Diff for /gforth/float.fs between versions 1.30 and 1.38

version 1.30, 2000/01/17 00:04:29 version 1.38, 2002/11/02 15:26:03
Line 16 Line 16
   
 \ You should have received a copy of the GNU General Public License  \ You should have received a copy of the GNU General Public License
 \ along with this program; if not, write to the Free Software  \ along with this program; if not, write to the Free Software
 \ Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.  \ Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111, USA.
   
 \ 1 cells 4 = [IF]  \ 1 cells 4 = [IF]
 \ ' cells   Alias sfloats  \ ' cells   Alias sfloats
Line 52 Line 52
     here dup dfaligned swap ?DO  bl c,  LOOP ;      here dup dfaligned swap ?DO  bl c,  LOOP ;
   
 1 sfloats (Field) sfloat+ , ( sf-addr1 -- sf-addr2 ) \ float-ext s-float-plus  1 sfloats (Field) sfloat+ , ( sf-addr1 -- sf-addr2 ) \ float-ext s-float-plus
 \G Increment @i{sf-addr1} by the number of address units corresponding to the size of  \G @code{1 sfloats +}.
 \G a single-precision IEEE floating-point number, to give @i{sf-addr2}.""  
   
 1 dfloats (Field) dfloat+ , ( df-addr1 -- df-addr2 ) \ float-ext d-float-plus  1 dfloats (Field) dfloat+ , ( df-addr1 -- df-addr2 ) \ float-ext d-float-plus
 \G Increment @i{df-addr1} by the number of address units corresponding to the size of  \G @code{1 dfloats +}.
 \G a double-precision IEEE floating-point number, to give @i{df-addr2}.""  
   
 : f, ( f -- ) \ gforth  : f, ( f -- ) \ gforth
     \G Reserve data space for one floating-point number and store      \G Reserve data space for one floating-point number and store
Line 102  DOES> ( -- r ) Line 100  DOES> ( -- r )
   IF  '- emit  THEN ;    IF  '- emit  THEN ;
   
 : f.  ( r -- ) \ float-ext f-dot  : f.  ( r -- ) \ float-ext f-dot
 \G Display (the floating-point number) @i{r} using fixed-point notation,  \G Display (the floating-point number) @i{r} without exponent,
 \G followed by a space.  \G followed by a space.
   f$ dup >r 0<    f$ dup >r 0<=
   IF    '0 emit    IF    '0 emit
   ELSE  scratch r@ min type  r@ precision - zeros  THEN    ELSE  scratch r@ min type  r@ precision - zeros  THEN
   '. emit r@ negate zeros    '. emit r@ negate zeros
Line 112  DOES> ( -- r ) Line 110  DOES> ( -- r )
 \ I'm afraid this does not really implement ansi semantics wrt precision.  \ I'm afraid this does not really implement ansi semantics wrt precision.
 \ Shouldn't precision indicate the number of places shown after the point?  \ Shouldn't precision indicate the number of places shown after the point?
   
   \ Why do you think so? ANS Forth appears ambiguous on this point. -anton.
   
 : fe. ( r -- ) \ float-ext f-e-dot  : fe. ( r -- ) \ float-ext f-e-dot
 \G Display @i{r} using engineering notation, followed by a space.  \G Display @i{r} using engineering notation (with exponent dividable
   \G by 3), followed by a space.
   f$ 1- s>d 3 fm/mod 3 * >r 1+ >r    f$ 1- s>d 3 fm/mod 3 * >r 1+ >r
   scratch r@ min type '. emit  scratch r> /string type    scratch r@ tuck min tuck - >r type r> zeros
     '. emit scratch r> /string type
   'E emit r> . ;    'E emit r> . ;
   
 : fs. ( r -- ) \ float-ext f-s-dot  : fs. ( r -- ) \ float-ext f-s-dot
 \G Display @i{r} using scientific notation, followed by a space.  \G Display @i{r} using scientific notation (with exponent), followed
   \G by a space.
   f$ 1-    f$ 1-
   scratch over c@ emit '. emit 1 /string type    scratch over c@ emit '. emit 1 /string type
   'E emit . ;    'E emit . ;
Line 177  IS interpreter-notfound Line 180  IS interpreter-notfound
     \G Divide 1.0e0 by @i{r1}.      \G Divide 1.0e0 by @i{r1}.
     1.0e0 fswap f/ ;      1.0e0 fswap f/ ;
   
   get-current environment-wordlist set-current
   1.7976931348623157e308 FConstant max-float
   set-current
   
 \ We now have primitives for these, so we need not define them  \ We now have primitives for these, so we need not define them
   
Line 201  IS interpreter-notfound Line 207  IS interpreter-notfound
         f- fabs frot frot f* f< ;          f- fabs frot frot f* f< ;
   
 : f~ ( r1 r2 r3 -- flag ) \ float-ext f-proximate  : f~ ( r1 r2 r3 -- flag ) \ float-ext f-proximate
     \G ANS Forth medley: r3>0: @code{f~abs}; r3=0: r1=r2; r3<0: @code{fnegate f~rel}.      \G ANS Forth medley for comparing r1 and r2 for equality: r3>0:
       \G @code{f~abs}; r3=0: bitwise comparison; r3<0: @code{fnegate f~rel}.
     fdup f0=      fdup f0=
     IF      IF \ bitwise comparison
         fdrop f=  \ !! this does not work, because 0=-0 with f= on Linux-Intel          fp@ float+ 1 floats over float+ -text 0=
                   \ the standard says they should compare unequal          fdrop fdrop fdrop
                   \ the comparison should be done with COMPARE  
         EXIT          EXIT
     THEN      THEN
     fdup f0>      fdup f0>
Line 216  IS interpreter-notfound Line 222  IS interpreter-notfound
         fnegate f~rel          fnegate f~rel
     THEN ;      THEN ;
   
   1e 53 0 [do] f2* [loop] fconstant fround-offset \ 2^53
   
   : fround ( r1 -- r2 ) \ float f-round
   \G Round to nearest integral value.  Break ties with round-to-even.
       \ assumes IEEE DP FP in round-to-even mode.
       \ for an explanation of this code read
       \ <2002Oct26.113823@a0.complang.tuwien.ac.at> ff.
       fdup f0> if
           fround-offset f- fround-offset f+
       else fdup f0< if \ leave 0e and -0e as is
               fround-offset f+ fround-offset f-
       then then ;
   
 : f.s ( -- ) \ gforth f-dot-s  : f.s ( -- ) \ gforth f-dot-s
     \G Display the number of items on the floating-point stack,      \G Display the number of items on the floating-point stack,
     \G followed by a list of the items; TOS is the right-most item.      \G followed by a list of the items; TOS is the right-most item.

Removed from v.1.30  
changed lines
  Added in v.1.38


FreeBSD-CVSweb <freebsd-cvsweb@FreeBSD.org>