The ABINIT Users Mailing List ( CLOSED )

["Eric J. Walter" <ewalter@particle.physics.wm.edu>]

Dear Abinit users,

I am seeing some strange behavior with abinip which might be just due
to an improper input file, but I can't be sure.

The problem deals with a string of parallel GS and then RF
calculations. Specifically when localrdwf is set to 0 and kptopt is
changed from one calculation to another.  I run on a cluster
which uses a separate file system for each processor, so 
using localrdwf correctly is important for me.

I am finding that if I do the first/GS calculation with kptopt=1 as
usual and then proceed to, say, a d/dk RF calculation with kptopt=2,
if localrdwf is 0, incorrect results are generated (ETOT is differs by
about 0.5 Ha from the serial calculation).

If I use localrdwf=1 I get the same results as a serial version of
this calculation.

Finally, if I make kptopt=2 for the GS calculation, localrdwf=0 and 1
works fine and give the same output as the serial version.

So, this seems to say that localrdwf=0 should never be used when kptopt
is changed from one calculation to another.  Is this true?  Have 
others seen this?  

I am running on a Linux Alpha machine and I have tried this test with
both v3.3.4 and 3.4.3 of abinit.  I would be happy to provide more
information if it would be helpful.

Many thanks for any help anyone can give,

Eric Walter
Department of Physics
College of William and Mary