面向众核系统的层次化栅栏同步机制

计算机工程与科学 ›› 2022, Vol. 44 ›› Issue (11): 1901-1908.

面向众核系统的层次化栅栏同步机制

臧照虎，李晨，王耀华，陈小文，郭阳

(国防科技大学计算机学院，湖南长沙 410073)

收稿日期:2021-11-22 修回日期:2022-03-22 接受日期:2022-11-25 出版日期:2022-11-25 发布日期:2022-11-25
基金资助:
国防科技大学科研计划（ZK20-04）

A hierarchical hardware barrier synchronization design for many-core processors

ZANG Zhao-hu，LI Chen，WANG Yao-hua，CHEN Xiao-wen，GUO Yang

（College of Computer Science and Technology，National University of Defense Technology，Changsha 410073,China）

Received:2021-11-22 Revised:2022-03-22 Accepted:2022-11-25 Online:2022-11-25 Published:2022-11-25

摘要/Abstract

摘要： 同步操作在保证多核处理器线程的数据一致性和正确性等方面起着重要作用。随着处理器内核数量的不断增加，同步操作的开销也越来越大。栅栏同步是并行应用中多核同步的重要方法之一。软件同步方法通常需要数千个周期才能完成多个内核之间的同步，这种高延迟和串行化同步会导致多核程序性能的显著下降。相比于软件栅栏同步方法，硬件栅栏能够实现较低的同步延迟，然而传统集中式硬件栅栏的可扩展性有限，难以适应众核处理器系统的同步需求。面向众核处理器提出了一种层次化硬件栅栏机制——HSync,它由本地栅栏单元和全局栅栏单元组成,二者协调配合,以实现低硬件开销的快速同步。实验结果表明，与传统的集中式硬件栅栏相比，层次化硬件栅栏机制将众核处理器系统性能提高了1.13倍，同时网络流量减少了74%。

关键词: 硬件同步, 栅栏, 众核系统, 并行计算

Abstract: Synchronization plays an important role in ensuring data consistency and correctness of multicore processor threads. As the number of processor cores increases, the cost of synchronization increases. Barrier synchro-nization is one of the effective methods for multi-core synchronization in parallel applications. Software synchronization methods typically require thousands of cycles to complete synchronization among multiple cores. This high latency and serialization synchronization can result in significant performance degradation of multicore programs. Compared with the software barrier synchronization method, the hardware barrier can achieve lower synchronization delay, but the scalability of the centralized hardware barrier is limited and it is difficult to adapt to the multicore processor systems. This paper proposes a hierarchical hardware barrier mechanism called HSync for multicore processors. It consists of local and global barrier units, which work together to achieve fast synchronization with low hardware overhead. The experimental results show that the hierarchical hardware barrier mechanism improves the performance of the multicore proces-sor system by 1.13 times and reduces network traffic by 74% compared with the traditional centralized hardware barrier.

Key words: hardware synchronization, barrier, many-core processors, parallel computing

臧照虎, 李晨, 王耀华, 陈小文, 郭阳. 面向众核系统的层次化栅栏同步机制[J]. 计算机工程与科学, 2022, 44(11): 1901-1908.

ZANG Zhao-hu, LI Chen, WANG Yao-hua, CHEN Xiao-wen, GUO Yang . A hierarchical hardware barrier synchronization design for many-core processors[J]. Computer Engineering & Science, 2022, 44(11): 1901-1908.

