Lin Peng

Lin Peng

Auto-generated profile, based on publicly available information

Lin Peng

College of Computer Science and Technology, National University of Defense Technology, Changsha, China

Research Focus & Metrics

Academic Career Length

6Years (2021-Present)

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

Representative Publications

View All Publications
20 November 2024 in ACM Transactions on Architecture and Code Optimization
Verified
Journal Article
Open access
XLXiaobo Lu
Xiaobo Lu
Jianbin FangJianbin Fang
Jianbin Fang
LPLin Peng
Lin Peng
CHChun Huang
Chun Huang
ZDZidong Du
Zidong Du
ZWZheng Wang
Zheng Wang
Sparse-dense matrix multiplication (SpMM) is the performance bottleneck of many high-performance and deep-learning applications, making it attractive to design specialized SpMM hardware accelerators. Unfortunately, existing hardware solutions do not take full advantage of data reuse opportunities of the input and output matrices or suffer from irregular memory access patterns. Their strategies increase the off-chip memory traffic and bandwidth pressure, leaving much room for improvement. We present Mentor , a new approach to designing SpMM accelerators. Our key insight is that column-wise dataflow, while rarely exploited in prior works, can address these issues in SpMM computations. Mentor is a software-hardware co-design approach for leveraging column-wise dataflow to improve data reuse and regular memory accesses of SpMM. On the software level, Mentor incorporates a novel streaming construction scheme to preprocess the input matrix for enabling a streaming access pattern. On the hardware level, it employs a fully pipelined design to unlock the potential of column-wise dataflow further. The design of Mentor is underpinned by a carefully designed analytical model to find the tradeoff between performance and hardware resources. We have implemented an FPGA prototype of Mentor . Experimental results show that Mentor achieves speedup by geomean 2.05× (up to 3.98×), reduces the memory traffic by geomean 2.92× (up to 4.93×), and improves bandwidth utilization by geomean 1.38× (up to 2.89×), compared with the state-of-the-art hardware solutions.

1 August 2021 in Genome
Verified
Journal Article
Closed access
SZShengchi Zhang
Shengchi Zhang
YZYongzhe Zheng
Yongzhe Zheng
GZGuimin Zhang
Guimin Zhang
PLPeng Lin
Peng Lin
WWWei Wang
Wei Wang
The purpose of this study was to explore the relationship between autophagy and DNA methylation, and to identify key genes for autophagy-regulated thyroid cancer progression. We divided patients with thyroid cancer into high-autophagy score (AS) group and low-AS group based on their AS values. The results found that AS was associated with the distant metastasis of thyroid cancer, and adversely affected prognosis. Then, we screened 359 differently expressed genes (DEGs) with DNA methylation status consistent with gene expression change. Functional classification analysis demonstrated that the 359 DEGs consistent with DNA methylation status were significantly involved in adhesion, migration, and differentiation of immune cells. To further screen the key genes in the autophagy-related thyroid cancer progression, we constructed a protein–protein interactions (PPI) network and performed prognostic analysis. B cell linker (BLNK) was identified as the key potential gene affecting autophagy-related thyroid cancer progression. Finally, we verified that BLNK promoted the proliferation of thyroid cancer cells, and BLNK expression was regulated by DNA methylation. Our research provides a new perspective for exploring the relationship between autophagy and DNA methylation during the progression of thyroid cancer and provides a new target for the treatment of metastatic thyroid cancer.

30 June 2025 in ACM Transactions on Architecture and Code Optimization
Verified
Journal Article
Open access
XLXiaobo Lu
Xiaobo Lu
Jianbin FangJianbin Fang
Jianbin Fang
LPLin Peng
Lin Peng
CHChun Huang
Chun Huang
ZYZixiao Yu
Zixiao Yu
TLTiejun Li
Tiejun Li
Graph attention networks (GATs) have advanced performance in various application domains by introducing the attention mechanism into the graph neural networks (GNNs). The inefficiency of running GATs on CPUs or GPUs necessitates specialized hardware designs. Unfortunately, previous specialized architecture designs have focused on either the GNN architecture or the attention mechanism, resulting in limited performance and leaving ample room for improvement. This article presents Gator , a joint optimization approach with software–hardware co-designs for GAT inference. On the software level, Gator leverages degree-weighted graph partitioning and parameter-adaptive feature selection techniques to preprocess the input graph data, mining subgraph-level parallelism and mitigating the computation bottleneck of the dedicated dataflow. On the hardware level, Gator designs a unified processing engine to support various kernels by extracting a common computation pattern and a dimension-aware microarchitecture for efficient partial sum reduction. Extensive experiments show that our approach can achieve 11.5× more efficiency compared to NVIDIA RTX 4090 and provide a speedup of 3× to 9.4×, along with a 2.6× to 4.7× reduction in memory traffic, when compared to six state-of-the-art methods, with minimal accuracy loss.

17 May 2024 in CCF Transactions on High Performance Computing
Verified
Journal Article
Closed access
PDF file
JLJianfeng Liu
Jianfeng Liu
WGWanrong Gao
Wanrong Gao
HLHanzheng Liang
Hanzheng Liang
LPLin Peng
Lin Peng
TWTing Wang
Ting Wang

26 August 2024
Verified
Book Chapter
Closed access
PDF file
XFXiaokang Fan
Xiaokang Fan
ZGZhen Ge
Zhen Ge
SLSifan Long
Sifan Long
TTTao Tang
Tao Tang
CHChun Huang
Chun Huang
CYCanqun Yang
Canqun Yang

12 August 2024
Verified
Conference Paper
Open access
XWXianglin Wang
Xianglin Wang
XYXin Yi
Xin Yi
HYHengbiao Yu
Hengbiao Yu
CHChun Huang
Chun Huang
LPLin Peng
Lin Peng
Correctly rounded elementary mathematical functions are crucial for numerical computations and scientific applications. Generating these functions accurately is a challenging task. The latest methods automate this process by transforming the problem of generating correctly rounded elementary mathematical functions into a linear programming problem. However, this generation process is serial, and the inefficiency of serialization hinders the creation of new elementary mathematical functions and limits the broader application of the technique. This paper introduces a novel parallel method to accelerate the generation of correctly rounded elementary mathematical functions. The primary approach is to optimize the generation process through two dimensions: task decomposition and data parallelism, effectively mitigating data dependency and competition. This adaptation renders it suitable for conventional parallel models. We implemented our approach to generate six elementary functions. Our experimental results demonstrate that our parallel method can automate the generation of correctly rounded elementary math functions, achieving a speedup of 9.2x over the state-of-the-art approach, RLIBM.