[BalanceChen]

I am using the LINUX CFX12.1 parallel computation to modeling an axial compressor flow. The total grid elements is about nine million. Eight computers were used with a HP MPI distributed parallel.

The problem is that the computation stopped at about 40th step, with no error pop out. I am sure the CPU and memory is sufficient for the computation, but why i got the interruption?

Another problem is that should I do anything when I manually stopped a MPI parallel computation. I find that, the computation become slow when it is started after a parallel computation stopped.

Thanks firstly,

BalanceChen

[BalanceChen]

I choose the partition type to "uniform" and then the problem dissappears~

[mohammad]

Originally Posted by BalanceChen

I am using the LINUX CFX12.1 parallel computation to modeling an axial compressor flow. The total grid elements is about nine million. Eight computers were used with a HP MPI distributed parallel.

The problem is that the computation stopped at about 40th step, with no error pop out. I am sure the CPU and memory is sufficient for the computation, but why i got the interruption?

Another problem is that should I do anything when I manually stopped a MPI parallel computation. I find that, the computation become slow when it is started after a parallel computation stopped.

Thanks firstly,

BalanceChen

Hi BalanceChen,
I have a question to ask.

I'm using the same CFX and I want to use a supercomputer with 4nodes*4 ppn= 16
my problem is that when I use serial it solve about 12 iterations per hour.

when I use any parallel or Distributed parallel the speed is slowerrrr.. 2 iterations per hour.

do you know why?

regards.