Doug Manzer wrote:
>
> Hi David, thanks for sending the file (AN575). You wrote:
>
> >FLOAT produces - I had to divide all numbers passed to fp routines by 4 to
 get
> >correct results.
>
> Oh, no, not more bad news!
>
> >I'm using fp24.a16 v 2.02 downloaded in early March and it
> >seems to work OK.
>
> What an adventure.

The floating-point math for PIC16/17 released on the bbs has been fully
tested (100000 test vectors per operation). So I am confident about its
accuracy. As for the C program, I do not believe that it was updated
properly :-(

I am attaching a fortran version plus the executable. Hope that it is OK.
I will post on the bbs as well.
--
Regards,
Al Lovrich                        email: Al.Lovrich@Microchip.COM
Microchip Technology, Inc.        Tel: (602) 786-7602
2355 W Chandler Blvd              Tel: (602) 917-4311
Chandler, AZ  85224

Content-Type: application/octet-stream
Content-Disposition: attachment; filename="FPREP.EXE"

Attachment converted: wonderlandthree:FPREP.EXE (????/----) (00003F0D)
      PROGRAM FPREP
C
      IMPLICIT NONE
C
      REAL*8 X,Y,Z
      REAL*4 W
      INTEGER*4 IW,IU
      INTEGER*1 B(4),C(4)
      EQUIVALENCE(W,B(1))
      EQUIVALENCE(W,IW)
      EQUIVALENCE(IU,C(1))
C
      INTEGER K,N,P,A(32)
      N=32
C
1     PRINT 100
100   FORMAT(1X,'ENTER Y: ',$)
      READ *,Y
      W=SNGL(Y)
C
      Z=DLOG(DABS(Y))/DLOG(2.D0)
      P=INT(Z)
      IF(DBLE(P).GT.Z)P=P-1
      X=DABS(Y)*2.D0**(-P)
      DO K=1,N
         IF(X.GE.2.D0**(-K+1))THEN
            A(K)=1
         ELSE
            A(K)=0
         ENDIF
      X=X-DBLE(A(K))*2.D0**(-K+1)
      ENDDO
C
      PRINT 300,P,A(1),(A(K),K=2,N)
300   FORMAT(1X,'e = ',I4,3X,'f = ',I1,'.',3I1,1X,7(4I1,1X))
C
      PRINT 200,W
200   FORMAT(1X,'IEEE FORMAT: ',Z8)
C
      C(1)=B(1)
      C(2)=B(2)
      C(3)=B(3)
      IW=ISHC(IW,1)
      C(4)=B(4)
      IF(Y.LT.0.)THEN
          IU=IBSET(IU,23)
      ELSE
          IU=IBCLR(IU,23)
      ENDIF
      PRINT 400,IU
400   FORMAT(1X,'PIC  FORMAT: ',Z8)
C
      GOTO  1
C
      END