Continuing weirdness, cool division (and also cool subtraction)

and comparasion:
a<b
coolsub(a,b)&high_bit
a<=b
(coolsub(b,a)&high_bit)==0
> and >= is the same.

edit:as about subtraction via addition,

i'd rather do it in that way:
int coolsub2(int a,int b){
return cooladd(cooladd(~a,~b),2);
};
,rather than using coolmultiply to multiply with (~0) :).

And important thing, coolmultiply need simple fix to work with signed numbers, same as one used in cooldiv.

[Edited by - Dmytry on September 2, 2004 6:46:30 AM]

I have already written your cooladd,coolmultiply, etc. functions in INTERCAL and TriINTERCAL (which is a super-obfuscated and quirky language which should never have been born - Blame Eric S. Raymond). Did the interviewer specify a language to write in?

In INTERCAL or any dialect thereof(http://www.catb.org/~esr/intercal), the program becomes:

(100) DO (1020) NEXT
DO RESUME #1

Which increments the 16-bit variable .1. Note the lack of equals signs. It also has the advantage of working in binary, ternary, all the way up to base 7.


As an extra, memory-guzzling idea, I thought also you could use Standard Template Library containers and the functional algorithms.
You could create a linked list and use the functional library to fill it with a numeric sequence, starting at 0, ending at 2^(sizeof(unsigned int)*8)-1, so that list[x] == x for all possible unsigned x values.
Get an iterator to list[x], increment the iterator, and you have &list[x+1], which should equal x+1.

I dont know how you increment the iterator without using +'s , but i'm sure there's a way. In any case, the INTERCAL routine was the main addition to this thread, useless though it may be. :)

Quote:Original post by Dmytry
And important thing, coolmultiply need simple fix to work with signed numbers, same as one used in cooldiv.

Actually, my coolmultiply works for signed numbers too, I have tested it. Or did you find some specific numbers it fails for?

oh,my bad... it's because i know processor have different operations for signed multiply and unsigned ,and have never thinked about it.
In fact it's just only because result of processor's 32bit multiply is 64-bit and there's real difference.

Quote:
Melekor:
Actually, my coolmultiply works for signed numbers too, I have tested it. Or did you find some specific numbers it fails for?

Yes, my intercal routine does the same thing - it was written for unsigned integers, but it works for signed numbers in 2's complement format and the ternary version works in unsigned form, 3's complement form and even balanced ternary..