A talk given by Ansgar Schaefer studying convergence behaviour for rutiles is here (pdf).
General suggestions for helping GPAW convergence are here.
A discussion and suggestions for converging some simple systems can be found here.
Other convergence experience:
System |
Who |
Action |
---|---|---|
Graphene with vacancy |
Felix/JensH |
Increase Fermi Temp from 0.1 to 0.2, use cg |
Graphene with vacancy |
cpo |
change nbands from -10 to -20, MixerDif(beta=0.03, nmaxold=5, weight=50.0) |
Nitrogenase FeVCo for CO2 reduction |
grabow |
use Davidson solver (faster as well?), although later jvarley said MixerSum |
Several surfaces |
aap |
Broyden mixer with Beta=0.5 |
The get a guess for the right number of nodes to run on for GPAW, run the
following line interactively:
gpaw-python <yourjob>.py --dry-run=<numberofnodes> (e.g. gpaw-python graphene.py --dry-run=16) |
Number of nodes should be a multiple of 8. This will run quickly
(because it doesn't do the calculation). Then check that the
following number is <3GB for the 8-core farm, <4GB for the 12-core farm:
Memory estimate --------------- Calculator 574.32 MiB |
privgpaw.csh build privgpaw.csh test privgpaw.csh bsub <arguments> |
Some notes:
You can find a Jacapo parallel NEB example in /afs/slac/g/suncat/example/parneb.py. Some lines need to change for a restart. An example is in the same directory in parneb_restart.py.
Note that this restart example turns on the "FIRE" and and "climb" parameters: this is for later in the calculation. The documentation here discusses the reasons for that (although perhaps doesn't spell out clearly what the criteria are for turning those on).