[Wien] SCF crash, XLF IBM

[Laurence Marks]

Peter may want to know the full list (his email crossed as I composed
this); if there is anything on the mixer or mini please send those to
me.

One thing as I have been working with someone else who also has aix.
The essl library has some routines with the same names as those in
lapack but different arguments. For certain parts of the mixer do not
work if they are linked against essl first, so compilation options
need to have (if essl is used)  "-llapack and -lessl" so lapack is
searched first. They probably already know this.

On Fri, Mar 22, 2013 at 8:41 AM, Luis Ogando <lcodacal at gmail.com> wrote:
> Dear Prof. Blaha, Marks and WIEN2k community,
>
>    First of all, I would like to thank your comments on my IBM/XLF problem.
>    About the suggestion given by Prof. Blaha, people from RES believe that
> the problem is in the writing of case.energy (number of lines, ...) and not
> in the reading. Could you please indicate "who" writes case.energy ?
>
>    About the changes made in the code (asked by Prof. Marks), they sent me
> some information:
>
> ===================================================================================================
>
> SRC_lapw0:
>
> lapw0.F
>
> *line 256: line break inside a string, solution:
>
> if ((alphax .lt. 0d0) .or. (alphax .gt. 1d0)) stop 'error in case.inhf:&
> & the fraction of HF exchange should be between 0 and 1'
>
> *several lines: comparison among logical values
>  .neq. and .eq. changed by .neqv. and .eqv.
>
> vx_screened.f:
>
> * line 47: .eqv.
>
> SRC_lapw1:
>
> inilpw.f
>
> *line 229: comment with # changed by !
>
> SRC_hf:
>
> calc_exhf.f
>
> several lines : line break inside a format: solution:
>
>
> write(6,'("valence-valence HF exchange energy in sphere ",i4,&
> & " (mult=",i4,") = ",f22.8," Ry")') &
>
> calc_exhfvv_tmp_.F
>
> line 138.40: 1512-050 (W) Field separator is missing, in literal FMT
> specifier, after edit descriptor ".  A comma is assumed.
>
> WE DO NOT UNDERSTAND THIS ASSUMPTION
>
> * line 269: changed
> igv = (/ 1:ng /) by
>
> igv = (/ (i,i=1,ng) /)
>
> * line 272: changed
> igv = (/ 1:(igs-1),(igs+1):ng /) by
>
> igv = (/ (i,i=1,igs-1),(i,i=igs+1,ng) /)
>
> calc_h.F:
>
> * line 290:
>
> indexb(1:nbf,1) = (/1:nbf/) by
>
> indexb(1:nbf,1) = (/ (i,i=1,nbf) /)
>
> * line 421:
>
> the same
>
> * line 424:
>
> again
>
> * problems with line break (&) in many lines
> (1037,1038,1039, 1053, 1054, 1055)
>
> calc_rhovalvxsl.F
>
> * line breaks
>
> * line 25:
> r(1:nr) = r0(iat)*exp(dx(iat)*(dble((/1:nr/))-1d0)) by
>
> r(1:nr) = r0(iat)*exp(dx(iat)*(dble((/ (i,i=1,nr) /))-1d0))
>
> calc_ucuchucuc.f
>
> line 33:
>
>
> r(1:nr) = r0(iat)*exp(dx(iat)*(dble((/ (j,j=1,nr) /))-1d0)
>
>
> calc_uu.f
>
> line 23:
>
>
>  r(1:nr) = r0(iat)*exp(dx(iat)*(dble((/ (i,i=1,nr) /))-1d0))
>
> calc_uuchucu.f
>
> line 33:
>
>
> r(1:nr) = r0(iat)*exp(dx(iat)*(dble((/ (j,j=1,nr) /))-1d0))
>
>
> calc_uuchucuh.f
>
> line 35:
>
>
> r(1:nr) = r0(iat)*exp(dx(iat)*(dble((/ (j,j=1,nr) /))-1d0))
>
> calc_uuguu.f
>
> line 33
>
>
> r(1:nr) = r0(iat)*exp(dx(iat)*(dble((/ (j,j=1,nr) /))-1d0))
>
> calc_uuguuh.f
>
> line 35: the same
>
> calc_uui.f
>
> line 33: the same
>
> calc_uuih.f.
>
> line 35: the same
>
> calc_uvxslug.f
>
>
>    And it goes on repeating these kinds of problems !!!
>
> ===================================================================================================
>
>    Now, they are compiling WIEN2k with lower optimization flags. Let's see
> if anything changes.
>    All the best,
>                   Luis



-- 
Professor Laurence Marks
Department of Materials Science and Engineering
Northwestern University
www.numis.northwestern.edu 1-847-491-3996
"Research is to see what everybody else has seen, and to think what
nobody else has thought"
Albert Szent-Gyorgi