Liang Wang

Dr. Liang Wang


Research Focus & Metrics

Academic Career Length

16Years (2011-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
10 May 2023 in IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems
Verified
Journal Article
Closed access
MHMeng Han
Meng Han
LWLiang Wang
Liang Wang
LXLimin Xiao
Limin Xiao
HZHao Zhang
Hao Zhang
CZChenhao Zhang
Chenhao Zhang
JZJianfeng Zhu
Jianfeng Zhu
Point clouds have been employed extensively in machine perception applications. Farthest point sampling (FPS) is a critical kernel for point cloud processing. With the rapid growth of point cloud scale, FPS introduces a large number of memory accesses, which become the bottleneck of the large-scale point cloud processing. In this paper, we present QuickFPS, an architecture and algorithm co-design of farthest point sampling in large-scale point clouds. First, We systemically analyze the characteristics of FPS and put forward a bucket-based farthest point sampling algorithm. The algorithm introduces a two-level tree data structure to organize the large-scale point cloud into multiple buckets. By using two mechanisms named merged computation and implicit computation for the buckets, the external memory accesses and compute cost are significantly reduced. Then, we design an efficient domain-specific accelerator for farthest point sampling in large-scale point clouds. The accelerator takes advantage of different forms of parallelism and further improves the accelerator’s efficiency. Finally, we evaluate QuickFPS with several widely used point cloud datasets, which include small-scale and large-scale point clouds (up to 120,000 points). Overall, QuickFPS achieves performance speedups of 43.4x and 12.2x compared to GTX 1080Ti GPU and state-of-the-art point cloud accelerator PointAcc, respectively.

8 December 2015 in IEEE Transactions on Computers
Verified
Journal Article
Closed access
LWLiang Wang
Liang Wang
XWXiaohang Wang
Xiaohang Wang
TMTerrence Mak
Terrence Mak
Technology scaling leads to the reliability issue as a primary concern in Network-on-Chip (NoC) design. We observe that due to routing algorithm some routers age much faster than others which becomes a bottleneck for NoC lifetime. In this paper, lifetime is modeled as a resource consumed over time. A metric lifetime budget is associated with each router, indicating the maximum allowed workload for current period. Since the heterogeneity in router lifetime reliability has strong correlation with the routing algorithm, we define a problem to optimize the lifetime by routing packets along the path with maximum lifetime budgets. The problem is then extended for both performance and lifetime reliability optimization. The lifetime is optimized in long-term time scale while performance is optimized in short-term time scale. Two dynamic programming-based adaptive routing algorithms (lifetime aware routing and multi-objective routing) are proposed to solve the problems. In the experiments, the lifetime aware routing and multi-objective routing algorithms are evaluated with synthetic traffic and real benchmarks respectively. The experimental results show that the lifetime aware routing has around 20, 45 and 55 percent minimal lifetime improvement than XY routing, NoP routing and Oddeven routing, respectively. In addition, the multi-objective adaptive routing algorithm can effectively improve both performance and lifetime.

9 May 2024 in IEEE Transactions on Computers
Verified
Journal Article
Closed access
MHMeng Han
Meng Han
LWLiang Wang
Liang Wang
LXLimin Xiao
Limin Xiao
TCTianhao Cai
Tianhao Cai
ZWZeyu Wang
Zeyu Wang
CZChenhao Zhang
Chenhao Zhang
The systolic accelerator is one of the premier architectural choices for DNN acceleration. However, the conventional systolic architecture suffers from low PE utilization due to the mismatch between the fixed array and diverse DNN workloads. Recent studies have proposed flexible systolic array architectures to adapt to DNN models. However, these designs support only coarse-grained reshaping or significantly increase hardware overhead. In this study, we propose ReDas, a flexible and lightweight systolic array that supports dynamic fine-grained reshaping and multiple dataflows. First, ReDas integrates lightweight and reconfigurable roundabout data paths, which achieve fine-grained reshaping using only short connections between adjacent PEs. Second, we redesign the PE microarchitecture and integrate a set of multi-mode data buffers around the array. The PE structure enables additional data bypassing and flexible data switching. Simultaneously, the multi-mode buffers facilitate fine-grained reallocation of on-chip memory resources, adapting to various dataflow requirements. ReDas can dynamically reconfigure to up to 129 different logical shapes and 3 dataflows for a 128 × 128 array. Finally, we propose an efficient mapper to generate appropriate configurations for each layer of DNN workloads. Compared to the conventional systolic array, ReDas can achieve about 4.6× speedup and 8.3× energy-delay product (EDP) reduction.

