logan-nc
commented
7 years ago Cool. At this point, with multiple people following/using FOCUS, you might want to consider using releases instead of these issues to record significant steps in the development.
I've sent it before, but I still recomend the suggestion on http://nvie.com/posts/a-successful-git-branching-model/ of releasing ever time you push to master (excepting accidents like when I made the norm pull request to master instead of develop). As per the semantic versioning standards (http://semver.org/) you would start with release 0.1.0. This way, assuming you include the version in the h5 output, you can go back and checkout the version you used to run something when coming back to it a year from now.
zhucaoxiang
Dear all,
I pushed several updates in both 'master' and 'develop' branches. Here are the key points:
Add options to freely control the optimization focus : currents or geometries The two input namelist weight_Inorm and weight_Gnorm are added. Thanks to @logan-nc , the details can be seen in pull_12.
Add additional criteria to exit optimizations The new input variable exit_tol could be used to exit from the running optimizations when the relative change of last 5 outputs is small than exit_tol. By default, it's 1.0E-04. You can use this to accelerate your jobs. Refer to initial.pdf.
Replace the repeated output information in CG Now the second column of output information when using CG is the time from the start of optimization. Just for a reference. Any other informations can be added.
And, the bug that wrong cpu time is shown when using gfortran compiler is now fixed by replacing
include "mpif.h"withuse mpi.fixed other minor stuffs
If you have any questions, please let me know. Thanks!