Cannot read an unformatted file generated on IBM platform

[Ludovic_Sauge1]

Hi all!
This this my first post on the Intel Software Network!

I have an issue using ifort (10.1.015). A simple code read an unformatted file created on an IBM plateform. This is reading and then some results writing on he standart output.
This file is opend using the following statements,

open(10,file=myfile,form='unformatted',access='direct',recl=irecl&
&,convert='big_endian')

The code is compiled using '-assume byterecl' flag.
Unfortunately, the results appear false (including some NaNs!).

But using gfortran, the results are now correct.

Any idea to reproduce the gfortran beahavior with ifort or the unformatted file generatded on the IBM platform is only campatible with gfortran.

Regards,
Ludo

[Ron_Green]

Please show the gfortran open statement and any compiler switches used to specify IO characteristics.

thanks

ron

[TimP]

Many IBM platforms are little-endian. Some compilers in use on IBM platforms have different direct access file formats, even when the endian-ness matches what you set.

[Ludovic_Sauge1]

Hi ron, thanks for your support.
I compiled the gfortran version without any flag.
Version of gfortran is GNU Fortran (GCC) 4.1.2 20070626 (Red Hat 4.1.2-14)
cheers,
Ludo

[Steven_L_Intel1]

I would suggest, then, that this is not a big-endian file and you should try the ifort defaults as well.

You can do an "od -t x4" on the file to examine the record lengths and data layout.