C help

Hello. I've recently taken a minor interest in random number generators, and have been studying the Mersenne Twister algorithm. (http://www.math.sci.hiroshima-u.ac.jp/~m-mat/MT/emt.html)

The first of the seeding functions uses an unsigned long to seed the array, but it does a curious thing. Here is the code:

    void init_genrand(unsigned long s)
    {
        mt[0]= s & 0xffffffffUL;

As far as I can tell this is an identity operation: i.e. for all values of s, s & 0xffffffff == s

In other words it does nothing.

Later on in the function he does it again, but this time with a comment that appears to explain it:

    mt[mti] &= 0xffffffffUL;
    /* for >32 bit machines */

Can anyone suggest why the author put this in his code? I can't see how a 64 bit machine would affect the calculation at all, but I don't know much C which probably explains my ignorance.

Here's the whole function along with the (ew) global variables it uses:

    #define N 624
    static unsigned long mt[N]; /* the array for the state vector  */
    static int mti = N+1; /* mti == N+1 means mt is not initialized */

    /* initializes mt with a seed */
    void init_genrand(unsigned long s)
    {
        mt[0] = s & 0xffffffffUL;
        for (mti = 1; mti < N; mti++)
        {
            mt[mti] =
            (1812433253UL * (mt[mti-1] ^ (mt[mti-1] >> 30)) + mti);
            /* See Knuth TAOCP Vol2. 3rd Ed. P.106 for multiplier.
             * In the previous versions, MSBs of the seed affect
             * only MSBs of the array mt.
             * 002/01/09 modified by Makoto Matsumoto */
             mt[mti] &= 0xffffffffUL;
             /* for >32 bit machines */
        }
    }

It's bitmasking. The coder obviously doesn't want more than 32-bits. Why? Probably due to some analysis of the algorithm assuming 32-bits.

Why is DES limited to 56-bits? Due to the algorithm.

Note the "is a transformation which assures equidistribution of the generated numbers in 623 dimensions" in the Wikipedia. This only applies for 32-bit numbers.

Why in the world would you want to do this in C?

...you must be stupid. This problem is almost made for C.

Lots of bitwise operations? Check. Need code transparency? Check. Need speed? Check. Needs to be readable to all programmers/incorporated in libraries? Check.

>>2
I know it's a mask, but I thought unsigned longs were limited to 32 bits anyway. amirong? Guess I must be.

>>3 like >>4 said.

Personally I try to avoid C for most tasks, but this is clearly a job for C. In fact the reason I'm studying it is because I want to implement it in Scheme. I know it'll be slow as hell and the code will be ugly, but what can I say, I'm a masochist.

No. Be very careful when making assumptions about the lengths of types. There are ways to ensure that a type is 32, 64, or whatever bits, but the usual types of int, long, and even char, depend on the machine architecture.

OLD SKOOL BUMP JUSTICE

>>1
HAHA DISREGARD I SUCK COCKS

>>8
Dictionary attack?

>>9
Your virginity is showing.

>>10
He used the same tripcode as OP.

>>10
STOP SEEING HIS POSTS YOU STUPID FUCKER

>>12
Settle down, FrozenVoid's screaming companion.

>>1
Oh wow I remember this tripfag. Always thought his name was awesome.

awesome

>>14
No, his name was Christy McJesus.

There are ways to ensure that a type is 32, 64, or whatever bits
HOW?

>>17
man stdint.h

>>18
THATS c99 FAGGOT

>>19
__int32

>>17
int mt:32;

>>17
autoconf

>>21
Bit fields considered harmful

>>20-23
Same peson

>>23
Bit fields considered helpful

HAY EVERYBODY I USED A GOTO YESTERDAY. I AM SO COOL