【摘 要】
：
开发了基于图形处理器(GPU)的Cholesky分解并行算法,应用于模态计算程序中,对计算进行加速.算例测试表明该算法相对串行算法计算性能大幅提升,且加速比随矩阵阶数增加而增加,与串行程序相比加速比可达到19.6,此时GPU浮点运算能力达到298Gflops.GPU程序固有频率计算结果与Abaqus计算结果的误差在2%以内,具有足够的计算精度.
论文部分内容阅读
开发了基于图形处理器(GPU)的Cholesky分解并行算法,应用于模态计算程序中,对计算进行加速.算例测试表明该算法相对串行算法计算性能大幅提升,且加速比随矩阵阶数增加而增加,与串行程序相比加速比可达到19.6,此时GPU浮点运算能力达到298Gflops.GPU程序固有频率计算结果与Abaqus计算结果的误差在2%以内,具有足够的计算精度.
其他文献
Land use change (LUC) is widely recognized as one of the most important driving forces of global carbon cycles. The soil organic carbon (SOC) and labile organic carbon (LOC) stores were investigated a
Identifying the flow patterns is vital for understanding the complicated physical mechanisms in multiphase flows.For this purpose,electrical capacitance tomography(ECT) technique is considered as a pr
Identification of the meta-instable stress state and study of its mechanism and evolution of relevant physical fields would be of great significance for determination of potential seismic risks and es
On 12 May 2008, the magnitude 8.0 Wenchuan Earthquake occurred along the Longmen Shan nappe, Sichuan, China. This devastating earthquake led to a heavy death toll of greater than 80,000. The seismic o
采用开顶式气室法(Open-top Chamber,OTC)模拟升温,分别于升温1a和升温2a后,测定崇明东滩围垦湿地代表植物——芦苇快速生长期叶片的光响应进程,分析最大净光合速率(Pmax)、暗呼吸速率(Rd)、表观量子效率(AQY)、光补偿点(LCP)、光饱和点(LSP)等光合参数对升温时间的累积响应。结果表明:与对照样地相比,升温1a后,Pmax、LSP显著增加,Rd、LCP显著降低,表现出
Fast inversion of source rupture process for significant earthquake is introduced and applied to earthquake emergency response. We retrieve and use the apparent source time functions (ASTFs) to stabil
Earthquake rupture process generally involves several faults activities instead of a single fault.A new method using both fuzzy clustering and principal component analysis makes it possible to reconst
Magnetic induction tomography(MIT) is one of the newest industrial process imaging techniques.Main industrial applications of the MIT imaging are in high conductive flow imaging.However,recently it ha
针对光弹性应力计条纹图像人工判读存在的问题,首先,在不同级数的条纹图像颜色变化最明显的方向上提取RGB颜色变化的色差向量作为样本,计算待识别条纹图像与样本空间的相关系数,得到初步识别的级数。其次,由于工程实际中条纹图像难以与样本完全一致,使得不同方向得到的初步识别的级数并不一定相同,因此,采用模糊推理方法对初步识别的级数进行综合分析,确定最终条纹图像的级数。实际应用表明,该方法具有较好的识别精度。
In this paper, a new design of multi-parallel-beam (MPB) collimators with projection multiplexing is proposed. In the MPB system, two different oblique parallel channels are introduced in a convention