CODEWITHSUNDEEP
exceptionfloating-pointfortrangfortran
distico
distico

Reputation: 45

gfortran IEEE exception inexact

We are using gfortran (5.3.1), Fedora 23, in a new 64 b machine. Compiling with simple gfortran -o (we are not using -ffpe-trap options !), excites the "classical- trivial" warning: 

Note: The following floating-point exceptions are signalling: IEEE_INVALID_FLAG

Its due to INEXACT exception (type 2.0/3.0). The DDD debugger points towards a real constant (180d0/pi; pi = 3.141518...). We don't understand why this flag appears, with this basic compilation, because these exceptions are reached all the time...

Some code here:

    Implicit none !real*8(a-h,o-z)   
    real*8 pi,dpi,radgra,TSI,TOL,xlsol,fi,W      
    Integer year, T1, k,m

      open(10,file='stof-elem.sol') 

          pi = 4.d0 * datan(1.d0)
         dpi = 2.d0 * pi
       radgra = 360.d0 / dpi !!!!!!!! HERE POINTS THE EXCEPTION!!!!!!!!!!

         T1 = -9998 !800d0 !1450d0 !

        TSI = 1360.d0 !1364.5d0 !1367d0 

        TOL = 0.7d0 / radgra ! dont' use smaller 

C...Name of the output file
      open(12,file='midmonth-2000.sal')

C-----------------------------------------------------------------------
            k = 0 ! outputs counter

            write(12,*)T1            

      DO m = 1, 12  ! select month
           IF(T1.lt.0) then
             xlsol = (270.d0 - dble(m-1) * 30.d0) / radgra !from Dec
               if(xlsol.lt.0d0) xlsol = xlsol + dpi
           ELSE
              xlsol = dble(m-3) * 30.d0 / radgra !from Jan
               if(xlsol.lt.0d0) xlsol = xlsol + dpi
           ENDIF 

            CALL MEANINSOLA(pi,dpi,radgra,TOL,T1,TSI,xlsol,fi,k,W)

            rewind(10) ! better rewind...
      ENDDO

       write(*,*) 'Outputs:', k,'lines'

The EXCEPTION appears at the definition of RADGRA ... as indicated. If redefine the constant (i.e.,, RADGRA = 57.2d0), the exception migrates to another parts using RADGRA... and so on...

Upvotes: 3

Views: 5017

Answers (1)

Vladimir F Героям слава
Vladimir F Героям слава

Reputation: 60113

As per https://gcc.gnu.org/ml/fortran/2013-06/msg00072.html the Fortran standard requires printing of these notes after executing the STOP statement.

"If any exception (14) is signaling on that image, the processor shall issue a warning indicating which exceptions are signaling; this warning shall be on the unit identified by the named constant ERROR UNIT (13.8.2.8)."

Note that even if you request the Fortran 95 standard by -std=f95 the note is still displayed.

You can control this behaviour by -ffpe-summary=, consult you compiler's manual. By default, a summary for all exceptions but ‘inexact’ is shown. Have you enabled inexact yourself somewhere?

Why is that exception signalling is a different matter, you must examine your code whether it is something you should be worried about or not. Probably you should not, inexact floating point operations are very common.

Because the message is invoked by the STOP statement, a simple way to get rid of these messages is to not terminate your program by a STOP statement, but let it reach the END PROGRAM.

Upvotes: 4

Related Questions