[uosilos]

I have coded this switch to write and end the simulation when pressure reaches a defined value.

The code works nice in serial mode but in parallel it gets stuck while writing, it creates only a few fields in the processor0 folder and nothing in others proc folders.

What could be the problem?

pimple
{
...
rho = thermo.rho();
         
  runTime.write();
                    
Info<< "ExecutionTime = " << runTime.elapsedCpuTime() << " s"
            << "  ClockTime = " << runTime.elapsedClockTime() << " s"
            << nl << endl;
                
if (pressureOpening=="on")
    {

        #include "Panels.H"
                
        if (valueDiffAvg >= pStatAbs.value()) 
        {
            Info << "Aborting:  pressure threshold has been reached. \n" ;
            Info << "valueDiffAvg: " << valueDiffAvg << "   pStatAbs: " << pStatAbs.value() << endl ;
            
            runTime.writeAndEnd();
        }          
     }
 }
   Info<< "End\n" << endl;
 
  return 0;

[olesen]

Originally Posted by uosilos

I have coded this switch to write and end the simulation when pressure reaches a defined value.

The code works nice in serial mode but in parallel it gets stuck while writing, it creates only a few fields in the processor0 folder and nothing in others proc folders.

What could be the problem?

Your code very likely has different branching logic. I.e. some processes think they are finished, but others do not. The volField write however is parallel-aware, so if you only call it from one process things will simply "block". This usually looks like some really ugly MPI error at exit.



You probably want to test something like the following:

if (returnReduce((valueDiffAvg >= pStatAbs.value()), orOp<bool>())) 
{
     Info<< "Aborting:  pressure threshold has been reached" << nl
          << "valueDiffAvg: " << valueDiffAvg << "   pStatAbs: " << pStatAbs.value() << endl ;
            
     runTime.writeAndEnd();
}

Of course, without knowing what your code is, can't say if you want an andOp<bool>() logic, reduce the values ahead or whatever. Either way, I hope you get the idea.

[uosilos]

Originally Posted by olesen

Your code very likely has different branching logic. I.e. some processes think they are finished, but others do not. The volField write however is parallel-aware, so if you only call it from one process things will simply "block". This usually looks like some really ugly MPI error at exit.

Thanks a lot mate, you gave sense to the OF "community".


I wondered it should be something related with processors but my attempt was to include an if Pstream::master(), write, what neither work.


In a new attempt with your improvement seems that at least it is stopping and writing nicely, I will try further validation.


Thanks again