11 August 2016 in Journal of Polymer Engineering
Verified
Journal Article
Closed access
XYXiao-chun Yin
Xiao-chun Yin
LWLiang Wang
Liang Wang
SLSai Li
Sai Li
GHGuangjian He
Guangjian He
ZYZhitao Yang
Zhitao Yang
JQJinping Qu
Jinping Qu
Various quantities of carbon fibers (CFs) (from 5% to 20% in weight) were added to matrix by melt blending to produce polylactic acid (PLA)/thermoplastic polyurethane (TPU)/CF composites. Differential scanning calorimetry measurements revealed that the CF content and mixing time had little influence on the crystallization and melting behavior of PLA. Thermogravimetric analysis showed that the introduction of CFs tended to decrease the thermal stability of PLA/TPU/CF composites, and the increase of mixing time tended to increase the thermal stability of PLA/TPU/CF composites when the mixing time is <5 min. Rheological results showed that all the samples exhibited non-Newtonian and shear thinning characteristics. The storage modulus and complex viscosity both increased with the increase of the CF content. It also showed that the increase of mixing time tended to increase the storage modulus and complex viscosity of PLA/TPU/CF composites when the mixing time is <5 min. Scanning electron microscopy images showed that the TPU/PLA blends contain a continuous PLA phase with evenly distributed TPU particles in the size range of 0.25–3 μm, and the blends are immiscible at the micron scale. Mechanical properties showed that the addition of proper CF content could lead to an obvious increase (about 11.43%) in tensile strength.

11 November 2020 in IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems
Verified
Journal Article
Closed access
YWYibo Wu
Yibo Wu
LWLiang Wang
Liang Wang
Xiaohang WangXiaohang Wang
Xiaohang Wang
JHJie Han
Jie Han
SYShouyi Yin
Shouyi Yin
LLLeibo Liu
Leibo Liu
Deadlock is a critical issue in faulty NoCs. Existing deadlock-free approaches on faulty NoCs suffer from low throughput and poor fairness when the network becomes oversaturated. This problem hinders their practical use as oversaturation scenarios are more frequent on faulty NoCs. To address this issue, a deflection-based deadlock recovery framework is proposed for higher over-saturation performance on faulty NoCs. First, we observe the low over-saturation performance of existing deadlock recovery approaches, and analyze the positive feedback loop that can amplify the negative impact of deadlocks and congestions, which necessitate handling both deadlocks and congestions in a deadlock recovery framework. Second, we propose a novel deadlock recovery framework, which includes an accurate, timely deadlock detection and a highly-efficient deadlock recovery. Both the deadlock detection and recovery reduce the average packet traversal latency, thereby improving the average over-saturation throughput. Third, we propose a distributed implementation to make the entire network enter and exit the deflection mode, which is conducted by broadcasting special messages via a bufferless subnetwork. An average oversaturation throughput improvement of 1.1 8.1× over state-ofthe-art approaches is achieved. In terms of fairness, the minimal over-saturation throughput is improved from near zero to half of the peak throughput.

1 October 2020 in Journal of Systems Architecture
Verified
Journal Article
Closed access
YZYiming Zhao
Yiming Zhao
XWXiaohang Wang
Xiaohang Wang
YJYingtao Jiang
Yingtao Jiang
LWLiang Wang
Liang Wang
MYMei Yang
Mei Yang
TMTerrence Mak
Terrence Mak