petsc.h error on new cluster

[Alex Kaiser]

Hi everyone, 

Trying to install on a new cluster, and have an error related to petsc.h. It says "configure: error: could not find header file petsc.h" but I'm not sure this is fully descriptive, given the log file. Relevant part of config.log is: 

configure:30268: using PETSC_DIR: /home/adk354/sfw/petsc/petsc

configure:30270: using PETSC_ARCH: linux-opt

configure:30279: checking petsc.h usability

configure:30279: mpicxx -c   -O3 -Wall -I/home/adk354/sfw/petsc/petsc/include -I/home/adk354/sfw/petsc/petsc/linux-opt/include -I/home/adk354/sfw/ibamr/IBAMR/ibtk/contrib/muparser/include -I/home/adk354/sfw/linux/hdf5/1.8.14/include -I/home/adk354/sfw/ibamr/IBAMR/ibtk/contrib/eigen -I/home/adk354/sfw/ibamr/IBAMR/ibtk/contrib/boost_1_60_0  -I/home/adk354/sfw/ibamr/IBAMR/ibtk/contrib/boost_1_60_0  -DOMPI_SKIP_MPICXX -fpermissive conftest.cpp >&5

In file included from /home/adk354/sfw/petsc/petsc/include/petscbag.h:4:0,

                 from /home/adk354/sfw/petsc/petsc/include/petsc.h:5,

                 from conftest.cpp:69:

/home/adk354/sfw/petsc/petsc/include/petscsys.h:150:6: error: #error "PETSc was configured with OpenMPI but now appears to be compiling using a non-OpenMPI mpi.h"

 #    error "PETSc was configured with OpenMPI but now appears to be compiling using a non-OpenMPI mpi.h"

      ^

Any idea what is going on here? Full log is attached. 

Thank you, 

Alex 

[Boyce Griffith]

It looks like different MPIs are being used to build different libraries. Did you build everything from scratch, or are some of these libraries provided by the system?

Any idea what is going on here? Full log is attached. 

Thank you, 

Alex 

--
You received this message because you are subscribed to the Google Groups "IBAMR Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ibamr-users...@googlegroups.com.
To post to this group, send email to ibamr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.
<config.log>

[Alex Kaiser]

All from scratch, except the OpenMPI and gcc compilers, which are system. I did a quick test with which and -v and the compilers look consistent. 

A previous script added a system LAPACK because petsc was having trouble finding it on the old cluster, but I just removed that and the problem is the same. It also had "--with-shared-libraries" in the petsc config line. Removed that too with no effect on IBAMR configure. Petsc compiles and passes make check. 

Thanks, 

To unsubscribe from this group and stop receiving emails from it, send an email to ibamr-users+unsubscribe@googlegroups.com.

To post to this group, send email to ibamr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.
<config.log>

--
You received this message because you are subscribed to a topic in the Google Groups "IBAMR Users" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/ibamr-users/3QOb39gqaJ0/unsubscribe.
To unsubscribe from this group and all its topics, send an email to ibamr-users+unsubscribe@googlegroups.com.

[Boyce Griffith]

Can you check that PETSc is using the same MPI compliers? E.g.

$ tail -n80 $PETSC_DIR/$PETSC_ARCH/lib/petsc/conf/configure.log

It might be help to give full paths to the MPI C and C++ compliers.

To unsubscribe from this group and stop receiving emails from it, send an email to ibamr-users...@googlegroups.com.

[Alex Kaiser]

Thank you for suggesting that command; it revealed that something was not being cleaned properly and Petsc was still trying to link to the system LAPACK that was the source of the problem.

I set Petsc to download and build BLAS and LAPACK, and that worked fine. Will consult with the cluster people about the built in one, because system BLAS is probably better if they have really tried to tune it. 

Thank you, 

Alex 

To unsubscribe from this group and stop receiving emails from it, send an email to ibamr-users+unsubscribe@googlegroups.com.

To post to this group, send email to ibamr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Boyce Griffith]

On Apr 23, 2017, at 4:37 PM, Alex Kaiser <adka...@gmail.com> wrote:

Thank you for suggesting that command; it revealed that something was not being cleaned properly and Petsc was still trying to link to the system LAPACK that was the source of the problem.

I set Petsc to download and build BLAS and LAPACK, and that worked fine. Will consult with the cluster people about the built in one, because system BLAS is probably better if they have really tried to tune it. 

I think this is unlikely to make a big difference (and would be interested to know whether I am wrong!).

-- Boyce

To unsubscribe from this group and stop receiving emails from it, send an email to ibamr-users...@googlegroups.com.