Diff for /gforth/float.fs between versions 1.8 and 1.68

version 1.8, 1994/09/12 19:00:29 version 1.68, 2012/12/31 15:25:18
Line 1 Line 1
 \ High level floating point                            14jan94py  \ High level floating point                            14jan94py
   
 1 cells 4 = [IF]  \ Copyright (C) 1995,1997,2003,2004,2005,2006,2007,2009,2010,2011,2012 Free Software Foundation, Inc.
 ' cells   Alias sfloats  
 ' cell+   Alias sfloat+  
 ' align   Alias sfalign  
 ' aligned Alias sfaligned  
 [ELSE]  
 : sfloats  2* 2* ;  
 : sfloat+  4 + ;  
 : sfaligned ( addr -- addr' )  3 + -4 and ;  
 : sfalign ( -- )  here dup sfaligned swap ?DO  bl c,  LOOP ;  
 [THEN]  
   
 1 floats 8 = [IF]  
 ' floats   Alias dfloats  
 ' float+   Alias dfloat+  
 ' falign   Alias dfalign  
 ' faligned Alias dfaligned  
 [ELSE]  
 : dfloats  2* 2* 2* ;  
 : dfloat+  8 + ;  
 : dfaligned ( addr -- addr' )  7 + -8 and ;  
 : dfalign ( -- )  here dup dfaligned swap ?DO  bl c,  LOOP ;  
 [THEN]  
   
 : f, ( f -- )  here 1 floats allot f! ;  
   
 \ !! have create produce faligned pfas  
 : fconstant  ( r -- )  
   falign here f,  Create A,  
   DOES>  @ f@ ;  
   
 : fvariable  \ This file is part of Gforth.
   falign here 0. d>f f, AConstant ;  
   
 : fdepth  ( -- n )  f0 @ fp@ - [ 1 floats ] Literal / ;  \ Gforth is free software; you can redistribute it and/or
   \ modify it under the terms of the GNU General Public License
   \ as published by the Free Software Foundation, either version 3
   \ of the License, or (at your option) any later version.
   
   \ This program is distributed in the hope that it will be useful,
   \ but WITHOUT ANY WARRANTY; without even the implied warranty of
   \ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
   \ GNU General Public License for more details.
   
   \ You should have received a copy of the GNU General Public License
   \ along with this program. If not, see http://www.gnu.org/licenses/.
   
   \ 1 cells 4 = [IF]
   \ ' cells   Alias sfloats
   \ ' cell+   Alias sfloat+
   \ ' align   Alias sfalign
   \ ' aligned Alias sfaligned
   \ [ELSE]
   \ : sfloats  2* 2* ;
   \ : sfloat+  4 + ;
   \ : sfaligned ( addr -- addr' )  3 + -4 and ;
   \ : sfalign ( -- )  here dup sfaligned swap ?DO  bl c,  LOOP ;
   \ [THEN]
   
   \ 1 floats 8 = [IF]
   \ ' floats   Alias dfloats
   \ ' float+   Alias dfloat+
   \ ' falign   Alias dfalign
   \ ' faligned Alias dfaligned
   \ [ELSE]
   \ : dfloats  2* 2* 2* ;
   \ : dfloat+  8 + ;
   \ : dfaligned ( addr -- addr' )  7 + -8 and ;
   \ : dfalign ( -- )  here dup dfaligned swap ?DO  bl c,  LOOP ;
   \ [THEN]
   
   : sfalign ( -- ) \ float-ext s-f-align
       \G If the data-space pointer is not single-float-aligned, reserve
       \G enough space to align it.
       here dup sfaligned swap ?DO  bl c,  LOOP ;
   : dfalign ( -- ) \ float-ext d-f-align
       \G If the data-space pointer is not double-float-aligned, reserve
       \G enough space to align it.
       here dup dfaligned swap ?DO  bl c,  LOOP ;
   
   (Field) sfloat+ ( sf-addr1 -- sf-addr2 ) \ float-ext s-float-plus
   \G @code{1 sfloats +}.
       1 sfloats ,
   
   (Field) dfloat+ ( df-addr1 -- df-addr2 ) \ float-ext d-float-plus
   \G @code{1 dfloats +}.
       1 dfloats ,
       
   : f, ( f -- ) \ gforth
       \G Reserve data space for one floating-point number and store
       \G @i{f} in the space.
       here 1 floats allot f! ;
   
   : fconstant  ( r "name" -- ) \ float f-constant
       Create f,
   DOES> ( -- r )
       f@ ;
   
   : fdepth ( -- +n ) \ float f-depth
       \G @i{+n} is the current number of (floating-point) values on the
       \G floating-point stack.
       fp0 @ fp@ - [ 1 floats ] Literal / ;
   
