>> Who can write the shortest multiply and division routines ? Let's have a
>> contest... The one who can divide (16 bit / 8 bit) and multiply ( 16 *
16)
>> with the least clockcycles, wins two free entrances for THE MSX FAIR OF
THE
>> WORLD (Tilburg 97).
What do you want? a SHORTEST or a FASTEST routine? Shortest does not
necessarily
imply fastest! Please make up your mind ;)
And what multiply/division routines do we have to make? I see the
following possibilities
operands result note
8b * 8b => 8b with overflow check
8b * 8b => 8b no overflow check
8b * 8b => 16b
16b * 16b => 16b with overflow check
16b * 16b => 16b no overflow check
16b * 16b => 32b
16b / 16b => 16b
8b / 8b => 8b
Please let us know what routines are desired.
Oh, and one last thing. How do you count the speed? The speed of multiply
and division routines depends on the values of the operands. So, most
multiply/division routines use a minimum, a maximum and an average number
of clockcycles. What do you want to have minimized (assuming you want a
fastest routine), the minimum, the maximum or the average clockcycle count?
And one more last question: do you want us to calculate the clockcycles
ourselves or do you have some independent commision/group calculating the
clockcycles.
By the way: what do you plan to count, the number of T-cycles as printed in
almost each Z80 programming book, or the real number of cycles used on the
MSX where a wait-state is inserted in each M1 (instruction fetch) cycle.
And I keep asking questions: what happens with the rights of the routines?
Does sending a routine to the MSX GG Tilburg imply that they get the
(intellectual) rights to the routine? Or do these rights remain property
of the author? Or do you have some other idea about what should happen
with these rights?
Best regards,
Alex Wulms
Ps. this memo is not ment as anti-contest or something like that. I support
the
idea of such the contest completely! This are only some questions which
rose to
my mind when I decided to try to win the free entrance cards...
