Types in C++ for large numbers

Question

I want to calculate the number of inversions for a very big array, something like 200,000 ints, and the number I get is quite big. So big it can't be stored in an int value.

The answer I get is something like -8,353,514,212, while for simple cases it works, so I think that the problem is the type of the variable I use to store the number of inversions.

I also tried with long int and the output is the same, but if I try with double 4.0755e+009 is the output. I don't know what the problem is.

Possible duplicate of http://stackoverflow.com/questions/117429/inputting-large-numbers-in-c — miqbal, Jul 07 '12 at 12:10
`unsigned long long` will take you far enough to store the number of inversions in arrays of `6*10^9` elements or so. — Daniel Fischer, Jul 07 '12 at 12:12

Otto Allmendinger · Answer 1 · 2012-07-07T12:36:49.887

3

use an unsigned data type
use unsigned long (usually 2^32-1) or unsigned long long (usually 2^64-1)

For full reference see this article.

edited Jul 07 '12 at 12:36

answered Jul 07 '12 at 12:21

Otto Allmendinger

27,448
7
68
79

The size of these `unsigned long` and `unsigned long long` are platform specific. It would be good to mention this in you answer. – Kiril Kirov Jul 07 '12 at 12:27
1

@KirilKirov Yes, but the given numbers are the smallest allowed maximal values. If the types are actually larger, nothing's lost. – Daniel Fischer Jul 07 '12 at 12:36
@DanielFischer - it's true, but still, this could be misleading for someone. – Kiril Kirov Jul 07 '12 at 12:38

score 0 · Answer 2 · answered Jul 07 '12 at 12:53

If the native types of the compiler aren't fit to hold the result of your computation, you could consider using a bignum library.

A quick search revealed these two:

http://www.ttmath.org/

http://gmplib.org/

I've no experience with either, but gmp seems to be the more popular choice around SO, so maybe that's what you should try first

Types in C++ for large numbers

2 Answers2