Logic Required

[rohit]

hi!
actually,i want to write a program in C.the specifications are as below:-
1. Input should be a two digit number(say ab)
2. If a>b then Output -1
3. If a<b then Output +1
4. If a==b then Output 0
5. But we have not to use divide(/) or modulo(%) operator.
6. I think,by using bitwise operators,we may get our goal.But
HOW?
Please give logic behind it??????????????

regards

Rohit..

[axion_sa]

why get so complicated? it's a very small program and any
increase in speed (if any) will be insignificant.

just use normal equality expressions (==, >, <)

[DarkElf007]

You can mutiply adding up shifts e.g.

x * 10 = x<<1 + x<<3

<<1 (*2)
<<2 (*4)
<<3 (*8)
<<4 (*16)

etc.

[CornedBee]

The problem is seperating A and B. The ususal way is using / and % with 10, but this is forbidden.

You could use hex digits and claim that they are digits too. Then it is easy:
a = ab >> 4;
b = ab & 0x0f;

Decimal digits are far harder.
it takes a lot of bitmask and shifting (I can't think of a way), or you can simulate the operation a cpu does when dividing: substract and count

// do a division and modulo at the same time
int ab = GetInput();
int temp = ab;
int cnt = 0;
int result = -2; // error by default
while(temp >= 10)
{
  temp -= 10;
  ++cnt;
}
// now cnt == ab / 10, temp == ab % 10
if(cnt == ab) result = 0;
if(cnt < ab) result = -1;
if(cnt > ab) result = 1;
return result;

That's a dirty trick however, and may be forbidden.

[kedaman]

Cheating now hehe, Msvc7 came up with this solution:

  00013    b8 cd cc cc cc     mov     eax, -858993459        ; cccccccdH
  00018    f7 64 24 0c     mul     DWORD PTR _x$[esp+16]
  0001c    c1 ea 03     shr     edx, 3 

so in C++ that would be:

    unsigned int x,y;
    cin>>x;
    x-=10*(y=(x*0xcccd)>>19);
    if(y == x) result = 0;
    if(y < x) result = -1;
    if(y > x) result = 1;
    cout<<result; 

[CornedBee]

so the workaround is the same as:
x / 10 == x * (m/10) / m
and we are searching for an m that is a power of 2 and can be divided by 10 evenly. There is no such number, but at 0xcccd, which is 52429, the difference to 2^19 (= 524288) is so small compared to the magnitude of the numbers, that the relatively unprecise shift-right division will come up with the right result.

MSVC 7 is really impressive. That thing must be possible with many numbers.
There should be some optimization like "optimize calls to pow, sqrt, sin etc. with constant parameters away". This would preserve code readability without the necessity of compile time template calculations, yet cause no impact on runtime.

[kedaman]

There should be some optimization like "optimize calls to pow, sqrt, sin etc. with constant parameters away". This would preserve code readability without the necessity of compile time template calculations, yet cause no impact on runtime.

Might be, I haven't tried out yet, that would quite beat the purpose of templates in a lot of cases indeed in that case it could threaten BORK but it has a lot of other intuitive features, which a C++ programmer can't dream of