CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   Phoenics (http://www.cfd-online.com/Forums/phoenics/)
-   -   Problems in Parallel PHOENICS (http://www.cfd-online.com/Forums/phoenics/51158-problems-parallel-phoenics.html)

Zeng February 22, 2001 09:52

Problems in Parallel PHOENICS
 
I have encountered problems list below in parallel PHOENICS, any of your sugestion is appreciated. (1) phi file size is different between the simulation with single- and multi-CPU. (2)The simulation seems ok for the 401 case in VR library with single CPU, but the simulation is divergent by the parallel PHOENICS with multi-CPU, and of course its result is unbelievable. (3)I have coded ground.f, it works well for single CPU simulation, but it does not work after parallelisation. These is no error to build paraear, but the simulation seems to be suspended when it runs.

Has any one here experienced same problem in parallel PHOENICS? How do you overcome these problems? Is there any rules for coding ground.f with parallel calculation?

Thanks in advance

Zeng

George Bergantz February 22, 2001 13:38

Re: Problems in Parallel PHOENICS
 
I am running parallel PHOENICS and can perhaps provide a few useful comments. I may be dead wrong about some of these things, but here is what I've come to understand:

1) The phi file is only for the first processor. Each processor creates its own restart file. Hence visualization of the entire domain cannot be done with say PHOTON. Again, I may be wrong about this.

2) I can't help you with the divergence issues. Be sure that you appreciate that re-gridding to a higher number of nodes, as we often do to exploit the power of parallel, can impact convergence. But that may not be your problem.

3) If you are using MPI it is my understanding and experience that you cannot simply code ground.f as you would with a sequential algorithm. You must be mindful of coding and compilation in the context of the MPI commands and framework. That is too extensive a topic to be treated here, perhaps a CHAM engineer can weigh-in on that point. I have been given example ground files that CHAM has written for specific parallel problems where the source terms have been parallelized, and things are a bit different than the usual approach. For one thing you are working in the X or Y to Z plane so the usual GETYX, SETYX and other niceities of grabbing things from the F array are now moot.

Zeng February 22, 2001 22:40

Re: Problems in Parallel PHOENICS
 
Hi, George

Thanks. If it is possible, can I have a look at the example ground file, which you received from CHAM. I have sent my groud and q1 files to their agent two weeks before, and I am waiting their reply.

Thank again,

Zeng

Sandeep Karode February 27, 2001 14:28

Re: Problems in Parallel PHOENICS
 
I have encountered similar problems too!


All times are GMT -4. The time now is 00:57.