i < x or i != x?

Which would be the best/fastest way for evaluating the loop condition,

Code:

while (i < x)

or

Code:

while (i != x)

? Assuming that the loop is incrementing i for each iteration, to avoid elitist remarks

Well assuming you know that i will at some point be exactly x, then there is no difference. But consider any possibility that might result in i never equaling x. Such as this:

Code:

float i = 0.0;
while (i != 1.0) {
  printf("%.2f\n", i);
  i += .1;
}

Yes, the value of i will be the same as x's at some point, they are both integers and i is incrementing by one at a time.

Originally Posted by OnionKnight

Which would be the best/fastest way for evaluating the loop condition,

Code:

while (i < x)

or

Code:

while (i != x)

? Assuming that the loop is incrementing i for each iteration, to avoid elitist remarks

I'm not an assembly expert, but I know on at least one architecture (ARM) those will both take exactly the same amount of time; cmp and then beq or bge (or bne vs blt); half a dozen in one hand and 6 in the other.

But even if some architectures have one compare go faster than the other, don't give it any thought. Micro-optimization is bad. Forget "fastest" and just concentrate on "best" - whatever makes sense, is easy to understand and maintain. Always go good coding style.

Yes I usually leave stuff like that up to the compiler but I figured that != would be the same as ! and ==, two operations whereas < would be just one but it seems like they actually do the same amount of work.
I've come to the conclusion that I'll use != in loops where the value will become the other at some point and < otherwise.

just for fun:

Code:

int
main(void)
{
	unsigned i, start, stop, first, second, max = UINT_MAX;
	i = 0;
	start = clock();
	while(i < max) i++;
	stop = clock();
	first = stop - start;
	i = 0;
	start = clock();
	while(i != max) i++;
	stop = clock();
	second = stop - start;
	cout << "(i < max): " << first << endl;
	cout << "(i != max): " << second << endl;
}

the first one seems to execute *slightly* faster than the second on my system, but it's negligible. I'd say follow moi's advice.