possible ifort bug ???

From: flubie (fluberman72_at_wanadoo.nl)
Date: 02/24/04 

Date: Tue, 24 Feb 2004 11:56:29 +0100

I'm not sure if I'm correct, but I think I found a bug in ifort (intel
fortran 8). I did a test to arrays and print out the results. The source
file is attached. I compiled it with both ifort and lahey fortran under
linux (suse 9) running on intel P-4, and ifort gave incorrect results.

Here are the output of lahey fortran, which is correct (I cut only to
several lines):

3 5 alpha= 0.1554151707572100 wm= 0.3626469000000000 mu= 0.6950514000000000
3 6 alpha= 0.000000000000000E+00 wm= 0.1609517000000000 mu=
0.9656013000000000
4 1 alpha= 0.1554151707572100 wm= 0.1609517000000000 mu= -0.9656013000000000

and the following are from ifort:

3 5 alpha= 0.155415170757210 wm= 0.362646900000000 mu= 0.695051400000000
3 6 alpha= 0.155415170757210 wm= 0.000000000000000E+000 mu=
0.965601300000000
4 1 alpha= 0.155415170757210 wm= 0.160951700000000 mu= -0.965601300000000

Please take a look at the second line of each output.

lahey fortran: wm= 0.1609517037868500
ifort: wm= 0.000000000000000E+000

in the source code, wm(3,6) is defined as

        wm(3,6) = wm1

and wm1 is

        real(8), parameter :: wm1 = 0.1609517d0

Can anybody confirm this?

Relevant Pages

  • Re: allocating arrays, trouble
    ... the funny thing I now tried to compile it with g95. ... But anyway this if Fortran. ... The ifort ... program to work with ifort, perhaps with smaller arrays, in a test version, ...
    (comp.lang.fortran)
  • Re: [Intel Compiler 8.1] C++ calling Fortran Routine
    ... Fortran compiler earlier than ifort 8.0, you will need that library, and you ... but you need to know the additional libraries involved in icpc ...
    (comp.lang.fortran)
  • Re: [Intel Compiler 8.1] C++ calling Fortran Routine
    ... Fortran compiler earlier than ifort 8.0, you will need that library, and you ... but you need to know the additional libraries involved in icpc ...
    (comp.lang.cpp)
  • slow I/O with ifort compared to g77 !!!
    ... I have been comparing the performance of g77 ... I note that ifort gives generally slightly ... I wrote the following little FORTRAN code: ... CPU: 17.2200003Wall: 18. ...
    (comp.lang.fortran)
  • Re: bye bye IVF (was: bye bye CVF)
    ... >> I personally thought CVF was a great product. ... >> one is expected to migrate to IVF. ... If Intel, in contrast to AMD, can't even get its chip microcode right, why ... The Help button is not yet supported in the Fortran AppWizard ...
    (comp.lang.fortran)