The ABINIT Users Mailing List ( CLOSED )

["Ludwig, Christian" <ludwigc@uni-mainz.de>]

Hi Matteo,

I did what you suggested and the compilation went a little further. But then 
I got the same kind of error for the file m_sigma_results. I did the same for 
this file, enclosing the executable part and commenting the calls. Is this ok 
or is m_sigma_results needed for something important? The compilation seemed 
to be fine, up to a point where I got

ar: hdr_vs_dtset.o: No such file or directory         (not surprising, since 
it was not compiled)
make[3]: *** [lib68_gw.a] Error 1                            (I did not find 
this lib68_gw anywhere) 

Now what? The complete output of the compilation is

1501-510  Compilation successful for file split_sigc_cpp.f90.
ABINIT WRAPPER END
rm -f lib68_gw.a
ar rc lib68_gw.a assemblychi0_sym.o assemblychi0sfq0.o assemblychi0q0_sym.o 
bands_classification.o bstruct_init.o calc_coh.o calc_density.o calc_ffm.o 
calc_rpa_functional.o calc_sig_noppm.o calc_sig_ppm.o calc_sig_ppm_comp.o 
calc_vHxc_braket.o calc_wf_qp.o cchi0.o cchi0q0.o cppm1par.o cppm2par.o 
cppm3par.o cppm4par.o cqratio.o csigme.o cutoff_m_elem.o debug_tools.o 
dosym.o dotproductqrc.o eps1_tc.o fourdp_6d.o get_bands_sym_GW.o 
gwcompleteness.o gw_tools.o hdr_init.o hdr_vs_dtset.o kramerskronig.o 
linear_optics_paw.o make_epsm1_driver.o mlwfovlp_qp.o m_bz_mesh.o 
m_commutator_vkbr.o m_coulombian.o m_crystal.o m_dyson_solver.o m_electrons.o 
m_finite_cylinder.o m_fft_mesh.o m_gsphere.o m_gwannier.o m_gwdefs.o 
m_io_screening.o m_little_group.o m_ppmodel.o m_qparticles.o m_screening.o 
m_sigma_results.o m_spectra.o m_wfs.o outeps.o pawr.o paw_mknewh0.o 
paw_mkrhox.o paw_mkrhox_spl.o paw_rho_tw_g.o paw_symcprj.o print_psps.o 
q0fit.o rho_tw_g.o setup_qmesh.o setup_screening.o setup_sigma.o 
set_gwdistrb.o symmetrize_afm_chi0.o symf12.o spectral.o split_sigc.o 
interfaces_68_gw.o
ar: hdr_vs_dtset.o: No such file or directory
make[3]: *** [lib68_gw.a] Error 1
make[3]: Leaving directory `/usr/local/abinit/abinit-5.8.3/build/src/68_gw'
make[2]: *** [all-recursive] Error 1
make[2]: Leaving directory `/usr/local/abinit/abinit-5.8.3/build/src'
make[1]: *** [all-recursive] Error 1
make[1]: Leaving directory `/usr/local/abinit/abinit-5.8.3/build'
make: *** [all] Error 2

________________________________________
Von: Matteo Giantomassi [Matteo.Giantomassi@uclouvain.be]
Gesendet: Dienstag, 30. Juni 2009 18:00
An: forum@abinit.org
Betreff: Re: [abinit-forum] still problems with compilation on ibm power

> After searching through the archive of the mailing list I found a similar
> error. The suggestion was to reduce optimisation, but after reducing it to
> "-O0" I still get the very same error.
>
> Don't be shy to share your thoughts with me! :)
>
> I was thinking: The directory, where the error occurs, is 68_gw. I guess
> this has to do with the DFT method GW. If I do not use this, can I simply
> omit 68_gw during compilation?

 Dear Christian,

 It's not so straightforward as some routines in 95_drive
 (screening and sigma) and  93_rdm as well 98_main/mrgscr.F90
 depend on 68_gw. One should hack a bit
 the code such that all these procedures are not used. A much easier
 workaroud would be:

 1) Enclose the executable part of hdr_vs_dtset.F90 between the CPP if
statement:

 #if 0
  !executable section
 #endif

  In this way, the body of the routine is not compiled, and XLF should not
go off.

 2) Comment all calls to hdr_vs_dtset in the GW code.
    They are just used for consistency check.
    and are not necessary if you are not planning to run GW calculations.
    Only four files have to changed.

    grep hdr_vs_dtset */*.F90 | grep call gives:

    68_gw/setup_screening.F90: call hdr_vs_dtset(Hdr_kss,Dtset)
    68_gw/setup_sigma.F90: call hdr_vs_dtset(Hdr_kss,Dtset)
    68_gw/setup_sigma.F90: ! call hdr_vs_dtset(Er%Hscr%Hdr,Dtset)
    93_rdm/rdm.F90: call hdr_vs_dtset(Hdr_kss,Dtset)

 3) rerun make to see if XLF still complains about 68_gw.

 PS: Which version of XLF are you using? Could you provide the
 preprocessed version of hdr_vs_dtset (hdr_vs_dtset_cpp.F90)

 PPS: Simple hack to avoid having to recompile the entire
 source tree just to change the optimization of one routine.
 Kill the compilation when you reach hdr_vs_dtset, cut&paste
 the command line produced by make, change -O2 with -O0.
 Then use the new command line to produce the object file with reduced
optimization.

 A more elegant solution consists in using per-directory optimization,
 I recently saw a post from Yann in which he was explaining how
 to specify the optimization flags per each directory.

 Best Regards
 Matteo Giantomassi