Page 1 of 1

problems about compiling

Posted: Wed Jun 18, 2008 7:25 am
by dkkim
I complied the OpenSees ver. 1.7.5 to a parallel machine of my univ.
So, It works with OpenSees Multiple interpreter. thanks

by the way I have some questions

1.
I used the analysis system, ParallelProfileSPD.
but because my model is big(over 2000 elements for a processor), it takes much time to analze.

so I downloaded and compiled Petsc and Mumps to my system.
As I compiled the OpenSees with Petsc and Mumps, some errors(below lines) occured.

"PetscSOE.cpp", line 251.42: 1540-0256 (S) A parameter of type "_p_Mat **" cannot be initialized with an expression of type "int".
"PetscSOE.cpp", line 251.42: 1540-1205 (I) The error occurred while converting to parameter 2 of "MatCreate(MPI_Comm, Mat *)".

I think that it's a parameter definition problem. what should I do to fix these problems?
If I have to modify "PetscSOE.cpp", would you give me a example?



2.
When using ver. 2.0, a syntax about Mumps system is " system Mumps -ICNTL14 100 "
Is the syntax available in ver. 1.7.5 ??
or should I use " system Mumps " ?

3.
The reason why I used ver. 1.7.5 is a problem about compiling the MPICH2 for ver. 2.0.
As compiling the MPICH2 to my system, are there some options to be modified ??

DK

Posted: Mon Jun 23, 2008 12:51 pm
by fmk
1. you want to try the MUMPS compiler, it is at the moment the fastest by far. for the moment forget the Petsc solver and concentrate on the Mumps one.

2. that syntax was not available in 1.7.5. you really should try to get 2.0.0 working.

3, what is your system? mpich2 is usually a snap to install and there are binary versions available for some systems, eg windows.