generating random integers from 0 to 17576999

[chickenandfries]

RAND_MAX for my system is 32767 so rand() cannot generate an integer greater than this. The only method I think would work is

int generateRandomInt() {
int randomInt=0;
for (int i=0; i<536; ++i)
randomInt+=rand();
return randomInt;
}

Is there any other way?

[tabstop]

You can do this, but you won't get a uniform distribution (for example, to get 0 out of this function, rand() would have to be 0 536 times in a row). If this is a problem, well, you can learn how to write a random number generator, or see if you can find another one.

[chickenandfries]

You can do this, but you won't get a uniform distribution (for example, to get 0 out of this function, rand() would have to be 0 536 times in a row).

I realized that too. I tried modifying the code to

int generateRandInt() {
	int randInt=0, numOfLoops=rand()%537;
	for (int i=0; i<numOfLoops; ++i)
		randInt+=rand();
	return randInt;
}

and the distribution got a lot better.

[cpjust]

Couldn't you call rand() twice, shift the first one left 16 bits and then OR them together to get a 32 bit random number?

unsigned int randHigh = (rand() << 16);
unsigned int randNum = (randHigh | rand());

Whether this produces a very good random number or not, I'm not sure? To make the range smaller, just tweak the left shift a bit and add a % operator...

[tabstop]

I think you might get something more uniform with code like

int multipilier = rand()%537; 
int randInt = multiplier * 32768 + rand();

and of course *32768 might be represented as a bit-shift, if you feel like that sort of thing. Whether or not this matters depends on what you're doing with this thing.

Edit: My point being that you've swung the pendulum the other way: 0 is now way more likely than other numbers (a 1/537 chance instead of a 1/17million chance).

[laserlight]

Instead of relying on rand(), you could use the Mersenne Twister. Prelude has an implementation that will leave you eternally confuzzled.

[Thantos]

do a float divide (vs an integer divide) of the number you get from rand() by RAND_MAX, multiply it by the difference between your high and low, and then add your low.

[anon]

This way you'd never get certain values (if RAND_MAX is smaller than desired range)?

[lruc]

I not sure at all about this, but since RAND_MAX is just a #define, couldn't you just change the max value to the number you want?

[Thantos]

I not sure at all about this, but since RAND_MAX is just a #define, couldn't you just change the max value to the number you want?

I don't think that will work. That would be like trying to change INT_MAX in order to change the range of an int. Those are more for reporting then an actual setting.

[laserlight]

I not sure at all about this, but since RAND_MAX is just a #define, couldn't you just change the max value to the number you want?

If I read the C99 standard correctly, that will result in undefined behaviour. In any case, the purpose of RAND_MAX is to show the upper limit of the range of the return value of rand(). Merely changing RAND_MAX without changing the implementation of rand() is asking for trouble.

[robwhit]

No. RAND_MAX is indicative of the capabilities of rand(), not a parameter.