Enabling Bitwise Reproducibility for the Unstructured Computational Motif

被引:0
|
作者
Siklosi, Balint [1 ]
Mudalige, Gihan R. [2 ]
Reguly, Istvan Z. [1 ]
机构
[1] Pazmany Peter Catholic Univ, Fac Informat Technol & Bion, H-1088 Budapest, Hungary
[2] Univ Warwick, Dept Comp Sci, Coventry CV4 7AL, England
来源
APPLIED SCIENCES-BASEL | 2024年 / 14卷 / 02期
关键词
floating-point; bitwise reproducibility; unstructured-mesh computation; DSL; CPU; GPU; MPI; NUMERICAL REPRODUCIBILITY; POINT;
D O I
10.3390/app14020639
中图分类号
O6 [化学];
学科分类号
0703 ;
摘要
In this paper we identify the causes of numerical non-reproducibility in the unstructured mesh computational motif, a class of algorithms commonly used for the solution of PDEs. We introduce a number of parallel and distributed algorithms to address nondeterminism in the order of floating-point computations, in particular, a new graph coloring scheme that produces identical coloring results regardless of how many parts the graph is partitioned to. We implement these in the OP2 domain specific language (DSL) and show how it can be automatically deployed to any application that uses OP2 without user intervention. We contrast differences in results without reproducibility and then demonstrate how bitwise reproducibility can be gained using our methods on a variety of applications including a production CFD application used at Rolls-Royce. We evaluate the performance and overheads of enforcing bitwise reproducibility on a cluster of CPUs and GPUs.
引用
收藏
页数:21
相关论文
共 50 条
  • [1] Enabling the Verification of Computational Results An Empirical Evaluation of Computational Reproducibility
    Stodden, Victoria
    Krafczyk, Matthew S.
    Bhaskar, Adhithya
    PROCEEDINGS OF THE 1ST INTERNATIONAL WORKSHOP ON PRACTICAL REPRODUCIBLE EVALUATION OF COMPUTER SYSTEMS (P-RECS'18), 2018,
  • [2] Bitwise Reproducible task execution on unstructured mesh applications
    Siklosi, Balint
    Reguly, Istvan Z.
    Mudalige, Gihan R.
    2020 20TH IEEE/ACM INTERNATIONAL SYMPOSIUM ON CLUSTER, CLOUD AND INTERNET COMPUTING (CCGRID 2020), 2020, : 889 - 892
  • [3] Bitwise Signature Comparison: Enabling More Efficient Similarity Estimation
    Reviriego, Pedro
    Pontarelli, Salvatore
    Martinez, Jorge
    IEEE TRANSACTIONS ON EMERGING TOPICS IN COMPUTING, 2023, 11 (03) : 798 - 804
  • [4] Reproducible Software Environment: a tool enabling computational reproducibility in geospace sciences and facilitating collaboration
    Bhatt, Asti
    Valentic, Todd
    Reimer, Ashton
    Lamarche, Leslie
    Reyes, Pablo
    Cosgrove, Russell
    JOURNAL OF SPACE WEATHER AND SPACE CLIMATE, 2020, 10
  • [5] Behavioural bitwise scheduling based on computational effort balancing
    Molina, MC
    Ruiz-Sautua, R
    Mendías, JM
    Hermida, R
    DESIGN, AUTOMATION AND TEST IN EUROPE CONFERENCE AND EXHIBITION, VOLS 1 AND 2, PROCEEDINGS, 2004, : 684 - 685
  • [6] Bitwise scheduling to balance the computational cost of behavioral specifications
    Molina, MC
    Ruiz-Sautua, R
    Mendías, JM
    Hermida, R
    IEEE TRANSACTIONS ON COMPUTER-AIDED DESIGN OF INTEGRATED CIRCUITS AND SYSTEMS, 2006, 25 (01) : 31 - 46
  • [7] Bitwise identical compiling setup: prospective for reproducibility and reliability of Earth system modeling
    Li, R.
    Liu, L.
    Yang, G.
    Zhang, C.
    Wang, B.
    GEOSCIENTIFIC MODEL DEVELOPMENT, 2016, 9 (02) : 731 - 748
  • [8] Computational reproducibility in computational social science
    Schoch, David
    Chan, Chung-hong
    Wagner, Claudia
    Bleier, Arnim
    EPJ DATA SCIENCE, 2024, 13 (01)
  • [9] The (computational) means and the motif
    Jenny Bangham
    Nature Reviews Genetics, 2005, 6 : 161 - 161
  • [10] Reproducibility of Computational Models
    Erdemir, A.
    Sauro, H. M.
    IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, 2016, 63 (10) : 1995 - 1996