Re: Random Number generator

Jean:

Jesus Christ! If you are making a game (is this
`comp.games.development.design' or is this
`mil.communications.digital.cryptography' ?) just use rand().
If you are not making a game then I suggest you enroll in university for a
graduate mathematics degree, because you'll need one to understand why we
can't make `truely random numbers'. Or you can just take us at face value
when we say `one can not, with a PC, in any convenient manner, make truely
random numbers' and use rand().

I am not w/o experience in this matter, as I had to have some rudimentary
understanding of pseudorandom number generation when I had to deal with
cryptography while in the Air Force back in the early 90s, but I had a
head-start before that from programming

ANSI has established a standard that is both pretty fair and fairly
pretty; via con rand();.

</RANT>

In article <3aGpb.6294$Bv6.1933688@news1.epix.net>,
Timothy J. Bruce <uniblab@hotmail.com> wrote:
>If you are making a game (is this `comp.games.development.design' or
>is this `mil.communications.digital.cryptography' ?) just use
>rand().

This may not be the best choice if programming under Windows, as
the rand() implemented under MS Visual C++ 6.x/.NET (and probably
other versions) has RAND_MAX defined to 32767. Thus, you're only going
to get 15 usable bits out of it, tops. [rand() under most Unix
libc-versions has 31 usable bits, which is far more usable.]

The Mersenne Twister (free BSD-license source, see a site like
http://www.math.keio.ac.jp/~matumoto/emt.html ) has a period of
2^19937-1, and claims that it is sometimes faster than rand(). It's
not industrial/crypto-grade, but it's not MS's Fisher-Price toy
either.

Nathan Mates
--
<*> Nathan Mates - personal webpage http://www.visi.com/~nathan/
# Programmer at Pandemic Studios -- http://www.pandemicstudios.com/
# NOT speaking for Pandemic Studios. "Care not what the neighbors
# think. What are the facts, and to how many decimal places?" -R.A. Heinlein

Nate:

Thanks for the tip!

</RANT>

In article <3fa73b16$0$41291$a1866201@newsreader.visi.com>, nathan@visi.com (Nathan Mates) wrote:
>In article <3aGpb.6294$Bv6.1933688@news1.epix.net>,
>Timothy J. Bruce <uniblab@hotmail.com> wrote:
>>If you are making a game (is this `comp.games.development.design' or
>>is this `mil.communications.digital.cryptography' ?) just use
>>rand().
>
> This may not be the best choice if programming under Windows, as
>the rand() implemented under MS Visual C++ 6.x/.NET (and probably
>other versions) has RAND_MAX defined to 32767. Thus, you're only going
>to get 15 usable bits out of it, tops. [rand() under most Unix
>libc-versions has 31 usable bits, which is far more usable.]

What matters is how usable the bits are, rather than how many one
obtains on each iteration. The MSVC rand() has a 32-bit seed, and this
is probably a better criterion (it may well be that Unix versions have a
longer seed, I don't know).

In general, rand() is sufficient for the majority of games, but there
are game certain applications (map-making, generating IDs for online
play, to give two obvious possibilities) in which it would be wise to
look for something better.

Gerry Quinn
--
http://bindweed.com
Screensavers and Games for Windows
Download free trial versions
New arcade-puzzler just out - "Volcano"

Re: Random Number generator - Java-Games