[Wien] Compiling Wien under Linux on PowerPC

[Dr. Grigori V. Vajenine]

A brief report on solving my installation problems just in case somebody 
has the same situation:

1. upgrading the system to sles9 and the compilers to xlf 9.1 and xlc 
7.0 as well as essl did help.
2. the -q64 option was needed not only for both compilers but also in 
the linking options:
xlf90_r -O5 -qstrict -qarch=auto -q64
xlc_r -q64
linking options: -q64 -L../SRC_lib
3. I used the -qstrict option in addition to others in order to avoid 
warnings about possible change in semantics due to code optimisation 
(could that be a problem?).
4. to the problem of compiling the SRC_aim code (see below): we found 
out, that the underscore symbol "_" in the names of the functions in 
d1mach.c caused the problem for some reason. After deleting these in all 
function names the code compiled normally. I hope this does not affect 
the functionality!

Perhaps a silly question of a novice:
does such a compilation use both processors by default (I am planning to 
use only one computer) or should I compile it parallel or somehow setup 
a parallel calculation?
Preprocessor flags were not altered: '-WF,-DParallel'

Thanks for help in advance!

Grigori Vajenine
MPI-FKF
Stuttgart
Email: G.Vajenine at fkf.mpg.de


Peter Blaha wrote:
> Why did you use  FOPT  without the -q64 option ? Maybe -qarch=auto sets 
> this automatically, but maybe you are still running in 32bit mode.
> This could eventually explain the nn,kgen and lapw2 segmentation faults.
>
>  > current:FOPT:-O3 -w -qarch=auto
>  > current:FPOPT:-O5 -qarch=pwr5 -q64
>
> Otherwise your reports hint very much that updates of compiler and essl 
> are necessary, but of course there's never a guarantee.
>
>
>   
>> compilation resulted in a few error messages by compiling the source in 
>> SRC_aim, SRC_broadening, SRC_eosfit6, and SRC_pairhess (the compile.msg 
>>     
>
> This does NOT affect the "core" functionality of WIEN2k.
>
>   
>> xlf90_r -o ./aim moduls.o charge.o  gbass.o  gener.o  aim.o main1.o  outin.o  radial.o  reduc.o  rsur.o rotat.o  rotate.o  rotato.o  rotdef1.o sternb.o sum.o  vnorm.o  ylm.o  dtylm.o  sumd.o  grhosphe.o dradial.o  gen_change.o  rotat_back.o  rotate_back.o gen_brav.o  interst.o  sphere.o  init.o  onestep.o rsurf.o  gauleg.o  critic.o  surf.o integrho.o follow.o interp.o cputim.o jacobi.o ludcmp.o spline.o doit.o dtdtylm.o sumdd.o hrhosphe.o matprod.o readcs.o dipole.o rho.o cossin.o  intlib.o davint.o d1mach.o sortag.o follown.o SplineFit.o CheckCSpline.o  -L../SRC_lib -lessl -llapack_lapw
>> intlib.o: In function `gaus8':
>> intlib.o(.text+0x7054): undefined reference to `i1mach'
>> intlib.o(.text+0x7054): relocation truncated to fit: R_PPC_REL24 i1mach
>> intlib.o(.text+0x7068): undefined reference to `d1mach'
>>     
>
> d1mach and i1mach are contained in d1mach.c. For some reason it seems 
> your make command did not compile this c file. What is your C compiler ?
> Maybe it needs something like   xlc90 -q64   or -qarch=auto
>
>
> eosfit6 compiled without problems. The "error" is no real error.
>
> broadening and pairhess have some small non-standard f90 part. It will 
> be fixed with the next release.
>
>                                         P.Blaha
> --------------------------------------------------------------------------
> Peter BLAHA, Inst.f. Materials Chemistry, TU Vienna, A-1060 Vienna
> Phone: +43-1-58801-15671             FAX: +43-1-58801-15698
> Email: blaha at theochem.tuwien.ac.at    WWW: 
> http://info.tuwien.ac.at/theochem/
> --------------------------------------------------------------------------
> _______________________________________________
> Wien mailing list
> Wien at zeus.theochem.tuwien.ac.at
> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
>