Close

%0 Conference Proceedings
%4 sid.inpe.br/plutao/2019/12.03.15.29.56
%2 sid.inpe.br/plutao/2019/12.03.15.29.57
%@doi 10.5753/wscad.2019.8657
%F lattes: 6700293516531228 1 FreitasMend:2019:RoAnPe
%T Roofline analysis and performance optimization of the MGB hydrological model
%D 2019
%A Freitas, Henrique Rennó de Azeredo,
%A Mendes, Celso Luiz,
%@affiliation Instituto Nacional de Pesquisas Espaciais (INPE)
%@affiliation Instituto Nacional de Pesquisas Espaciais (INPE)
%@electronicmailaddress henrique.renno@inpe.br
%@electronicmailaddress celso.mendes@inpe.br
%B Simpósio em Sistemas Computacionais de Alto Desempenho, 20
%C Campo Grande, MS
%8 16-18 out.
%I Sociedade Brasileira de Computação
%P 61-72
%S Anais
%K Performance evaluation, Application optimization, Roofline model, CPU/GPU parallelism.
%X The Roofline model gives insights about the performance behaviorof applications bounded by either memory or processor limits, providing usefulguidelines for performance improvements. This work uses the Roofline model onthe analysis of the MGB model that simulates hydrological processes in large-scale watersheds. Real-world input data are used to characterize the performanceon two multicore architectures, one with only CPUs and one with CPUs/GPU.The MGB model performance is improved with optimizations for better memoryuse, and also with shared-memory (OpenMP) and GPU (OpenACC) parallelism.CPU performance achieves42:51 %and50:17 %of each systems peak, whereasGPU performance is low due to overheads caused by the MGB model structure.
%@language en
%3 freitas_roofline.pdf
%U https://sol.sbc.org.br/index.php/wscad/article/view/8657


Close