[matteoL]

Hello,
I Have managed to compile succesfully OpenFoam on a Bluegene/P cluster with mpicc (which is a wrapper for gcc+mpi) and it works pretty well, at leats on the first case I used (rasinterFoam on 3.5 mln cells mesh).

I have 2 problems tough:

1)I then wanted to try it on a new mesh, same topology, just a bit more refined and with more cells, 5.5 mln cells.

For some reason, the code crashes. It crashes both when I try to convert the mesh (i use fluentMeshToFoam) and also when i run the rasinterdymFoam. (The same mesh gets decomposed and runs smootly on other standard pc/clusters.)

When I run fluentMeshToFoam it says that:
Building Patch'less mesh.... new cannot satisfy memory request.
This doesn't necessarily mean you run out of memory.. bla bla bla..

I am defineetely not run out of memory, since it was using just 4%of total memory.

When I run rasInterFoam (serial or parallel) it stops while trying to create the mesh. it says:
...
Create Mesh for Time=0
floating exception

...

Any idea about what could be the problem? why on another pc runs and bluegene not while on a very simialr ot works on both smoothly?

2)Has anyone managed to compile OF on bluegene/p but with the native compiler (i.e mpixlc)? when i try to do it I get plenty of errors. (Many double declarations of basic function/classes, i.e. dummy_scalar_functions.. and problems with Pstream class)

Any help would be much appreciated since it seems that a code compiled with mpixlc is about twice as fast as a code compiled with mpicc

Thanks,
Matteo