The question posed by mostyfriedman was why not just loop from 0 to 1 million so that your are looping the same number of times and not having same odd construction the for loop?

The value of i never plays a part in your computation within your for loop other then the number of times the computation is done.

The code you have in the original post has i ranging from 3 million to 4 million meaning that you are looping 1 million times. You could have also made i range from 5 million to 6 million, or 7.5 million to 8.5 million and still get the same number of iterations.

Elkanah modi, on 23 August 2012 - 05:46 AM, said:

If you change the loop as you describe to go from 1 to 4 million, then now you are changing the number of iteration. Now you'll have 4 million iterations where as before you were only iterating 1 million times.

I don't understand your question, can you please clarify?

also why aren't you just iterating a 10^6-1 from 1 to 10^6, what's the reason for starting from 3 million and ending at 4 million

Allocating an array of that size on the stack will almost certainly exceed your platform's stack size limit. You should allocate it on the heap instead.

Note that on most (all?) platforms exceeding the allowed heap size won't cause an error until you actually try to access the parts of the memory that exceed the stack boundary. That's why you don't get an error for small inputs.

PS: What does this have to do with the Fibonacci series?

Rather than keeping a max, wouldn't it be easier to count down in the for loops and break when you find a palindrome?