Small numbercial analysis question

  • Thread starter Thread starter DrKareem
  • Start date Start date
  • Tags Tags
    Analysis
AI Thread Summary
The discussion revolves around challenges faced when performing numerical analysis with large numbers in C++. Initially, the user encountered issues with integer overflow, leading to negative values when calculations exceeded the integer range. Switching to the 'long' data type resolved some issues, but limitations persisted with larger numbers. Suggestions for solutions included using 'long long' or '__int64' for 64-bit integers, and considering Java's BigInteger class for handling extremely large integers. Additionally, participants recommended exploring C++ numerics libraries that provide support for larger integer types and suggested alternatives like 'double' or 'long double' for non-integer calculations. Resources for libraries that support big integers were shared, along with a mention of ubasic, which can handle very large integers and rational numbers. The conversation emphasizes the importance of selecting appropriate data types and libraries for managing large numerical computations effectively.
DrKareem
Messages
101
Reaction score
1
I was doing a small numbercal analysis problem that required the use of huge numbers. At first i was surprised that my code didn't work, even after double and triple checking, and after tracing the programme, i found out the mistake that i did. Firstly i used integers, and when the numbers started to increase they went bigger than the integer number range on C++ and started to give a negative number. Anyways, i used long and it worked ok for most of the numbers that needed calculations, but still some iteration couldn't be done since they also went out of range.

How could i solve this problem?
 
Computer science news on Phys.org
Wow, well in the case of integers it is a simpler problem. Extending floating point datatypes is more complicated.

It is not so hard to create a larger integer class as long as you follow the same bitwise rules than intrinsic integer types follow. And the basic arithmetic operators can be overriden in C++ to make it pretty transparent.

But to save you some trouble, there are many C++ numerics libraries out there that have already done this. Try starting with "c++ numerics library" in google.

I can't speak for any of them because I haven't used them before.
 
Thx, will do.

Still open to more suggestions though :)
 
Are 64 bits enough for you? If so then use long long (with g++) or __int64 (with VS); or their unsigned counterparts. Otherwise a quick solution might be to switch to Java and use the library class BigInteger.
 
Last edited by a moderator:
Last edited:
I'm thankful. It hasn't occurred to me to use Java, which i will be doing, and for educational purposed i'll also check out and test the links grady and robphy gave.
 
If you don't have to work in Unix, get ubasic (free download) from
http://www.simtel.net/category.php?id=299

It comes "ready to go" with built-in capability to handle integers and rational numbers up to 2600 DIGITS long, plus other impressive features.
 
Last edited by a moderator:
Back
Top