I don't have these at hand, but you should be able to write them
quickly using these algorithms:

24 x 24 Multiplication algorithm:
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D

        x5:0 =3D x2:0 * y2:0,

x2:0 - input1,
y2:0 - input2,
x5:0 - output,
0 - least significant byte.

1) clear sum x5:3 =3D 0
2) clear carry
3) shift x2:0 right

4) if carry set, add y2:0 to x5:3
5) shift x5:0 right (carry from previous addition shifted to MSb and
next bit of input to carry)
6) repeat 23 more times from 4

48 / 24 Division algorithm:
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D

        x5:0 =3D x5:0 / y2:0

1) Check y2:0 for zero
2) Clear temporary Temp2:0 =3D 0
3) Set up counter Counter =3D 48

4) Shift left x5:0
5) Shift left Temp2:0 to get next bit from x5:0
6) Save carry (MSb of Temp might be set), e.g. in Counter<7>
7) Substract y2:0 from Temp2:0
8) If carry=3D1 (no borrow), set Counter<7>
9) If Counter<7>=3D0 (borrow and Msb of Temp was zero), add y2:0 to Temp2=
:0 to
restore it
10) carry=3DCounter<7> (next bit of result)
11) clear Counter<7>
12) decrement Counter, and repeat from 4 if not zero
13) Final shift left x5:0

Since result is known to fit in 3 bytes, then first 24 loops can be
skipped. It would save also 3 temporary bytes Temp2:0.


 Nikolai

---- Original Message ----
From: K=FCbek Tony
Sent: Wednesday, August 16, 2000 17:20:54
  To: PICLIST@MITVMA.MIT.EDU
Subj: [PIC]: Number crunching.. 24x24 and 48/24

> Hi,
> I dont have much hope of anyone supplying these but
> anyway, as I'm far to busy at the moment, here goes:

> I need routines for:

> 24 bit X 24 bit =3D 48 bit=20

> and

> 48 bit / 24 bit =3D 24 bit ( last bit properly rounded )
> ( and yes the result will alwyas fit inside 24 bits )

> Anyone have these laying around.. ;-)

> Code size over performance, as few temp vars as possible.

> I really need the accuracy, the end result from a few
> formulas are 'only' 24 bit, but they are using quite a bit of
> math in between.

> In case of no replys, I'll make them myself and post them later
> ( next month or so ), when I have a bit more time.



> /Tony


> Tony K=FCbek, Flintab AB           =20
> =B2=B2=B2=B2=B2=B2=B2=B2=B2=B2=B2=B2=B2=B2=B2=B2=B2=B2=B2=B2=B2=B2=B2=B2=
=B2=B2=B2=B2=B2=B2=B2=B2=B2=B2=B2=B2=B2=B2=B2=B2=B2=B2=B2=B2=B2=B2=B2=B2=B2=
=B2=B2=B2=B2=B2=B2=B2=B2=B2=B2=B2
> E-mail: tony.kubek@flintab.com
> =B2=B2=B2=B2=B2=B2=B2=B2=B2=B2=B2=B2=B2=B2=B2=B2=B2=B2=B2=B2=B2=B2=B2=B2=
=B2=B2=B2=B2=B2=B2=B2=B2=B2=B2=B2=B2=B2=B2=B2=B2=B2=B2=B2=B2=B2=B2=B2=B2=B2=
=B2=B2=B2=B2=B2=B2=B2=B2=B2=B2=B2

--
http://www.piclist.com hint: To leave the PICList
mailto:piclist-unsubscribe-request@mitvma.mit.edu