--- gforth/float.fs 2000/08/09 20:04:05 1.31 +++ gforth/float.fs 2002/11/02 15:26:03 1.38 @@ -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 @@ -100,9 +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 @@ -110,14 +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 . ; @@ -175,6 +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 @@ -214,6 +222,19 @@ IS interpreter-notfound fnegate f~rel 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 \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.