The ABINIT Users Mailing List ( CLOSED )

[Paul Fons <paul-fons@aist.go.jp>]

  I have been interested in using the paw basis to reduce the basis  
set size for some of the large (for me anyway) supercells I am doing  
geometry relaxation in.  Although in the 4.6.2 version of abinit,   
the tutorial runs correctly, I find that when I try and relax ZnO  
(wurtzite, 4 atoms/cell) that a run time error occurs associated with  
freeing of memory in pawinit (I can send details if helpful). This is  
using g95 on a dual power mac G5.  It would appear that (from doing  
diffs) the code is changing quite a bit in the paw routines between  
minor versions.  Are the paw routines usable at this point or are  
they "under construction"?  Running the same ZnO input file on the  
latest linux pc binary (4.6.3), the run gives garbage (the SCF energy  
values jump all over the place).  One of the reasons I ask is that I  
cannot be sure that it is a compile issue or not (with g95).  Using  
the 4.4.3 (xlf?) binary for the mac from the web site gives  
reasonable energies and looks like it is working alright leading to  
the confusion.

  It would also appear from the release notes that paw was  
significantly improved in going form 4.4 to 4.5 to 4.6 (paw almost  
beta quality). In short, I would like to be able to compile abinit  
myself so I can run under mpich and use at least two cpus (if not  
more later) for my calculations and I would like to know if the code  
is stable.  Has anyone else compiled a serial/parallel version of the  
4.6 (I have yet to try 4.5) and is there a binary available somewhere  
(why do the mac binaries stop at 4.4?)

 In any case, thanks all for listening to my rambling.  I really  
would like to use PAW as it would appear to have many of the  
advantages of FLAP without the overhead (my lack of cpu power is  
showing!)

                                                                     
Paul