FEAP User Forum
FEAP => Parallel FEAP => Topic started by: JReinoso on April 29, 2024, 03:18:11 PM
-
Dear FEAP members,
I trying to configure pETSc library on a mac M3 Sonoma.
however, I get this error.
Fortran error! mpi_init() could not be located!
Any way to get rid of it?
Thanks
-
Are you building parFEAP or you building petsc when you get this error?
-
This is when I try to build petsc.
It is weird because I installed mpi via https://formulae.brew.sh/formula/open-mpi without any error.
Do I need to use a different way to proceed?
Maybe I need only to change the bulding options of petsc?
Thanks
-
My go to method is to allow petsc to install mpi for me. Here is my usual configure line
export PETSC_DIR=<set to the location of your petsc folder>
export PETSC_ARCH=gnu
./configure --download-{parmetis,superlu_dist,openmpi,ml,hypre,metis,mumps,scalapack,blacs}
-
If you want to use brew for mpi, then you will need to point petsc to your MPI installation. See the petsc install pages for numerous options.
-
Many thanks for your help, but there are still some issues.
Which version of petsc are you using? I tried with 3.13.2 and 3.19.2, with no success.
I managed to install with. no problems in another debian-based server, and therefore I wanna test the performance of parfeap86. Do you have some examples to do so apart from those of the manual (especailly invoking petsc library)?
Thanks again
JReinoso
-
Right now I use 3.19.4.
You should always use the most current version.
For installation help, the petsc mailing lists are highly recommended.
You can send them your configuration.log and they usually have excellent
suggestions on how to get your installation to proceed.
-
Many thanks, I try out.
I am dealing with Taylor hood elements with Q2Q1Q1 formulation (disp-pressure-temperature)
Do you please indicate me any info about the implementation os such user-defined elements in FEAP?
Many thanks
JReinoso
-
We do not have this element implemented. But there is a Taylor-Hood fluids element. See fluids/flu2d3.f.