what type to use when long double is not enough?

Question

Hey,

I'm writing a program that assigns prime numbers to each entry of a matrix and then I'll need to multiply some of them. The resulting number grows rapidly and I'm at a loss as to which type to use, as I'm getting "wrap-around" with long double :S

All the help is appreciated.

-Pickel

Answer 1

You may have to implement your own big integer type. Check out:

BigInt

Answer 2

If it's just integers, there's long long int (at least in C/C++). If we're talking about doubles... Use a BigDecimal class.

Answer 3

Assuming you are using "long double" as a type that it's a C or C++ assignment?

How big are your numbers?

http://stackoverflow.com/questions/1582529/maximum-value-vs-size-of-long-and-double-in-net

discusses some related stuff but may be beyond the scope of your assignment

Generally speaking you need an arbitrary precision library:

http://en.wikipedia.org/wiki/Arbitrary-precision_arithmetic

But chances are that for purposes of an assignment you are not required to make things so complicated as to require the use of an APL

Answer 4

if it's an integer, use the BigInteger class (in Java or .NET)

if it's a floating point, use BigDecimal (only in java, .net still does not have a arbitrary precision floating number

if you're in C/C++, you have to create your own type

Answer 5

Unless you're required to implement your own arbitrary precision type, use GMP. You'll want the mpz_t (integer) type. It's pretty well documented, and there are tutorials and StackOverflow questions you can look at.

Answer 6

If you refuse to use a library, and don't want to invent your own big types, why don't you keep track of extra factors of 2, or something like that.

while ( mybignum > BIGNUM_THRESH )
{
    twos++;
    mybignum /= 2; // use >>=1 if you use an integer type (you said you used double so therefore the /=)
}

then print out your answers as mybignum * 2**twos

Take a 64 bit int for twos and you're safe up to 2^2^64