参考文献［24］

［1］	Scott S L. Synchronization and communication in the multiprocessor［C］∥Proc of the 7th International Conference on Architectural Support for Programming Languages and Operating Systems,1996:26-36.
［2］	Beck B,Kasten B,Thakkar S.VLSI assist for a multiprocessor［J］.SIGARCH Computer Architecture News,1987,15(5):10-20.
［3］	Hensgen D,Finkel R,Manber U.Two algorithms for barrier synchronization［J］.International Journal of Parallel Programming,1998,17(1):1-17.
［4］	Mellor-Crummey J M,Scott M L.Algorithms for scalable synchronization on shared-memory multiprocessors［J］.ACM Transactions on Computer Systems,1991,9(1):21-65.
［5］	Luchangco V,Nussbaum D,Shavit N.A hierarchical CLH queue lock［C］∥Proc of European Conference on Parallel Processing,2006:801-810.
［6］	Yang M,Wieder A,Brandenburg B B.Global real-time semaphore protocols:A survey unified analysis and comparison［C］∥Proc of 2015 IEEE Real-Time Systems Symposium,2015:1-19.
［7］	Ahn J,Hong S,Yoo S,et al.A scalable processing-in-memory accelerator for parallel graph processing［C］∥Proc of the 42nd Annual International Symposium on Computer Architecture,2015:105-117.
［8］	Sampson J,Gonzalez R,Collard J F,et al.Exploiting fine-grained data parallelism with chip multiprocessors and fast barriers［C］∥Proc of the 39th Annual IEEE/ACM International Symposium on Microarchitecture,2006:235-246.
［9］	Lei Z W,Ding H,Xiong H,et al.Design and realization of synchronization technique for FH-π/4-DQPSK communication system［C］∥Proc of the 13th IEEE Conference on Industrial Electronics and Applications,2018:2533-2538.
［10］	Liang C K,Prvulovic M.Misar:Minimalistic synchronization accelerator with resource overflow management［J］.ACM SIGARCH Computer Architecture News,2015,43(3S):414-426.
［11］	Vallejo E,Beivide R,Cristal A,et al.Architectural support for fair reader-writer locking［C］∥Proc of the 43rd Annual IEEE/ACM International Symposium on Microarchitecture,2010:275-286.
［12］	Marongiu A,Benini L,Kandemir M.Lightweight barrier-based parallelization support for non-cache-coherent MPSoC platforms［C］∥Proc of the 2007 International Conference on Compilers Architecture and Synthesis for Embedded Systems,2007:145-149.
［13］	Krishnan V,Torrellas J.The need for fast communication in hardware-based speculative chip multiprocessors［C］∥Proc of 1999 International Conference on Parallel Architectures and Compolation Techniques,1999:24-33.
［14］	Zhu W, Sreedhar V C, Hu Z,et al.Synchronization state buffer:Supporting efficient fine-grain synchronization on many-core architectures［C］∥Proc of the 34th Annual International Symposium on Computer Architecture,2007:35-45.
［15］	Zeng K W,Ning M N,Wang Y H,et al.Hierarchical clustering with hard-batch triplet loss for person re-identification［C］∥Proc of the IEEE/CVF Conference on Computer Vision and Pattern Recognition,2020:13657-13665.
［16］	Chen Xiao-wen.Dual channel fast barrier synchronization mechanism based on cooperative communication ［EB/OL］.［2017-05-12］.https://www.doc88.com/p-5307498419896. html.(in Chinese)
［17］	Shang S,Hwang K.Distributed hardwired barrier synchronization for scalable multiprocessor clusters［J］.IEEE Transactions on Parallel and Distributed Systems,1995,6(6),591-605.
［18］	Villa O,Palermo G,Silvano C.Efficiency and scalability of barrier synchronization on NoC based many-core architectures［C］∥Proc of the 2008 International Conference on Compilers Architectures and Synthesis for Embedded Systems,2008:81-90.
［19］	Hsu W T Y,Yew P C.An effective synchronization network for hot-spot accesses［J］.ACM Transactions on Computer Systems,1992,10(3):167-189.
［20］	Monchiero M,Palermo G,Silvano C,et al.An efficient synchronization technique for multiprocessor systems on-chip［C］∥Proc of the 2005 Workshop on Memory Performance:Dealing with Applications Systems and Architecture,2005:33-40.
［21］	Giannoula C, Vijaykumar N, Papadopoulou N,et al. SynCron:Efficient synchronization support for near-data- processing architectures［C］∥Proc of 2021 IEEE International Symposium on High-Performance Computer Architecture,2021:263-276.
［22］	Hetland C,Tziantzioulis G,Suchy B,et al.Paths to fast barrier synchronization on the node［C］∥Proc of the 28th International Symposium on High-Performance Parallel and Distributed Computing,2019:109-120.
［23］	Liu Chang， Guo Yang.Instruction set verification method for high performance DSP based on coverage-driven［J］.Computer Engineering，2014，40(6)：317-320.(in Chinese)
［24］	Chen Hai-yan， Guo Yang， Chen Ji-hua. Computer aided design and verification practice of integrated circuit［M］.Changsha：National University of Defense Technology Press,2010.(in Chinese)
	附中文参考文献：
［16］	陈小文.基于协同通信的双通道快速栅栏同步机制［EB/OL］.［2017-05-12］.https://www.doc88.com/p-5307498419- 896.html.
［23］	刘畅，郭阳.基于覆盖率驱动的高性能DSP指令集验证方法［J］.计算机工程，2014,40(6)：317-320.
［24］	陈海燕，郭阳，陈吉华.集成电路计算机辅助设计与验证实践［M］.长沙：国防科技大学出版社，2010.