Ignoring ALGO Fast warnings

[jeremiah]

Hello,

Is it safe to ignore the messages "num prob" and "WARNING in EDDRMM" for ALGO Fast as suggested in this forum post?

forum/viewtopic.php?f=3&t=17822

The answer in that post mentioned commenting out the WRITE statements in rmm-diis.F, but I am still getting the warning in VASP 6.3.2. I am also wondering about the "num prob" messages that are sometimes printed out many times along with the EDDRMM warning.

Thank you,
Jeremiah

[marie-therese.huebsch]

Hi,

Yes, it is safe. The write statements still haven't been removed. That should be done in the future. Sorry, you invested time in searching for the meaning of this warning.

Marie-Therese

[jeremiah]

Good to know, thank you.

[william_watson]

Can I ask why it is safe to ignore the 'WARNING in EDDRMM' and 'numprob' errors? Is there a limit at which you cannot ignore these warnings e.g. if you see more than approximately n instances in your output (.o*) file?

[marie-therese.huebsch]

As explained by Jürgen Furthmüller:

It occurs if you are very close to machine
accuracy in the RMM-DIIS scheme and single band iteration only results in "random" progress
anymore (due to machine accuracy limits). It says then indirectly to you "could not optimize
this band further" -- and the internal action taken is then simply NOT to update the wave
function further and to proceed with the next band(s).

So you can always ignore that warning.