-
Notifications
You must be signed in to change notification settings - Fork 5
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 bare (no container) install for NCI #14
base: main
Are you sure you want to change the base?
Conversation
Fixed some incompatibilities with using packages built for the container in a bare install. @hiker could you please give the bare install a go and let me know how it goes? |
@ScottWales Now I get:
I did a clean rebuild (after previously trying to just update), same effect. Directory: /scratch/dp9/jxh903/ngmo-envs |
That's my mess up with git I think - I put some changes in the wrong branch. I've force-pushed a fix, you may need to delete your local branch to pick it up:
|
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I finally got it working (and I also realised that messages from this repo got filtered into the wrong folder, so I didn't notice your answer in time ... apologies).
I do have one problem though: the intel compilers are not in the path. E.g.:
[jxh903@gadi-login-07 ~]$ mpif90 -show
/apps/intel-ct/2021.10.0/compiler/linux/bin/intel64/ifort -I/apps/openmpi/4.1.5/include -I/apps/openmpi/4.1.5/include/Intel -L/apps/openmpi/4.1.5/lib -L/apps/openmpi/4.1.5/lib/Intel -L/system/lib64 -Wl,-rpath=/apps/openmpi/4.1.5/lib -L/apps/intel-oneapi/compiler/2023.2.0/linux/compiler/lib/intel64_lin -Wl,-rpath=/apps/intel-oneapi/compiler/2023.2.0/linux/compiler/lib/intel64_lin -lmpi_usempif08 -lmpi_usempi_ignore_tkr -lmpi_mpifh -lmpi
[jxh903@gadi-login-07 ~]$ ifort
bash: ifort: command not found
Fab relies on being able to run the wrapped compiler to make certain versions are consistent (e.g. it verifies that mpif90-ifort
does indeed return the same version as ifort
, to early detect if e.g. the wrong mpif90 wrapper is being started). Adding the intel compiler to the path fixes this, but I think the plain compilers should be available in the environment.
Add an installation method not using a container at NCI. This can help during package development as all of the Spack builds are on the local filesystem rather than the compute nodes' temporary disks so inspection is easier. It has the downside of installs having a much higher inode count.
To the user the container and bare environments appear exactly the same, in both cases the envrun script will ensure the environment is loaded for a command.