[NewKid]

While running decomposePar in a case on 360 cores, everything is OK.
But while running this command on 420 or 480 cores, it reports in some subdomains there is no points and cells, then sprayFoam solver can not go on after it shows the radiation model.
The decomposeParDict file is shown as follows:

numberOfSubdomains 480;

method simple;

simpleCoeffs
{
n (20 6 4);// 2);//(4 2 11);//(4 11 2);//(7 4 4);// 1 8);
delta 0.001;
}

hierarchicalCoeffs
{
n (1 1 1);
delta 0.001;
order xyz;
}

metisCoeffs
{
processorWeights
(
1
1
1
);
}

multiLevelCoeffs

{

level0 //node级别
{
numberOfSubdomains 2; //区域的数目，我笔记本4核心单cpu的，只有一个node
method scotch;
}

level1//core级别
{
numberOfSubdomains 4; //子区域数目4，因为有4个核心
method scotch;//使用scotch方法
}

}

manualCoeffs
{
dataFile "";
}

distributed no;

roots
(
);

[simrego]

Hi!


Maybe this is a silly guess but try to use the scotch method for decomposition, not the simple one. Maybe you have some cores with 0 cells because of the simple method.
Also scotch will give you a better balance for the cores.
(Or you can use pscotch if you compiled it.)

[NewKid]

Hi, simrego. Thanks so much for your suggestion!
What bothers me is that while I run decomposePar of OF 4.0 instead of OF 3.0 on my machine, everything is fine, no cells never happens.
And I use multiLevel instead of simple, which is said to fit the multi-nodes computation. And in the different levels (node and core), the decompose method is scotch.
Everything is OK now.
Anyway, thank you

Regards!