C# BigInteger and int How to save memory?

Question

I am testing a matrix if it has a prime number in every line.

The MR means that it should use the Miller Rabin algorithm. If its false it simply tries the divisors up to sqrt(n). Vansor is true if it has found a prime in every checked row, vanoszlop is true if it has found a prime in the actually checked row.

My question is: Is it possible to save the memory by not creating both the int and BigInteger values only if the tryParse is true? I mean something like

if (int.tryParse(akt, out new int szam))

Is something like this possible? (and how much memory does a BigInteger take when its not signed?)

    akt = Console.ReadLine();
    int szam; BigInteger szambig;

    if (int.TryParse(akt, out szam))
    {
       if (MR)  {
          if (MilRab(szam))
          { vansor = true; vanoszlop = true; } }

          else if (Prim(szam))
          { vansor = true; vanoszlop = true; }
    }
    else if (BigInteger.TryParse(akt, out szambig))
    {
       if (MR) {
          if (MilRab(szam))
          { vansor = true; vanoszlop = true; } }

       else if (Prim(szam))
       { vansor = true; vanoszlop = true; }
    }

Better ask if would be sensible. The answer would be No. You're talking about 2 local variables with a very small footprint. The bulk of BigInt is only allocated when it is assigned. — H H, Nov 10 '11 at 16:51
Why are you trying to save memory so much? Did you somehow find out that this part of the code is consuming too much memory? — svick, Nov 10 '11 at 16:51
Shouldn't the bottom `MilRab()` and `Prim()` use the `szambig` variable? — C.Evenhuis, Nov 10 '11 at 16:59
This is like someone with two billion dollars in the bank spending an hour haggling over a four-dollar difference in a million-dollar purchase. *You already have two billion bytes of address space available, and you've already spent a million on the stack.* Now you're worried about how to deal with *four* of those bytes efficiently. You almost certainly have far more important things to worry about. — Eric Lippert, Nov 10 '11 at 17:03
@Eric Lippert - Unfortunately, thats how most rich people act... — C.Evenhuis, Nov 10 '11 at 17:07

score 0 · Accepted Answer · answered Nov 10 '11 at 16:57

I'm not 100% sure on how .Net IL optimizes memory, but in general local "value" type (as opposed to "reference" types) are kept on the stack. IE, you would only save the ~4 bytes that an integer takes up by not instantiating it, and only for the lifetime of that one call. Once you exit the function, the stack is cleared.

Structures are "value" types, and are also placed on the stack. They reserve memory for all other value types & reference pointers as needed. The size of a BigInteger is the same whether or not "Signed" is true or false.

I suppose my real question is: why the memory obsession? With the code example you have, you'll take a few dozen bytes of memory that will all be freed up when the method exits.

score 0 · Answer 2 · answered Nov 10 '11 at 17:03

If you managed to do it, at best you would save 16 bytes that BigInteger takes up on 64-bit mode (I guess).

I'm pretty sure that's not worth worrying about. In situations where those 16 bytes are important:

you probably shouldn't be using .Net
you most likely don't have .Net available anyway

C# BigInteger and int How to save memory?

2 Answers2

Linked