Factors Inhibiting Reproducability Of Simulation Results

[Chris Gordon-Smith]

I just managed to get my artificial chemistry simulation producing the
same results when I run it twice with the same parameters. Perhaps some
members of the group will be interested in the two factors that led to
differing results:

1) My graphics library (GTKMM2) calling rand() at random times between
the calls made by my code
2) My code storing pointers in an std::set, and then iterating over the
set. The ordering is dependent on the memory locations of the dynamically
allocated objects to which the pointers point.

Chris Gordon-Smith
www.simsoup.info

 

[Jorgen Grahn]

I just managed to get my artificial chemistry simulation producing the
same results when I run it twice with the same parameters. Perhaps some
members of the group will be interested in the two factors that led to
differing results:

1) My graphics library (GTKMM2) calling rand() at random times between
the calls made by my code

Aargh! But in a sense that's just another example of the brokenness
of rand(). You usually want a PRNG where you keep the state manually.
I assume the TR1 random number stuff can do that.

/Jorgen

 

[Miles Bader]

Aargh! But in a sense that's just another example of the brokenness
of rand(). You usually want a PRNG where you keep the state manually.
I assume the TR1 random number stuff can do that.

The random state in both TR1 and C++11 is a part of the generator
object, so ... yeah. [Of course it would be super duper surprising if
this wasn't true!]

-Miles