Reliability prediction of GPGPU programs under soft error perturbations presents significant challenges, particularly when fault injection samples are limited. This paper proposes a novel method for constructing a high-accuracy reliability prediction model using a small number of fault injection samples. The approach introduces guiding rules to optimize fault injection experiments by reducing redundancy and enhancing the quality of collected samples. Furthermore, data augmentation techniques are employed to mitigate the challenges of insufficient training data, enabling effective model learning. Experimental validation on various benchmark programs demonstrates that the proposed method achieves an average prediction accuracy of 91.56% with only 300 fault injection samples. These results highlight the method's capability to reduce experimental overhead while maintaining high prediction performance.
每个线程有独立的寄存器文件用于存储局部数据,操作数通过操作数收集器(Operand collector)准备完毕后,warp中的32个并行线程按照指令的操作类型分配到不同的计算单元中并行执行,包括流处理器(Streaming processor, SP)、特殊函数计算单元(Special function unit, SFU)以及加载存储单元(Load store unit, LSU)。SM还配备了私有的L1数据缓存(L1 Data Cache),用于加速高频数据的访问,并提供软件管理的共享内存(Shared memory)以支持block内线程间的数据通信和共享。
LiA, SongS L, WijtvlietM, et al. Sfu-driven transparent approximation acceleration on gpus[C]//2016 International Conference on Supercomputing, Istanbul, Turkey, 2016: 1-14.
[2]
BuckI, FoleyT, HornD, et al. Brook for GPUs: stream computing on graphics hardware[J]. Acm Transactions on Graphics, 2004, 23(3): 777-786.
[3]
ZhangBo-yuan, TianJian-nan, DiSheng, et al. FZ-GPU: A Fast and high-ratio lossy compressor for scientific computing applications on GPUs[C]//32nd ACM International Symposium on High-Performance Parallel and Distributed Computing, New York, USA, 2023:129-142.
[4]
AndrewL, ScottG. Fast algorithms for convolutional neural networks[C]//2016 IEEE Conference on Computer Vision and Pattern Recognition, Las Vegas, NV, USA, 2016: 4013-4021.
[5]
GhorbelA, Ben AmorN, AbidM. GPGPU-Based parallel computing of viola and jones eyes detection algorithm to drive an intelligent wheelchair[J]. Journal of Signal Processing Systems, 2022, 94(12): 1365-1379.
[6]
AiquipaW A, CarpioJ D, GarciaJ, et al. Analysis of high resolution panchromatic satellite images, based on GPGPU programming[C]//2nd International Conference on Sensors, Signal and Image Processing, Prague, Czech Republic, 2019: 45-48.
YuQi-rui, ZhangZhan-gang, LiBin, et al. Trends in the contribution of atmospheric neutrons and α particles to the chip soft errors [J]. Electronics and Packaging, 2023, 23(8): 70-76.
[9]
TanJ, GoswamiN, LiT, et al. Analyzing soft-error vulnerability on GPGPU microarchitecture[C]//2011 IEEE International Symposium on Workload Characterization, Austin, TX, USA, 2011:226-235.
[10]
WeiX, JiangN, YueH, et al. ApproxDup: developing an approximate instruction duplication mechanism for efficient SDC detection in GPGPUs[J]. IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems, 2024, 43(4): 1051-1064.
[11]
YangL, NieB, JogA, et al. Enabling software resilience in GPGPU applications via partial thread protection[C]//43rd IEEE/ACM International Conference on Software Engineering, Madrid, Spain, 2021: 1248-1259.
[12]
LiG, PattabirmanK, CherC Y, et al. Understanding error propagation in GPGPU applications[C]//2016 International Conference for High Performance Computing, Networking, Storage and Analysis, Salt Lake City, UT, USA, 2016: 240-251.
[13]
FangB, PattabiramanK, RipeanuM, et al. GPU-Qin: A methodology for evaluating the error resilience of GPGPU applications[C]//2014 IEEE International Symposium on Performance Analysis of Systems and Software, Monterey, CA, USA, 2014: 221-230.
[14]
HariS K S, TsaiT, StephensonM, et al. SASSIFI: an architecture-level fault injection tool for GPU application resilience evaluation[C]//2017 IEEE International Symposium on Performance Analysis of Systems and Software, Santa Rosa, CA, USA, 2017: 249-258.
[15]
TsaiT, HariS K S, SullivanM, et al. NVBitFI: dynamic fault injection for GPUs[C]//51st Annual IEEE/IFIP International Conference on Dependable Systems and Networks, Taipei, China, 2021: 284-291.
YueHeng-shan. Research on soft error resilience analysis and fault-tolerance strategy for GPGPU architecture[D]. Changchun: College of Computer Science and Technology, Jilin University, 2022.
[18]
LindholmE, NickollsJ, ObermanS, et al. NVIDIA Tesla: a unified graphics and computing architecture[J]. IEEE Micro, 2008, 28(2): 39-55.
[19]
SangchoolieB, PattabiramanK, KarlssonJ. One bit is (Not) enough: an empirical study of the impact of single and multiple bit-flip errors[C]//47th Annual IEEE/IFIP International Conference on Dependable Systems and Networks, Denver, CO, USA, 2017: 97-108.
[20]
AnwerA R, LiG, PattabiramanK, et al. GPU-Trident: efficient modeling of error propagation in gpu programs[C]//International Conference for High Performance Computing, Networking, Storage and Analysis, Atlanta, Georgia, USA, 2020: 1-15.
MaJun-chi, WangYun, CaiZhen-bo, et al. An Approach for Identifying SDC-causing instructions by fault propagation analysis[J]. Journal of Computer Research and Development, 2016, 53(9): 1943-1952.
[23]
CRAMÉRH. Mathematical methods of statistics[M]. New Jersey: Princeton University Press, 1946.
[24]
BengioY. Learning deep architectures for AI[J]. Foundations and Trends in Machine Learning, 2009, 2(1): 1-127
[25]
BreimanL, FriedmanJ H, OlshenR A, et al. Classification and Regression Trees[J]. Belmont: Wadsworth, 1984.