Based on data on taxed-cropland area and on the number of households in historical documents, a probabilistic model of cropland distribution and a cropland area allocation model were designed and validated. Cropland a...Based on data on taxed-cropland area and on the number of households in historical documents, a probabilistic model of cropland distribution and a cropland area allocation model were designed and validated. Cropland areas for the years AD976, 997, 1066, and 1078 were estimated at the level of Lu（an administrative region of the Northern Song Dynasty）. The results indicated that（1） the cropland area of the whole study region for AD976, 997, 1066, and 1078 was about 468.27 million mu（a Chinese unit of area, with1 mu=666.7m2）, 495.53 million mu, 697.65 million mu, and 731.94 million mu, respectively. The fractional cropland area（FCA） increased from 10.7% to 16.8%, and the per capita cropland area decreased from 15.7 mu to 8.4 mu.（2） With regard to the cropland spatial pattern, the FCA of the southeast, north, and southwest regions of the Northern Song territory increased by 12.0%, 5.2%, and 1.2%, respectively. The FCA of some regions in the Yangtze River Plain increased to greater than 40%, and the FCA of the North China Plain increased to greater than 20%. However, the FCA of the southwest region（except for the Chengdu Plain） in the Northern Song territory was less than 6%.（3） There were 84.2% Lus whose absolute relative error was smaller than 20% in the mid Northern Song Dynasty. The validation results indicate that our models are reasonable and that the results of reconstruction are credible.展开更多
随着集成电路工艺的发展,众核处理器体系结构逐渐成为计算机体系结构设计者的研究热点。众核体系结构通过任务级的并行来提升整个处理器的性能。然而,指令级的并行性仍然是众核设计者需要认真考虑的问题。对浮点运算效率和加速比进行了...随着集成电路工艺的发展,众核处理器体系结构逐渐成为计算机体系结构设计者的研究热点。众核体系结构通过任务级的并行来提升整个处理器的性能。然而,指令级的并行性仍然是众核设计者需要认真考虑的问题。对浮点运算效率和加速比进行了形式化描述,验证了进行指令级调度的必要性。对处理器核内流水线进行详细分析,指出了指令级调度的一般性问题。提出了在众核结构上使用指令级调度和软件流水的方法。针对Splash2程序集里的LU分解算法,使用众核结构的硬件支持,在Scratched Pad Memory(SPM)上给出了调度指令的方案。在众核仿真器Godson-T上仿真了经过指令级调度后的算法,当使用64个线程处理512×512的矩阵时,程序性能达到调度前性能的4倍。展开更多
基金National Natural Science Foundation of China,No.41271227 The Special Program for Basic Work of the Ministry of Science and Technology,China,No.2014FY210900
文摘Based on data on taxed-cropland area and on the number of households in historical documents, a probabilistic model of cropland distribution and a cropland area allocation model were designed and validated. Cropland areas for the years AD976, 997, 1066, and 1078 were estimated at the level of Lu（an administrative region of the Northern Song Dynasty）. The results indicated that（1） the cropland area of the whole study region for AD976, 997, 1066, and 1078 was about 468.27 million mu（a Chinese unit of area, with1 mu=666.7m2）, 495.53 million mu, 697.65 million mu, and 731.94 million mu, respectively. The fractional cropland area（FCA） increased from 10.7% to 16.8%, and the per capita cropland area decreased from 15.7 mu to 8.4 mu.（2） With regard to the cropland spatial pattern, the FCA of the southeast, north, and southwest regions of the Northern Song territory increased by 12.0%, 5.2%, and 1.2%, respectively. The FCA of some regions in the Yangtze River Plain increased to greater than 40%, and the FCA of the North China Plain increased to greater than 20%. However, the FCA of the southwest region（except for the Chengdu Plain） in the Northern Song territory was less than 6%.（3） There were 84.2% Lus whose absolute relative error was smaller than 20% in the mid Northern Song Dynasty. The validation results indicate that our models are reasonable and that the results of reconstruction are credible.
文摘随着集成电路工艺的发展,众核处理器体系结构逐渐成为计算机体系结构设计者的研究热点。众核体系结构通过任务级的并行来提升整个处理器的性能。然而,指令级的并行性仍然是众核设计者需要认真考虑的问题。对浮点运算效率和加速比进行了形式化描述,验证了进行指令级调度的必要性。对处理器核内流水线进行详细分析,指出了指令级调度的一般性问题。提出了在众核结构上使用指令级调度和软件流水的方法。针对Splash2程序集里的LU分解算法,使用众核结构的硬件支持,在Scratched Pad Memory(SPM)上给出了调度指令的方案。在众核仿真器Godson-T上仿真了经过指令级调度后的算法,当使用64个线程处理512×512的矩阵时,程序性能达到调度前性能的4倍。