   : FLiteral ( compilation r -- ; run-time -- r ) \ float f-literal
       \G Compile appropriate code such that, at run-time, @i{r} is placed
       \G on the (floating-point) stack. Interpretation semantics are undefined.
       BEGIN  here cell+ cell+ dup faligned <>  WHILE  postpone noop  REPEAT
       postpone ahead here >r f, postpone then
       r> postpone literal postpone f@ ;  immediate
   
   &15 Value precision ( -- u ) \ float-ext
   \G @i{u} is the number of significant digits currently used by
   \G @code{F.} @code{FE.} and @code{FS.} 
   : set-precision ( u -- ) \ float-ext
       \G Set the number of significant digits currently used by
       \G @code{F.} @code{FE.} and @code{FS.} to @i{u}.
       to precision ;
   
 : FLit ( -- r )  r> faligned dup f@ float+ >r ;  : scratch ( -- addr len )
 : FLiteral ( r -- )  postpone FLit  falign f, ;  immediate  
   
 &16 Value precision  
 : set-precision  to precision ;  
   
 : scratch ( r -- addr len )  
   pad precision - precision ;    pad precision - precision ;
   
 : zeros ( n -- )   0 max 0 ?DO  '0 emit  LOOP ;  : zeros ( n -- )   0 max 0 ?DO  '0 emit  LOOP ;
Line 54 Line 100
   IF  2drop  scratch 3 min type  rdrop  EXIT  THEN    IF  2drop  scratch 3 min type  rdrop  EXIT  THEN
   IF  '- emit  THEN ;    IF  '- emit  THEN ;
   
 : f.  ( r -- )  f$ dup >r 0<  : f.  ( r -- ) \ float-ext f-dot
   \G Display (the floating-point number) @i{r} without exponent,
   \G followed by a space.
     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 62 Line 111
 \ 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?
   
 : fe. ( r -- )  f$ 1- s>d 3 fm/mod 3 * >r 1+ >r  \ Why do you think so? ANS Forth appears ambiguous on this point. -anton.
   scratch r@ min type '. emit  scratch r> /string type  
   : fe. ( r -- ) \ float-ext f-e-dot
   \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
     scratch r@ tuck min tuck - >r type r> zeros
     '. emit scratch r> /string type
   'E emit r> . ;    'E emit r> . ;
   
 : fs. ( r -- )  f$ 1-  : fs. ( r -- ) \ float-ext f-s-dot
   \G Display @i{r} using scientific notation (with exponent), followed
   \G by a space.
     f$ 1-
   scratch over c@ emit '. emit 1 /string type    scratch over c@ emit '. emit 1 /string type
   'E emit . ;    'E emit . ;
   
 : sfnumber ( c-addr u -- r / )  [IFDEF] fp-char
     2dup >float  : sfnumber ( c-addr u -- r true | false )
     IF      fp-char @ >float1 ;
         2drop state @  
   Create si-prefixes ," PTGMk munpf"
   si-prefixes count bl scan drop Constant zero-exp
   
   : prefix-number ( c-addr u -- r true | false )
       si-prefixes count bounds DO
           2dup I c@ scan nip dup 0<> IF
               1 = IF  1- fp-char @  ELSE  I c@  THEN
               >float1
               dup IF  1000 s>f zero-exp I - s>f f** f*  THEN
               UNLOOP  EXIT  THEN  drop
       LOOP
       sfnumber ;
   [ELSE]
   : sfnumber ( c-addr u -- r true | false )
       >float ;
   : prefix-number  sfnumber ;
   [THEN]
   
   [ifdef] recognizer:
       [IFDEF] 2lit,
           : flit, postpone Fliteral ;
           :noname ['] noop ;
           :noname ['] flit, ;
       [ELSE]
           ' noop
           :noname postpone Fliteral ;
       [THEN]
       dup
       recognizer: r:fnumber
   
       : fnum-recognizer ( addr u -- float int-table | addr u r:fail )
           2dup prefix-number
         IF          IF
             postpone FLiteral              2drop r:fnumber  EXIT
         THEN          THEN
     ELSE          r:fail ;
         defers notfound  
     THEN ;  
   
 ' sfnumber IS notfound  ' fnum-recognizer
   forth-recognizer get-recognizers
   1+ forth-recognizer set-recognizers
   [else]
   [ifundef] compiler-notfound1
   defer compiler-notfound1
   ' no.extensions IS compiler-notfound1
   
   :noname compiler-notfound1 execute ; is compiler-notfound
   
 1e0 fasin 2e0 f* fconstant pi  defer interpreter-notfound1
   ' no.extensions IS interpreter-notfound1
   
