Thread overview
[phobos] Issue6244, implementing 'powmod'
October 02
Hello,

I am trying to implement the 'powmod' functionality as described here [1].

The issue that I am having is that the algorithm uses multiplications which can cause overflow.
If the base is 32 bits, then I can use 64 bit variables to handle the result of multiplications, however the problem arises if the base is 64 bit.

The pseudocode would look the following:

while (exponent > 0)
{
    if (exponent & 1)
    {
        result = mulmod(result, base, modulus);
    }

    base = mulmod(base, base, modulus);
    exponent >>= 1;
}

return result;

The problem that I am facing is with the 'mulmod' function which should do multiplication and modulo of the result without overflow problems.

Do you think that it would be a good idea to limit the base to 32 bits?
Or does D have any facility similar to 'mulmod'?

Thanks,
Alex

[1] - https://en.wikipedia.org/wiki/Modular_exponentiation#Right-to-left_binary_method
_______________________________________________
phobos mailing list
phobos@puremagic.com
http://lists.puremagic.com/mailman/listinfo/phobos


October 02
A possible solution is to implement mul128 using primitives for 32- and 64-bit numbers. A possible signature would be:

void mul128(ulong a, ulong b, out ulong lo, out ulong hi);

The technique is to do the multiplication "by hand" as if you had two numbers having two digits each. You obtain a 4-digit number.

Consider the two-digit decimal numbers ab and cd, for example 43 and 65. Then the multiplication "by hand" is:

(10a + b) * (10c + d) = 100ac + 10(ad + bc) + bd

Since a, b, c, d are single digits it follows we only need 1-digit multiplication and addition with carry. Now of course our base is not 10 but 2^^32, so we multiply two numbers, each having two 32-bit "digits" and we get 128 bits worth of result.

mul128() would be a generally useful function to put in druntime. I think someone on the general forum has implemented it already, you may want to ask.


Andrei

On 10/2/17 12:48 PM, Alex Jercaianu via phobos wrote:
> Hello,
> 
> I am trying to implement the 'powmod' functionality as described here [1].
> 
> The issue that I am having is that the algorithm uses multiplications which can cause overflow.
> If the base is 32 bits, then I can use 64 bit variables to handle the result of multiplications, however the problem arises if the base is 64 bit.
> 
> The pseudocode would look the following:
> 
> while (exponent > 0)
> {
>      if (exponent & 1)
>      {
>          result = mulmod(result, base, modulus);
>      }
> 
>      base = mulmod(base, base, modulus);
>      exponent >>= 1;
> }
> 
> return result;
> 
> The problem that I am facing is with the 'mulmod' function which should do multiplication and modulo of the result without overflow problems.
> 
> Do you think that it would be a good idea to limit the base to 32 bits?
> Or does D have any facility similar to 'mulmod'?
> 
> Thanks,
> Alex
> 
> [1] - https://en.wikipedia.org/wiki/Modular_exponentiation#Right-to-left_binary_method 
> 
> _______________________________________________
> phobos mailing list
> phobos@puremagic.com
> http://lists.puremagic.com/mailman/listinfo/phobos
_______________________________________________
phobos mailing list
phobos@puremagic.com
http://lists.puremagic.com/mailman/listinfo/phobos
October 03
On Tuesday, 3 October 2017 at 01:42:44 UTC, Andrei Alexandrescu wrote:
> A possible solution is to implement mul128 using primitives for 32- and 64-bit numbers. A possible signature would be:
>
> void mul128(ulong a, ulong b, out ulong lo, out ulong hi);
>
> The technique is to do the multiplication "by hand" as if you had two numbers having two digits each. You obtain a 4-digit number.
>
> Consider the two-digit decimal numbers ab and cd, for example 43 and 65. Then the multiplication "by hand" is:
>
> (10a + b) * (10c + d) = 100ac + 10(ad + bc) + bd
>
> Since a, b, c, d are single digits it follows we only need 1-digit multiplication and addition with carry. Now of course our base is not 10 but 2^^32, so we multiply two numbers, each having two 32-bit "digits" and we get 128 bits worth of result.
>
> mul128() would be a generally useful function to put in druntime. I think someone on the general forum has implemented it already, you may want to ask.
>
>
> Andrei

Thanks for the suggestion!
I think I have also found a way to implement 'mulmod' without' mul128'.
Is overflow well defined for unsigned types in D?
For example, (MAX_UINT + 1) equals 0 for unsigned types?

I will also come back with a 'mul128' implementation, since there is none available for the moment in the standard library.

Thanks,
Alex
_______________________________________________
phobos mailing list
phobos@puremagic.com
http://lists.puremagic.com/mailman/listinfo/phobos