The ABINIT Users Mailing List ( CLOSED )

[Xavier Gonze <gonze@pcpm.ucl.ac.be>]

Dear Pio Baettig,

I see that your compiler has a problem with the routine CCHI0.F90 ,  
part of the
directory tests/15gw . There have been many modifications of cchi0  
since ABINITv4,
and with some compilers, the level of optimization cannot be as high  
as it was in ABINITv4.
You might try a lower optimization level for 15gw than for other  
directories,
something that you can do using    fcflags_opt_15gw  but this is not  
available with the command
line , instead see examples of config files in doc/config/ 
build_examples , e.g.
alphaev67-compaq_deccint.ac

Best,
Xavier


On 30 Aug 2007, at 10:34, Pio Baettig wrote:

> Dear Abinit-users,
>
> I am trying to compile Abinit on SR11000.
> I read the older messages on the forum (Jan2007) and tried to use  
> the options suggested there, but without success so far.
>
> I tried to use the flags suggested by Prof. Arai (see attached  
> forum-post below, and also tried to enhance them with the compiler- 
> flags  we are using for another code. So far, 'make abinip' always  
> failed.
>
> The compilers that are available are mpixf90 and mpixf90_r (and  
> xf90 for serial compilation). Apparently the compilers vary also  
> from SR11000 to SR11000...
>
> The flags I normally used for parallel compilation (HiLAPW) are:
>
> ---
> FC= mpixf90
> FFLAGS= -Os -64 -precexp=4 -nompi -noparallel
> LFLAGS= -64 -noparallel
> LIB= -L/usr/local/lib -lblas_s -llapack_s -noparallel
> ---
>
> So I tried with:
> ./configure --disable-smart-config --enable-64bit-flags \
> --with-mpi-fc="mpixf90 -nohugeary -Os -64 -precexp=4 -nompi \
> -noparallel" --with-mpi-fcflags="-Os -nohugeary -64 -precexp=4 \
> -nompi -noparallel" FC="xf90 -nohugeary -O -64 -precexp=4"  
> FCFLAGS="-O \
> -nohugeary -64 -precexp=4"
>
>
> or some shorter variant (also the original suggestion of Prof.  
> Arai, with the compilers replaced by mpixf90/mpixf90_r and xf90) ,  
> and then
>
> make abinip
>
> but the compilation always broke. Unfortunately, the error-messages  
> are not all too helpful.
>
> ------
> *program name = CCHI0
> *program units = 0001, 0013  diagnostics generated, highest  
> severity code is  12
> make: The error code from the last command is 12.
>
>
> Stop.
> make: The error code from the last command is 2.
>
>
> Stop.
> ------
>
> I would be very grateful if someone has an idea on how to get  
> abinip running.
>
>
>
>
>
>
> So far, I tried mostly with abinit 5.3.4.
>
> Abinit 5.4.2 seems to be easier,
> >./configure --with-mpi-fc="mpixf90" FC="xf90"
> >make abinip
> runs (nearly) through, but there seems to be a problem just at the  
> end of the compilation:
> ----
> *program name = WRITE_MOLDYNVALUENETCDF
> *end of compilation : WRITE_MOLDYNVALUENETCDF
> *program units = 0001, 0011  diagnostics generated, highest  
> severity code is  00
>         rm -f lib21drive.a
>         /usr/bin/ar cru lib21drive.a afterscfloop.o brdmin.o  
> delocint.o diisrelax.o driver.o  drivergw.o elpolariz.o gstate.o  
> handle_err_netcdf.o isotemp.o  loop3dte.o moldyn.o move.o   
> nonlinear.o outscfcv.o scfcv.o  screening.o sigma.o testfi.o   
> timana.o vtorhotf.o wannier.o  write_header_moldynnetcdf.o   
> write_moldynvaluenetcdf.o
>         ranlib lib21drive.a
>         cd src/main && make abinip
>         xf90     -o  abinip
> ----
> and it finishes, but there is no abinip in src/main.
>
>
>
>
>
> Thank you very much for any hints.
>
> Best regards,
>
> Pio Baettig
>
> P.S. if further information or a logfile is needed, I can provide  
> them. I did not want to stuff this mail too much.
>
>
>
>
>
>
> ------------------- Original Mail
>
> Thank you for your suggestions.
>
> Now, I succeded in compiling abinit-5.2.4 without any ad hoc hacks.
>
> I used the following configure options.
>
> ../configure --disable-smart-config --enable-64bit-flags \
> --enable-mpi --with-mpi-fc="mpif90_r -nohugeary -O" \
> --with-mpi-fcflags="-O -nohugeary" \
> FC="f90 -nohugeary -O" FCFLAGS="-O -nohugeary"
>
> The option " --disable-smart-config" solves the problem I met at the
> configuration.
>
> In addition, two modification is necessary to compile on sr11000.
>
> (1) The compile option "-nohugeary" must be used. Without this option,
> some intrinsic functions such as "size" return long integer. (This  
> is an
> extension of Hitachi Fortran. I think such extention violates the  
> Fortran90
> standard.) It causes the compile error at several points.
>
> (2) The source code must be corrected as I pointed out in my e-mail  
> sent to
> abinit mailing list on 2006-12-27.
>
>
> --
>
> Masao Arai
>
> ----------------------