GooFit: A library for massively parallelising maximum-likelihood fits

20th International Conference on Computing in High Energy and Nuclear Physics

物理学;计算机科学

Andreassen, R.^1 ; Meadows, B.T.^1 ; De Silva, M.^1 ; Sokoloff, M.D.^1 ; Tomko, K.^2

University of Cincinnati, Physics Department, ML0011, Cincinnati

OH

45221-0011, United States^1

Ohio Supercomputer Center, 1224 Kinnear Road, Columbus

OH

43212, United States^2

关键词: Large datasets;    Multi-core cpus;    Parallelisation;    Probability density functions (PDFs);    Real-world problem;   

Others  :  https://iopscience.iop.org/article/10.1088/1742-6596/513/5/052003/pdf
DOI  :  10.1088/1742-6596/513/5/052003

学科分类：计算机科学（综合）

来源: IOP

【 摘 要 】

Fitting complicated models to large datasets is a bottleneck of many analyses. We present GooFit, a library and tool for constructing arbitrarily-complex probability density functions (PDFs) to be evaluated on nVidia GPUs or on multicore CPUs using OpenMP. The massive parallelisation of dividing up event calculations between hundreds of processors can achieve speedups of factors 200-300 in real-world problems.