Allowing SGE schedulers to use the plugin #41
Conversation
There was a problem hiding this comment.
Thanks @JPchico. I left two minor comments. The rest looks fine to me.
| @@ -331,7 +337,9 @@ def get_inputs_common(calculation, code, remote, structure, options, label, desc | |||
|
|
|||
| if code: | |||
| inputs.code = code | |||
|
|
|||
| _sched = load_computer(code.computer.label).scheduler_type | |||
There was a problem hiding this comment.
Why is load_computer needed? Does code.computer not give you the computer already?
There was a problem hiding this comment.
Yeah, you are right, I copied this from some other function where I wanted to load the computer for some tests. I have changed it as you recommed.
examples/kkr_submit_scf.py
Outdated
| # Total number of mpi processes | ||
| 'tot_num_mpiprocs': 16, | ||
| # Name of the parallel environment | ||
| 'parallel_env': 'mpi', |
There was a problem hiding this comment.
Maybe you can put a comment here to tell that this is for a SGE scheduler?
There was a problem hiding this comment.
Sure thing, added now a check for whether the code is in an SGE or SLURM/PBS machine. And added some reasonable options
…E, slurm and pbspro machines.
|
I have added the changes that you requested @PhilippRue , plus some smaller formatting issues. |
|
@JPchico I will merge this in tomorrow if there is nothing to add |
|
@PhilippRue I think this is fine, I have not seen any other issue cropping up in my tests. |
The plugin could not work as the
get_inputs_commonadded submission options which are incompatible for the SGE scheduler. Now it checks which scheduler is provided and sets options which correspond to a serial simulation. Fixes #40