Hi Pooja,
Concerning libomp: As the full brew package list shows, I had that installed. I wonder a bit why I have it and you did not get it automatically even though you have the same brew leaves
I have. It may be that I had the libomp leftover as dependency of another package that I uninstalled in the meantime or that brew leaves is not entirely reliable. But that problem is solved.
The complaint about libiomp5 in my experience relates to NumPy built with the Intel MKL, which is default for conda. Usually, one can avoid this by creating a conda environment where
nomkl is the first package specified/installed. You can check this either by running conda list and check if nomkl or a number of mkl* packages are installed, or by importing numpy into Python and then running numpy.__config__.show(). If that mentions mkl_rt
libraries, you have NumPy with MKL in your environment, if it shows only openblas you don’t.
Best,
Hans Ekkehard
--
Prof. Dr. Hans Ekkehard Plesser
Head, Department of Data Science
Faculty of Science and Technology
Norwegian University of Life Sciences
PO Box 5003, 1432 Aas, Norway
Phone +47 6723 1560
On 18/01/2021, 20:30, "Babu, Pooja" <p.babu@fz-juelich.de> wrote:
Hello Hans,
Thank you for a detailed description of your setup and steps.
I checked them with mine and although I do not have all the brew packages that you have in your complete list, I have the ones you mentioned in brew leaves.
I tried to install nest with Apple’s Clang with the option -Dwith-mpi=OFF. I got the below errors during cmake.
-- Could NOT find OpenMP_C (missing: OpenMP_C_FLAGS OpenMP_C_LIB_NAMES)
-- Could NOT find OpenMP_CXX (missing: OpenMP_CXX_FLAGS OpenMP_CXX_LIB_NAMES)
-- Could NOT find OpenMP (missing: OpenMP_C_FOUND OpenMP_CXX_FOUND)
CMake Error at cmake/ProcessOptions.cmake:466 (message):
CMake can not find OpenMP.
Call Stack (most recent call first):
CMakeLists.txt:146 (nest_process_with_openmp)
I saw another thread with the same issue and found that I had
open-mpi installed via homebrew as mentioned there. The thread also mentions
libomp package. I installed that and the errors were gone. The
make install command was successful. After this, when I imported nest into python, I got the below errors
OMP: Error #15: Initializing libomp.dylib, but found libiomp5.dylib already initialized.
OMP: Hint This means that multiple copies of the OpenMP runtime have been linked into the program. That is dangerous, since it can degrade performance or cause incorrect results. The best thing to do is to ensure that only a single OpenMP runtime is linked
into the process, e.g. by avoiding static linking of the OpenMP runtime in any library. As an unsafe, unsupported, undocumented workaround you can set the environment variable KMP_DUPLICATE_LIB_OK=TRUE to allow the program to continue to execute, but that
may cause crashes or silently produce incorrect results. For more information, please see
http://openmp.llvm.org/
Abort trap: 6
I did the workaround as suggested in the message by setting KMP_DUPLICATE_LIB_OK=TRUE. This setting indeed got rid of the above error
and the import was successful, but as mentioned, the behaviour could be unpredictable.
I couldn’t find any other way to work around the initial OpenMP_C and OpenMP_CXX errors other than installing
libomp package. Do you know if there is any other way to resolve that error?
Thanks,
Pooja