 : f2*  2e0 f* ;  :noname interpreter-notfound1 execute ; is interpreter-notfound
 : f2/  2e0 f/ ;  [then]
 : 1/f  1e0 fswap f/ ;  
   :noname ( c-addr u -- ... xt )
       2dup sfnumber
       IF
           2drop [comp'] FLiteral
       ELSE
           defers compiler-notfound1
       ENDIF ;
   IS compiler-notfound1
   
   :noname ( c-addr u -- ... xt )
       2dup sfnumber
       IF
           2drop ['] noop
       ELSE
           defers interpreter-notfound1
       ENDIF ;
   IS interpreter-notfound1
   [then]
   
   : fvariable ( "name" -- ) \ float f-variable
       Create 0.0E0 f, ;
       \ does> ( -- f-addr )
   
   1.0e0 fasin 2.0e0 f* fconstant pi ( -- r ) \ gforth
   \G @code{Fconstant} -- @i{r} is the value pi; the ratio of a circle's area
   \G to its diameter.
   
   : f2* ( r1 -- r2 ) \ gforth
       \G Multiply @i{r1} by 2.0e0
       2.0e0 f* ;
   
   : f2/ ( r1 -- r2 ) \ gforth
       \G Multiply @i{r1} by 0.5e0
       0.5e0 f* ;
   
   : 1/f ( r1 -- r2 ) \ gforth
       \G Divide 1.0e0 by @i{r1}.
       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
   
   \ : falog ( f -- 10^f )  [ 10.0e0 fln ] FLiteral f* fexp ;
   
   \ : fsinh    fexpm1 fdup fdup 1.0e0 f+ f/ f+ f2/ ;
   \ : fcosh    fexp fdup 1/f f+ f2/ ;
   \ : ftanh    f2* fexpm1 fdup 2.0e0 f+ f/ ;
   
   \ : fatanh   fdup f0< >r fabs 1.0e0 fover f- f/  f2* flnp1 f2/
   \            r> IF  fnegate  THEN ;
   \ : facosh   fdup fdup f* 1.0e0 f- fsqrt f+ fln ;
   \ : fasinh   fdup fdup f* 1.0e0 f+ fsqrt f/ fatanh ;
   
   : f~abs ( r1 r2 r3 -- flag ) \ gforth
       \G Approximate equality with absolute error: |r1-r2|<r3.
       frot frot f- fabs fswap f< ;
   
   : f~rel ( r1 r2 r3 -- flag ) \ gforth
       \G Approximate equality with relative error: |r1-r2|<r3*|r1+r2|.
           frot frot fover fabs fover fabs f+ frot frot
           f- fabs frot frot f* f< ;
   
   : f~ ( r1 r2 r3 -- flag ) \ float-ext f-proximate
       \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=
       IF \ bitwise comparison
           fp@ float+ 1 floats over float+ over str=
           fdrop fdrop fdrop
           EXIT
       THEN
       fdup f0>
       IF
           f~abs
       ELSE
           fnegate f~rel
       THEN ;
   
 : falog ( f -- 10^f )  [ 10e0 fln ] FLiteral f* fexp ;  -0e 8 0 [do] fp@ [i] + c@ $80 = [if] [i] constant fsign-offset [then] [loop]
   
 : fsinh    fexpm1 fdup fdup 1e0 f+ f/ f+ f2/ ;  : fcopysign ( r1 r2 -- r3 ) \ gforth
 : fcosh    fexp fdup 1/f f+ f2/ ;  \G r3 takes its absolute value from r1 and its sign from r2
 : ftanh    f2* fexpm1 fdup 2e0 f+ f/ ;      \ !! implementation relies on IEEE DP format
       fp@ fsign-offset + dup c@ $80 and >r ( r1 r2 addr-r1sign )
       float+ dup c@ $7f and r> or swap c!
       fdrop ;
   
   \ proposals from Krishna Myeni in <cjsp2d$47l$1@ngspool-d02.news.aol.com>
   \ not sure if they are a good idea
   
   : ftrunc ( r1 -- r2 ) \ X:ftrunc
       \ round towards 0
       fdup fabs floor fswap fcopysign ;
   
   : FMOD ( r1 r2 -- r )
       \ remainder of r1/r2
       FOVER FOVER F/ ftrunc F* F- ;
   
 : fatanh   fdup f0< >r fabs 1e0 fover f- f/  f2* flnp1 f2/  
            r> IF  fnegate  THEN ;  
 : facosh   fdup fdup f* 1e0 f- fsqrt f+ fln ;  
 : fasinh   fdup fdup f* 1e0 f+ fsqrt f/ fatanh ;  

Removed from v.1.8  
changed lines
  Added in v.1.68


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