Question

请高人帮忙翻译一段话（有点难度），要专业的！翻译工具的就不要拿出来了！翻译好再加分！谢谢！

三）本次并行计算性能分析
本文的混波室属于电大尺寸，采用非均匀网格可以缩小其仿真规模，比例因子取1.3，我们计算了各种规模下的加速比和并行效率，其中规模的增大是通过增加网格数量和时间步数的方法来实现的，各计算规模大小如表2所示：
表2 不同网格划分产生的规模
表3 不同规模问题并行计算测试结果
由表3可见：
① 并行计算一个明显的优点就是大大缩短了运行时间，对于规模1，两机并行后时间
上较比串行缩短了4074s。规模2、3均相应的减少了近1倍的时间。可以看出，并行时域有限差分大大提高了运算速度，为具有时间限制的数值问题提供了解决的可能。本文中对混波室统计特性研究过程中，因取样值多，采取了规模1进行计算，在保证精度的同时，缩短计算时间。
② 两机获得较为理想的并行效率，81.3%～85%。规模2较之规模1，规模3较之规
模2，在网格数和时间步上有所增加，而加速比和并行效率均相应地提高。这是因为对着计算规模扩大，计算/通信比率升高，系统开销与计算开销分配的进一步合理，从而提高了并行效率。显而易见，当计算规模增大时，并行计算相对串行计算的优势逐渐的体现了出来。
正如第二章所讲，时域有限差分法耗用内存和子区域的体积是成正比的，因此区域分割对并行效率和加速比是有一定影响的。表4是不同区域划分方法进行了描述 ，其中时间步均取73172。表5则是对表4中不同区域分割后得到的并行性能的测试情况。
表4 不同区域分割情况

Answer 1

你好,希望能帮到你

C) The performance analysis of parallel computing
Reverberation-room of this article belongs to TV size, the use of non-uniform grid can reduce the size of the simulation, the ratio factor from 1.3, we calculate the size of a variety of speed-up ratio and parallel efficiency, which increases the size of the grid by increasing the number of and time-step approach to achieve the size of the calculation as shown in table 2:
Table 2 have a different mesh size
Table 3 the size of the different test results parallel computing
Can be seen from Table 3:
① parallel computing is an obvious advantage of greatly reducing the running time, for the size of 1, after two planes parallel time
On reducing the per pound Serial 4074s. 2,3 are the corresponding scale have decreased by almost 1 percent of the time. As can be seen, parallel finite-difference time-domain greatly enhanced the speed of operation, has a time limit for the numerical solution of the problem may be provided. In this article on the statistical characteristics of Reverberation-Chamber in the course of the study, due to the sampling value, and adopted a scale of 1 calculated at the same time ensuring the accuracy and shorten the computing time.
② two planes parallel to obtain better efficiency, 81.3% ~ 85%. 2 compared with the scale of the size of 1, compared with the size of regulation 3
Mode 2, the grid number and time step on the increase, while the speed-up ratio and parallel efficiency are improved accordingly. This is because the calculation at the scale of the expansion of computing / communication ratio increases, the system and calculation of overhead expenses to a reasonable allocation, thereby enhancing the efficiency of the parallel. Obviously, when calculating the size increases, the parallel calculation of the relative advantages of serial embodiment gradually out.
As the second chapter, the finite difference time domain method consume memory, and the promoter region is directly proportional to the volume, and therefore the regional split of the efficiency and speed-up ratio in parallel there is a certain effect. Table 4 is divided into different regions described methods, which are from time step 73172. Table 5 of Table 4 in the different regions after the partition of the parallel performance of the test case.
Table 4 partition situation in different regions.

Answer 2

C) The performance analysis of parallel computing

Reverberation-room of this article belongs to TV size, the use of non-uniform grid can reduce the size of the simulation, the ratio factor from 1.3, we calculate the size of a variety of speed-up ratio and parallel efficiency, which increases the size of the grid by increasing the number of and time-step approach to achieve the size of the calculation as shown in table 2:

Table 2 have a different mesh size

Table 3 the size of the different test results parallel computing

Can be seen from Table 3:

① parallel computing is an obvious advantage of greatly reducing the running time, for the size of 1, after two planes parallel time

On reducing the per pound Serial 4074s. 2,3 are the corresponding scale have decreased by almost 1 percent of the time. As can be seen, parallel finite-difference time-domain greatly enhanced the speed of operation, has a time limit for the numerical solution of the problem may be provided. In this article on the statistical characteristics of Reverberation-Chamber in the course of the study, due to the sampling value, and adopted a scale of 1 calculated at the same time ensuring the accuracy and shorten the computing time.

② two planes parallel to obtain better efficiency, 81.3% ~ 85%. 2 compared with the scale of the size of 1, compared with the size of regulation 3

Mode 2, the grid number and time step on the increase, while the speed-up ratio and parallel efficiency are improved accordingly. This is because the calculation at the scale of the expansion of computing / communication ratio increases, the system and calculation of overhead expenses to a reasonable allocation, thereby enhancing the efficiency of the parallel. Obviously, when calculating the size increases, the parallel calculation of the relative advantages of serial embodiment gradually out.

As the second chapter, the finite difference time domain method consume memory, and the promoter region is directly proportional to the volume, and therefore the regional split of the efficiency and speed-up ratio in parallel there is a certain effect. Table 4 is divided into different regions described methods, which are from time step 73172. Table 5 of Table 4 in the different regions after the partition of the parallel performance of the test case.

Table 4 partition situation in different regions