New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Add initial Apptainer/Singularity definition file #435
base: devel
Are you sure you want to change the base?
Conversation
Pinging @ettaka |
Here are some failing tests as I am running now the suite, which does take a while in my computer...
And I just want to point out that the failing tests are not failing due to my computer not having enough CPUs ;) |
HYPRE seems to be broken in this build.
|
Here is the summary of the results
Quite a few tests fail like this:
And like this, just right after running
|
Great thanks! I will have a look next neek. |
Latest devel branch in Elmer allows for externally provided UMFPACK - I tested with SuiteSparse. These were the configuration lines I added |
Thanks @tzwinger! I will update the Apptainer file to reflect this change and run the tests once again :) You may want to mention your solution on #434 I think having a toggle between prepackaged UMFPACK and external (GPL) UMFPACK is the best of both worlds. Also, one question. Why don't we make CMake add the linker flags Thanks, |
I just built the Apptainer image with your flags (adapted to use Debian's packaged libs) and it compiles. There are a few warnings regarding implicit declarations, but that is about it. Here is one of them:
I am running now the test suite once again to see if there are any new failures. EDIT: note that I am not sure if those UMFPACK warnings were already present or if they are new. I am just pointing them out now. |
I can report that no new tests are failing by using the system provided UMFPACK. However, the tests took a bit longer to run eventhough I changed the
I will push the changes now. |
Nice work! I tried to build
|
Hi @ettaka how are you entering the Singularity environment? The files/tests should be located inside the container under I am using the latest Apptainer version, but Singularity should also work... |
I used |
Mmmmm... Maybe singularity is using the If not, I would encourage you to rebuild the "image" and look in the build log in the terminal for the directory that is being used during the build. |
Hi @ettaka, did you have any luck with the Apptainer file? If you want anything modified or more help feel free to ask! Regards, |
Hi ElmerFEM team.
I created an Apptainer/Singularity definition file for Elmer for me to test it and have a "stable/reliable" way of building it. Here is the version I currently have. I build it with everything and I try as much as possible to link with the system libraries (this may not be desirable in some scenarios). This means it is MPI ready and it should have access to all the solvers/meshers. The image is currently quite raw. For example I do not clean the build directory (because I want to be able to test ElmerFEM) and I install a few things that may not be necessary. For example I install the system UMFPACK, but ElmerFEM builds its own version.
You can try and build it with
apptainer build elmer.sif elmer.def
. However, I would actually recommend to build it in sandbox mode in order to test it and debug it. This can be done by building withapptainer build --sandbox elmer_testing elmer.def
. This will create a folder where the entire image resides. We can run ElmerFEM by runningapptainer run -B /run elmer_testing
and ElmerGUI should show up. I had to add the option-B /run
in order for the GUI to detect/attach to the running display.In order to run the tests, use
apptainer shell --writable elmer_testing
. This will drop you into a shell within the container/folder and will make it writable (which is need in order to run the tests). Then, inside the shell session just do acd /root/build && make test
.Currently, it shows me that there are 1040 available tests. If I remember correctly, in the past it told me it could run more, but maybe I am missremembering or things may have changed. THERE ARE QUITE A FEW FAILING TESTS... I am not good enough with Elmer to debug them... I do remember that after running all of them and then rerunning the failing ones, some did pass. I do not know if this is an issues with running
make test -j14
or if there are memory bugs that only show up sporadically or something else... Maybe it has to do with me using some 64bit int math libraries...Anyhow, here is the initial draft. I would like to receive feedback and criticism. I am specially interested in why there are quite a few failing tests... I also hope that this helps the community and testing of ElmerFEM. And of course, you can propose changes, such as changing Debian for another base system or whatever :)
Best regards,
Fer