| 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 |
| 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> |
