The ABINIT Users Mailing List ( CLOSED )

[kesler@ciw.edu]

Hi,
    In the case of the ABINIT binary files, I wrote a simple hack that
appears to be at least somewhat portable.  It figures out the Fortran
record header length by searching for the first '.' in the version
number, and counting bytes:

  int recHeaderLength = 0;
  char ch='\0';
  while (ch != '.') {
    fread(&ch, sizeof(char), 1, fin);
    recHeaderLength++;
  }
  fseek (fin, (long)0, SEEK_SET);
  recHeaderLength -= 2;
  char dummy[100];
  fread (dummy, sizeof(char), recHeaderLength, fin);

  char vsn[7];
  fread (vsn, sizeof(char), 6, fin);
  vsn[6] = '\0';
  codvsn = vsn;

  fread (&headform, sizeof(int), 1, fin);
  fread (&fform, sizeof(int), 1, fin);

  // Skip FORTRAN record end and start:
  fread (dummy, sizeof(char), 2*recHeaderLength, fin);

  fread (&bantot,    sizeof(int),    1, fin);
  fread (&date,      sizeof(int),    1, fin);

...

If anyone wants the rest of the functions to read the header, etc. let me
know.

Sincerely,
Ken Esler

> Hello everyone,
>
> Joe is correct: fortran writes are ugly things that are very compiler
> dependent. The data fields will be ok, but there can be arbitrary (ie
> compiler-dependent) additional bytes at the beginning and end of each
> _record_ ie for each write()... statement.
>
>
>> I don't agree with this. I have analysed binary files before, either
>> from Fortran or from C. Moreover, the documentation explains
>> the header structure in terms of the Fortran write statement !
>> For example:
>>
>> write(unit=header) codvsn,headform,fform
>> write(unit=header) bantot,date,intxc,ixc,natom,ngfft(1:3),&
>> & nkpt,nspden,nspinor,nsppol,nsym,npsp,ntypat,occopt,pertcase,usepaw,&
>> &
>> ecut,ecutdg,ecutsm,ecut_eff,qptn(1:3),rprimd(1:3,1:3),stmbias,tphysel,tsmear
>
> In the source code:
>
> ...
> read (unitfi)   codvsn,headform,fform
> ...
> read(unitfi) bantot, hdr%date, hdr%intxc, hdr%ixc, natom, hdr%ngfft(1:3),&
> &    nkpt, hdr%nspden, hdr%nspinor, nsppol, nsym, npsp, ntypat,
> hdr%occopt, hdr%pertcase,&
> &    hdr%usepaw, hdr%ecut, hdr%ecutdg, hdr%ecutsm, hdr%ecut_eff,
> hdr%qptn(1:3), hdr%rprimd,&
> &    hdr%stmbias, hdr%tphysel, hdr%tsmear, hdr%usewvl
>
>
> see src/59_io_mpi/hdr_io.F90:
> http://www.abinit.org/downloads/browse-lastest-sources?currenturl=aHR0cDovL3d3d29sZC5hYmluaXQub3JnL3BhY2thZ2Uvcm9ib2RvY181LjguMi81OV9pb19tcGkvXzU5X2lvX21waV8uaHRtbA==&linkurl=Li9oZHJfaW9fRjkwLmh0bWwjcm9ibzQ3OQ==&attrtype=href
>
> You are correct that the documentation is out of date, but only by
> missing the usewvl variable on the end, but that's all. I have updated
> the documentation in our current development version.
>
> The added bytes you find between fform and date come from the
> compiler. Note that the 0e 00 00 00 is repeated before each record,
> then I'm not sure where the 228 comes from (can you check your output
> to see if bantot is 1 or 228? For the density, probably 1...).
>
> Try making a test binary file with your f90 compiler, looking at the
> beginning and end of each record (or check for the end of the 2nd
> record of you DEN file.
>
>
>> I assume an integer is the usual 4 bytes.
> this can also be platform dependent, of course, and abinit uses
> default ints, which can vary.
>
>> I want to access the binary files directly, without cut3d.
> The only safe way to read back a binary fortran file is with the same
> fortran compiler! Your code will never be portable, I promise.
> You could also wrap the fortran reading, importing the hdr_io bits you
> need, or even add your functionnality into cut3d :)
>
>> I can't find:
>>  headform, fform, bantot, pertcase, ecutdg, ecut_eff
> Indeed these are not input variables, they are internal ones. Some
> description is given in the source file cited above.
>
> Matthieu
>
>
> --
> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> Dr. Matthieu Verstraete
>
> European Theoretical Spectroscopy Facility (ETSF)
> Dpto. Fisica de Materiales,
> U. del Pais Vasco,
> Centro Joxe Mari Korta, Av. de Tolosa, 72,   Phone: +34-943018393
> E-20018 Donostia-San Sebastian, Spain        Fax  : +34-943018390
>
> Mail : matthieu.jean.verstraete@gmail.com
> http://www-users.york.ac.uk/~mjv500
>
>