Random Numbers Unix

**quizkiwi** · 08-30-2006

Does anyone know why this compiles but does not work on a Unix platform using g++? It works fine on a windows box using vs2005. It seems the random number is always 0. Thanks for your help!

Code:

srand(time(NULL)); 
    int random_integer, random_integernum; 
    int lowest=0, highest=200, lownum=1, highnum=50; 
    int range=(highest-lowest)+1;
	int rangenum=(highnum-lownum)+1;

	cout << "\n# of elements  Height of Tree   Log2(n)" << endl;
	Set Random;
	for(int i=0; i<20; i++)
	{
		random_integernum = lownum+int(rangenum*rand()/(RAND_MAX + 1.0));
		for(int index=0; index<random_integernum; index++)
		{ 
			random_integer = lowest+int(range*rand()/(RAND_MAX + 1.0)); 
			Insert(random_integer, Random.root);
		}
		cout << "     " << random_integernum << "             " << Maxheight(Random.root);
		cout << "            " << log2(double(random_integernum)) << endl;
		Makenull(Random.root);
	}

**Daved** · 08-30-2006

>> lownum+int(rangenum*rand()/(RAND_MAX + 1.0));
Perhaps that is causing an overflow of the integer type. RAND_MAX is different on different platforms.

**quizkiwi** · 08-30-2006

I did check on that and RAND_MAX is the same, in fact, this formula is specified in the Unix Man pages.

**DougDbug** · 08-30-2006

That does look fishy. If RAND_MAX is equal to INT_MAX, then when you increment, it “rolls-over" to zero. Now you are dividing by zero!

**Cat** · 08-30-2006

No, because + 1.0 means that RAND_MAX will be promoted to a floating-point type, which won't have rollover problems.

**major_small** · 08-30-2006

well... you're mulpitplying rangenum by a random number... who says that result can fit into an integer?

dont' forget that's happening before it gets promoted to a floating-point number...

this code

Code:

#include <iostream>
#include <ctime>

int main()
{
	srand(static_cast<short int>(time(0)));
	int x;
	for(int i=0;i<10;i++)
	{
		x=50*rand()/(RAND_MAX+1.0);
		std::cout<<x<<std::endl;
	}
}

results in this output

Code:

-1279308866
1370173278
-633427736
445617742
-2042006872
623654194
-806845720
-1524802060
431276324
542628534

and the rest of the equation:

Code:

#include <iostream>
#include <ctime>

int main()
{
	srand(static_cast<short int>(time(0)));
	int x;
	for(int i=0;i<10;i++)
	{
		x=1+int(50*rand()/(RAND_MAX+1.0));
		std::cout<<x<<std::endl;
	}
}

results in:

Code:

xxxxx@MCP ~/Programming/C++ $ g++ test.cpp -Wall -w -ansi -pedantic -o test.exe && ./test.exe
1
1
1
1
1
1
1
1
1
1

**quizkiwi** · 08-31-2006

Well, I understand what you are saying, and I did some tests to confirm it, but is there a solution to having a random number within a certain range? And why does this work perfectly on XP with VS2005? I would be content to leave it, but the homework will be graded on a Unix Server....bummer

**quizkiwi** · 08-31-2006

I fixed this by implementing what the Unix Man pages states never to do and that is use low-order bits...

Code:

random_integer = 1 + (rand()%50)

This gives a random number between 1 and 50. Since it is only for testing my code and is not actually my code, I guess I will let this slide. If anyone has a better solution, I would appreciate hearing about it.

**SlyMaelstrom** · 08-31-2006

http://www.eternallyconfuzzled.com/articles/rand.html

Our favorite female moderator explains a simple uniform random number algorithm.

EDIT: Owned
|
|
V

**Mario F.** · 08-31-2006

There is a better solution. Read Prelude's page on using rand().

http://eternallyconfuzzled.com/articles/rand.html

Thread: Random Numbers Unix

Thread Tools

Search Thread

Display

Random Numbers Unix

Similar Threads

questions....so many questions about random numbers....

Doubts regarding random numbers generation

random numbers

random numbers limit

Generate random numbers in Lucky7 project using C#