminating overflow and lowering congestion level of global routing results as much as possible is a crucial topic for reducing violations and hotspots in subsequent design phases. Different from current common practices of using maze routing according to some explicit orders to sequentially re-route particular nets of interest, this paper proposes a collaborative refinement framework that can generate multiple paths simultaneously to enlarge the solution space based on a multi-agent generative model, serving as a flexible post-processing plug-in on existing global routing results to reduce congestion. Experimental results well reveal its effectiveness.

Due to cost benefits, supply chains of integrated circuits (ICs) are largely outsourced nowadays. However, passing ICs through various third-party providers gives rise to many threats, like piracy of IC intellectual property or insertion of hardware Trojans, i.e., malicious circuit modifications.

Inverse lithography technology (ILT) is one of the promising resolution enhancement techniques (RETs) in modern design-for-manufacturing closure, however, it suffers from huge computational overhead and unaffordable mask writing time. In this paper, we propose A2-ILT, a GPU-accelerated ILT framework with spatial attention mechanism. Based on the previous GPU-accelerated ILT flow, we significantly improve the ILT quality by introducing spatial attention map and on-the-fly mask rectilinearization, and strengthen the robustness by Reinforcement-Learning deployment. Experimental results show that, comparing to the state-of-the-art solutions, A2-ILT achieves 5.06% and 11.60% reduction in printing error and process variation band with a lower mask complexity and superior runtime performance.

As the demand for semiconductor products increases and the integrated circuits (IC) processes become more and more complex, wafer failure pattern classification is gaining more attention from manufacturers and researchers to improve yield. To further cope with the real-world scenario that there are only very limited labeled data and without any unlabeled data in the early manufacturing stage of new products, this work proposes an efficient human-like staged learning framework for wafer failure pattern classification named WaferHSL. Inspired by human's knowledge acquisition process, a mutually reinforcing task fusion scheme is designed for guiding the deep learning model to simultaneously establish the knowledge of spatial relationships, geometry properties and semantics. Furthermore, a progressive stage controller is deployed to partition and control the learning process, so as to enable humanlike progressive advancement in the model. Experimental results show that with only 10% labeled samples and no unlabeled samples, WaferHSL can achieve better results than previous SOTA methods trained with 60% labeled samples and a large number of unlabeled samples, while the improvement is even more significant when using the same size of labeled training set.