quality of rand() in VC++

Anyone know what generator rand() in VC++6 (CRT) uses?

Is it good quality or not?  Is a RAND_MAX of 32767 far too low (as it would seem)?

A good reference on the subject would be welcome (specific to VC++ rand() please - I have a copy of TAOCP in front of me, enough random number generator theory for my lifetime :-)
Who is Participating?
jonninConnect With a Mentor Commented:
yea, it wants 32 bit numbers, but the tests are broken into chunks of bits, so it will partially work (the conclusion will be wrong, as the leading zeros will mess it up, but you can still figure it out and get results; the conclusion is just a summary of the chunk tests).
just stick rand into a 32 bit int to do this...

but I guess this is still invalid as it repeats at least every 32787 numbers, and dh wants a million, ~3 repeats in the set...

also, diehard was written for unix, so the high/low bit reports are backwards!

I would say that if you have to ask about the randomness of rand, you should not be using it.  A custom one is the way to go on important things. I think 15 bits is way too low, m$ probably did this because nt could orig run on several platforms and they picked lowest common hardware for some algorithms...  it crops up from time to time, whenever you see a 16 bit number used when a 32 should have been used...

Most of the random number generators shipped with compilers are not that good.  I was looking at a post on rec.games.programmer and one programmer noted that a test of VC++ 5 (I know you said you were using 6) gave a repeating cycle after several thousand numbers.  The poster noted that the source for rand was on the CD, so maybe they do that with version 6 also.  Anyway, two ways to limit the possibility of cycling:
Reseed after a set number of calls to the function.
Use a custom random number generator.

As for 32767 being too low, it really depends on what you are trying to do.  I am working with Linux, and gcc has RAND_MAX set to 2147483647.

Hope this helps,

Its not good, but better than many compilers' generators. Look up a battery of tests called diehard and test it (needs about 1 million numbers from the generator). Does statistics and predicibility tests on the numbers, bits, and more.

I have a home coded one for vc that passes all the tests, but its *very* slow, many computations / number...

if you want to try it, I can post it...

Free Tool: SSL Checker

Scans your site and returns information about your SSL implementation and certificate. Helpful for debugging and validating your SSL configuration.

One of a set of tools we are providing to everyone as a way of saying thank you for being a part of the community.

IainHereAuthor Commented:
I saw diehard, and thought that it needed 32 bit numbers - rand() is only 15 bit, isn't it?
CryptoAPI also supports generating a random number (rather an array of bytes). The function is CryptGenRandom. If interested I can post the code to generate the random sequence of bytes.
i;ve been working on some genetic algorythm stuff that requires random mutations - and i;ve been using rand() succesfully. it;s not great but seems to do the job. must remember to seed the rand generator - i include time.h and use srand(time(NULL))

Question has a verified solution.

Are you are experiencing a similar issue? Get a personalized answer when you ask a related question.

Have a better answer? Share it in a comment.

All Courses

From novice to tech pro — start learning today.