A Question

[scathenC]

Hi All,

I have an unsigned int that I use to keep track of the frame I am in. Through out my code I need to know the difference between the current frame I am in and other frames that happened in the past.

If the current frame was 5000 and the frame I am checking against is 4900, then that computation is easy (5000 - 4900 = 100 difference). The problem is when the frame reaches the limit of unsigned ints and loops back to 0, then I need to calculate the difference as follows. Lets say the current frame is 10, and the past frame is 4000. I cannot use 10 - 4000, because the difference is incorrect, so I have to use 10 - 4000 + UINT_MAX.

I was wondering if anyone had any ideas how to do this in one single formula without having to check if there was a loop around (or had any other ideas)?

[Zaccheus]

To flip the sign of unsigned integers, you normally flip the bits and add one:

 // 5 - 3 = -2
unsigned a = 5;
unsigned b = 3;
unsigned c = a + ((~b)+1);
unsigned cc = ((~c)+1);

I think that should work.

Why not just use the signed int type?

Do you really run out of numbers? Even at 1000 FPS you should have 4 million seconds, that's 46 days using a 32 bit unsigned int.

Alternatively, does your compiler support 64 bit ints?

[couling]

I cannot use 10 - 4000, because the difference is incorrect

Huh? Processors, When adding or subtracting should remain correct past the wrap around. They have a "carry" flag to idicate that this has happened allowing a 32bit processor to be used to add or subtract a 64bit integer.

But hey, I might be wrong. Try the following code and then post back its result:

#include <iostream>

using namespace std;

int main(int argc, char* argv[])
{
	unsigned int a = 0xFFFFFFFF; // max integer value
	unsigned int b = a + 1000;

	cout << "a is: " << a << endl
		 << "b is: " << b << endl
		 << "difference is: " << b-a << endl;

	return 0;
}

Flipping the sign wont help here.

To flip the sign of unsigned integers, you normally flip the bits and add one:

 // 5 - 3 = -2
unsigned a = 5;
unsigned b = 3;
unsigned c = a + ((~b)+1);
unsigned cc = ((~c)+1);

I think that should work.

And "two's complement" ((~x)+1) is the same as (0-x)

Hope this helps