[vox-tech] Re: random number in C

[Matt Holland]

Jeff Newmiller wrote:

> The random driver may be used to obtain a seed for random() or rand(), or
> it could be used to obtain random numbers on its own.  You run a risk of
> extracting all of the randomness from the system if you depend on random
> for long random sequences, though, and a /dev/random that has run out of
> entropy is a file that will block until that condition changes.
> 
> Having known psuedo-random sequences does allow you to "repeat"
> monte-carlo simulations.  To answer Peter, yes, rand() is supposed to be
> deterministic for this very reason.  I am puzzled by the same issue Ken is
> puzzled by, though... failing to call srand() should always cause rand()
> to yield the same pseudo-random sequence... it should never have yielded
> variable results in the first place!


I don't think there's any mystery here.  If you compare the two sets of 
runs (late in the day vs. early in the day), you will see that the early 
runs have "trials: 1000000", whereas the later runs have "trials: 1000". 
  I'm assuming that "trials" is some parameter that probably influences 
the number of random numbers generated within a simulation, so you 
wouldn't expect the same answer even with the same seed, since a 
parameter has been changed.

Matt

> 
> 
>>On Tuesday 02 April 2002 06:26 pm, you wrote:
>>
>>>>Date: Tue, 2 Apr 2002 14:47:20 -0800
>>>>[CODE SNIPPED]
>>>>
>>>So this is when the program is seeded to /dev/random:
>>>
>>>>   p@satan% ./ising2-jfunc 4.0
>>>>   T: 4.0e+00  beta: 2.500e-01  trials: 1000  N: 10  M: 10  seed:
>>>>3497451914 Average Energy: -1.273893
>>>>   Average magnetization: 0.425220
>>>>   Average |magnetization|: 0.531100
>>>>   Average magnetic susceptibility: 34.746600
>>>>
>>>And this is when the code is seeded to 1:
>>>
>>>>   T: 4.0e+00  beta: 2.500e-01  trials: 1000  N: 10  M: 10  seed:
>>>>3497451914 Average Energy: -1.326747
>>>>   Average magnetization: 0.495320
>>>>   Average |magnetization|: 0.495320
>>>>   Average magnetic susceptibility: 28.645040
>>>>
>>>And you've established that if you run it repeatedly you keep getting the 
>>>
>>same thing:
>>
>>>>   T: 4.0e+00  beta: 2.500e-01  trials: 1000  N: 10  M: 10  seed:
>>>>3497451914 Average Energy: -1.326747
>>>>   Average magnetization: 0.495320
>>>>   Average |magnetization|: 0.495320
>>>>   Average magnetic susceptibility: 28.645040
>>>>
>>>BUT four hours earlier:
>>>
>>>>Date: Tue, 2 Apr 2002 10:05:34 -0800
>>>>
>>>This is when you seed it to /dev/random:
>>>
>>>>	T: 4.0e+00  beta: 2.500e-01  trials: 1000000
>>>>	N: 10  M: 10  seed: 208006379
>>>>	Average Energy: -1.057143
>>>>	Average magnetization: 0.002008
>>>>	Average |magnetization|: 0.251765
>>>>	Average magnetic susceptibility: 8.948655
>>>>
>>>And this is when you seed it to 1:
>>>
>>>>	T: 4.0e+00  beta: 2.500e-01  trials: 1000000
>>>>	N: 10  M: 10  seed: 208006379
>>>>	Average Energy: -1.050182
>>>>	Average magnetization: 0.030723
>>>>	Average |magnetization|: 0.246092
>>>>	Average magnetic susceptibility: 8.692724
>>>>
>>>Why did 4 hours of time change the results when you seed it to 1 ???
>>>_______________________________________________
>>>vox-tech mailing list
>>>vox-tech@lists.lugod.org
>>>http://lists.lugod.org/mailman/listinfo/vox-tech
>>>
>>_______________________________________________
>>vox-tech mailing list
>>vox-tech@lists.lugod.org
>>http://lists.lugod.org/mailman/listinfo/vox-tech
>>
>>
> 
> ---------------------------------------------------------------------------
> Jeff Newmiller                        The     .....       .....  Go Live...
> DCN:<jdnewmil@dcn.davis.ca.us>        Basics: ##.#.       ##.#.  Live Go...
>                                       Live:   OO#.. Dead: OO#..  Playing
> Research Engineer (Solar/Batteries            O.O#.       #.O#.  with
> /Software/Embedded Controllers)               .OO#.       .OO#.  rocks...2k
> ---------------------------------------------------------------------------
> 
> 
> _______________________________________________
> vox-tech mailing list
> vox-tech@lists.lugod.org
> http://lists.lugod.org/mailman/listinfo/vox-tech
>