I have a version now that runs the following tests correctly:
bound states of C
e-C scattering in LS-coupling
e-C scattering in jK-coupling
photodetachment of C^- (LS-coupling)
There may be other cases as well (e.g., bound states of Ca), but I haven't checked.
I suggest we delete ALL input data from the Gateway, unless the codes (not just BSR) have been tested with these inputs. For example, if we don't have a BSR version that can do photoionization/photodetachment, then we need to remove any claims that this can be done, and we certainly shouldn't have input data for those cases.
Finally, individual jobs are so s..l.....o.......w on Expanse now (where BSR is supposedly running) that it makes practically no sense to try individual steps. We asked for one (1) minute on a shared node (one core) last week do run BSR_PREP (it really only needs a fraction of a second), and the job was waiting and waiting and waiting. We should set up workflows for all these test cases, so that one only has to wait once. I made proper run scripts based on what Oleg had, but some of the very early steps are not needed if we provide the appropriate input files (e.g., the .bsw files rather than formatted or .w files that need to be converted using some of the utilities that may or may not be available). While I was in the process, I also fixed the English bugs (there were many) in those scripts. While they don't change the ultimate result, a public version should look nice. I don't know how to make workflows on the Gateway, but I can provide the run scripts.
Finally, it is very confusing / misleading when the Gateway message for a job says that it's "Executing", when it is really just "Waiting".
I have a version now that runs the following tests correctly:
bound states of C e-C scattering in LS-coupling e-C scattering in jK-coupling photodetachment of C^- (LS-coupling)
There may be other cases as well (e.g., bound states of Ca), but I haven't checked.
I suggest we delete ALL input data from the Gateway, unless the codes (not just BSR) have been tested with these inputs. For example, if we don't have a BSR version that can do photoionization/photodetachment, then we need to remove any claims that this can be done, and we certainly shouldn't have input data for those cases.
Finally, individual jobs are so s..l.....o.......w on Expanse now (where BSR is supposedly running) that it makes practically no sense to try individual steps. We asked for one (1) minute on a shared node (one core) last week do run BSR_PREP (it really only needs a fraction of a second), and the job was waiting and waiting and waiting. We should set up workflows for all these test cases, so that one only has to wait once. I made proper run scripts based on what Oleg had, but some of the very early steps are not needed if we provide the appropriate input files (e.g., the .bsw files rather than formatted or .w files that need to be converted using some of the utilities that may or may not be available). While I was in the process, I also fixed the English bugs (there were many) in those scripts. While they don't change the ultimate result, a public version should look nice. I don't know how to make workflows on the Gateway, but I can provide the run scripts.
Finally, it is very confusing / misleading when the Gateway message for a job says that it's "Executing", when it is really just "Waiting".