NCTU-GR 2.0: Multithreaded Collision-Aware Global Routing With Bounded-Length Maze Routing

Liu, Wenhao; Kao, Wei-Chun; Li, Yih-Lang; Chao, Kai-Yuan

doi:10.1109/tcad.2012.2235124

Cited by 123 publications

(3 citation statements)

References 20 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The proposed SLDPSO-TA is implemented in C/C++ language on a Linux platform with a 2.6GHz CPU and 64GB RAM. To obtain global routing results as benchmarks for SLDPSO-TA, we adopt a global router NCTUgr [39] to route the placement results of NTUplace4 [40] that were obtained by placing test circuit DAC2021 [41]. Due to the large scale of the problem, we set M = 20, and iter = 400.…”

Section: Expeimental Resultsmentioning

confidence: 99%

SLDPSO-TA: Track Assignment Based on Social Learning Discrete Particle Swarm Optimization

Liu

Jing

Zhou

et al. 2023

Preprint

View full text Add to dashboard Cite

In the modern Very Large Scale Integration (VLSI) circuit design, the short-circuit problem is one of the key factors affecting routability. Due to the increased circuit size and net density, the short-circuit problem grows significantly in detailed routing. Furthermore, global routing ignores many problems in detailed routing, making for an increased mismatch between global routing and detailed routing. As a crucial intermediate phase between global routing and detailed routing, track assignment is an excellent phase to pre-process the short-circuit problem. However, existing track assignment algorithms face the problem of falling into local optimality. As one of the typical representatives of the swarm intelligence techniques, Particle Swarm Optimization (PSO) is a powerful tool to solve large-scale discrete problems. Therefore, we propose an effective Track Assignment Algorithm Based on Social Learning Discrete Particle Swarm Optimization (SLDPSO-TA). First, the proposed algorithm considers local nets to better guide detailed routing and an effective encoding method is designed to adapt the evolutionary algorithms better. Second, the social learning mode based on example pool mechanism is presented to improve the algorithm performance. Finally, a negotiation-based refining strategy is utilized to further reduce the overlap. Experimental results show that SLDPSO-TA can obtain the best overlap cost optimization among similar works, and reduce the congestion in key routing areas.

show abstract

Section: Expeimental Resultsmentioning

confidence: 99%

SLDPSO-TA: Track Assignment Based on Social Learning Discrete Particle Swarm Optimization

Liu

Jing

Zhou

et al. 2023

Preprint

View full text Add to dashboard Cite

show abstract

“…After performing 2D global routing, 2D solutions are extended to 3D solutions with layer assignment techniques. Most global routers adopt this two-step routing strategy and achieve high-performance routing results, such as NCTU-GR 2.0 [11], FastRoute 4.0 [12], NTU-GR [13] and NTHU-Route 2.0 [14]. However, these routers consider routes on a fixed placement result which does not allow cell movement.…”

Section: Previous Workmentioning

confidence: 99%

Detailed Placement and Global Routing Co-Optimization with Complex Constraints

Huang

Shi

et al. 2021

Electronics

View full text Add to dashboard Cite

With several divided stages, placement and routing are the most critical and challenging steps in VLSI physical design. To ensure that physical implementation problems can be manageable and converged in a reasonable runtime, placement/routing problems are usually further split into several sub-problems, which may cause conservative margin reservation and mis-correlation. Therefore, it is desirable to design an algorithm that can accurately and efficiently consider placement and routing simultaneously. In this paper, we propose a detailed placement and global routing co-optimization algorithm while considering complex routing constraints to avoid conservative margin reservation and mis-correlation in placement/routing stages. Firstly, we present a rapidly preprocessing technology based on R-tree to improve the initial routing results. After that, a BFS-based approximate optimal addressing algorithm in 3D is designed to find a proper destination for cell movement. We propose an optimal region selection algorithm based on the partial routing solution to jump out of the local optimal solution. Further, a fast partial net rip-up and rerouted algorithm is used in the process of cell movement. Finally, we adopt an efficient refinement technique to reduce the routing length further. Compared with the top 3 winners according to the 2020 ICCAD CAD contest benchmarks, the experimental results show that our algorithm achieves the best routing length reduction for all cases with a shorter runtime. On average, our algorithm can improve 0.7%, 1.5%, and 1.7% for the first, second, and third place, respectively. In addition, we can still obtain the best results after relaxing the maximum cell movement constraint, which further illustrates the effectiveness of our algorithm.

show abstract

“…This paper is aimed at improving the design of a global router by utilizing the conjoint optimization ability of DRL, which outperforms heuristic methods in terms of overflow and wire length [14]. In addition, this research seeks to surpass existing DRL-based global routers in terms of wire length, routability, and convergence performance.…”

Section: Introductionmentioning

confidence: 99%

An Enhanced Deep Reinforcement Learning-Based Global Router for VLSI Design

Yang

Liu

2023

Wireless Communications and Mobile Computing

View full text Add to dashboard Cite

Global routing is a crucial step in the design of Very Large-Scale Integration (VLSI) circuits. However, most of the existing methods are heuristic algorithms, which cannot conjointly optimize the subproblems of global routing, resulting in congestion and overflow. In response to this challenge, an enhanced Deep Reinforcement Learning- (DRL-) based global router has been proposed, which comprises the following effective strategies. First, to avoid the overestimation problem generated by Q -learning, the proposed global router adopts the Double Deep Q -Network (DDQN) model. The DDQN-based global router has better performance in wire length optimization and convergence. Second, to avoid the agent from learning redundant information, an action elimination method is added to the action selection part, which significantly enhances the convergence performance of the training process. Third, to avoid the unfair allocation problem of routing resources in serial training, concurrent training is proposed to enhance the routability. Fourth, to reduce wire length and disperse routing resources, a new reward function is proposed to guide the agent to learn better routing solutions regarding wire length and congestion standard deviation. Experimental results demonstrate that the proposed algorithm outperforms others in several important performance metrics, including wire length, convergence performance, routability, and congestion standard deviation. In conclusion, the proposed enhanced DRL-based global router is a promising approach for solving the global routing problem in VLSI design, which can achieve superior performance compared to the heuristic method and DRL-based global router.

show abstract

NCTU-GR 2.0: Multithreaded Collision-Aware Global Routing With Bounded-Length Maze Routing

Cited by 123 publications

References 20 publications

SLDPSO-TA: Track Assignment Based on Social Learning Discrete Particle Swarm Optimization

SLDPSO-TA: Track Assignment Based on Social Learning Discrete Particle Swarm Optimization

Detailed Placement and Global Routing Co-Optimization with Complex Constraints

An Enhanced Deep Reinforcement Learning-Based Global Router for VLSI Design

Contact Info

Product

Resources

About