QQ截图20190911102057.png
再补充一下,用不同的电脑跑了 damBreak 算例,得到的三个 ExecutionTime 和 ClockTime 。 并不总是 clockTime 大于 ExecutionTime 。@wwzhao
但是仔细检查发现,造成右边两个有差异的主要原因是两个时间的精度问题,如下所示:
ExecutionTime = 1.58 s ClockTime = 1 s
ExecutionTime = 1.59 s ClockTime = 1 s
ExecutionTime = 1.59 s ClockTime = 1 s
ExecutionTime = 1.6 s ClockTime = 1 s
ExecutionTime = 1.6 s ClockTime = 1 s
ExecutionTime = 1.61 s ClockTime = 1 s
ExecutionTime = 1.62 s ClockTime = 1 s
ExecutionTime = 1.62 s ClockTime = 1 s
ExecutionTime = 1.63 s ClockTime = 1 s ExecutionTime = 1.64 s ClockTime = 1 s
ExecutionTime = 1.64 s ClockTime = 2 s
ExecutionTime = 1.65 s ClockTime = 2 s
ExecutionTime = 1.66 s ClockTime = 2 s
ExecutionTime = 1.66 s ClockTime = 2 s
ExecutionTime = 1.67 s ClockTime = 2 s
ExecutionTime = 1.67 s ClockTime = 2 s
ExecutionTime = 1.68 s ClockTime = 2 s
因为是一个很小的单核算例,IO也不大,所以IO基本上没有花时间,导致在保留精度不同的情况下出现了上述差异。
多说两句:
上图右上角是在WSL2 根目录下的结果,
右下角是在一个 Ubuntu18.0.4 服务器上的结果,
左边的结果是用 WSL2 在 /mnt/h 盘下的结果,
可以看到在即使是很小的算例。这个IO耗时也是相当大。这是 WSL2目前的问题,相较于WSL 它在 根目录下确实IO能力有很大提升。但是在 /mnt/* 下 真是果断弃疗。
版本:
OpenFOAM-v7
The 'ExecutionTime' is the elapsed CPU time, and the ClockTime is the elapsed wall clock time for the latest time step(approximate!!)
https://www.cfd-online.com/Forums/openfoam-solving/67695-difference-between-executiontime-clocktime.html