【摘要】 本文针对传统应急评估软件计算速度偏慢、评估时间过长的问题,利用GPU加速计算技术,开展应急快速评估算法的并行化研究。在分析串行评估算法性能瓶颈的基础上,运用计算任务并行化和数据处理并行化的方法,提出了基于CPU-GPU混合架构的并行化评估模型,给出了分区和分层的数据并行处理方案。与传统的串行评估模型相比,并行评估模型可以充分发挥当前主流计算机的计算能力,计算速度更快,数据处理能力更强,更适合震后应急救援工作的实际需求。该模型经软件优化后,可大大缩短震后快速评估所需的时间,为震后早期决策提供更为及时有效的支持。更多还原

【Abstract】 Most traditional rapid seismic assessment software are suffering from slow computing speed and long evaluation time. To solve this problem, the authors use GPU accelerated computing technology to develop parallel algorithms for rapid seismic assessment in this paper. We analyzed the performance bottlenecks of serial evaluation algorithms at first. Then a parallel assessment model based on CPU-GPU hybrid architecture was proposed by using the methods of computing task parallelism and data processing parallelism, and a data parallel processing scheme based on partitioning and layering was given. Compared with traditional serial assessment models, the parallel assessment model can make full use of the computing power of the current mainstream computer. With faster calculation speed and stronger data processing ability, this model is more suitable for the actual needs of post-earthquake emergency and rescue.After optimization in software, the proposed model can greatly shorten the time required for rapid post-earthquake assessment, which will provide more timely and efficient support for early post-earthquake decision-making.更多还原