--- gforth/float.fs 1999/12/03 18:24:22 1.29 +++ gforth/float.fs 2002/10/27 09:57:10 1.37 @@ -16,7 +16,7 @@ \ You should have received a copy of the GNU General Public License \ 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] \ ' cells Alias sfloats @@ -51,15 +51,11 @@ \G enough space to align it. here dup dfaligned swap ?DO bl c, LOOP ; -1 sfloats constant 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 a single-precision IEEE floating-point number, to give @i{sf-addr2}."" -dofield: lastxt code-address! \ change the constant into a field - -1 dfloats constant 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 a double-precision IEEE floating-point number, to give @i{df-addr2}."" -dofield: lastxt code-address! \ change the constant into a field +1 sfloats (Field) sfloat+ , ( sf-addr1 -- sf-addr2 ) \ float-ext s-float-plus +\G @code{1 sfloats +}. + +1 dfloats (Field) dfloat+ , ( df-addr1 -- df-addr2 ) \ float-ext d-float-plus +\G @code{1 dfloats +}. : f, ( f -- ) \ gforth \G Reserve data space for one floating-point number and store @@ -104,9 +100,9 @@ DOES> ( -- r ) IF '- emit THEN ; : 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. - f\$ dup >r 0< + f\$ dup >r 0<= IF '0 emit ELSE scratch r@ min type r@ precision - zeros THEN '. emit r@ negate zeros @@ -114,14 +110,19 @@ DOES> ( -- r ) \ I'm afraid this does not really implement ansi semantics wrt precision. \ 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 -\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 - 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> . ; : 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- scratch over c@ emit '. emit 1 /string type 'E emit . ; @@ -179,6 +180,9 @@ IS interpreter-notfound \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 @@ -203,12 +207,12 @@ IS interpreter-notfound f- fabs frot frot f* f< ; : 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~abs}. + \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 - fdrop f= \ !! this does not work, because 0=-0 with f= on Linux-Intel - \ the standard says they should compare unequal - \ the comparison should be done with COMPARE + IF \ bitwise comparison + fp@ float+ 1 floats over float+ -text 0= + fdrop fdrop fdrop EXIT THEN fdup f0> @@ -218,6 +222,19 @@ IS interpreter-notfound fnegate f~rel THEN ; +2e 53e f** fconstant fround-offset \ !! does f** work correctly? + +: 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 \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.