[ODE] Re: Deterministic simulation on multiple clients

[George Birbilis]

it could be some bug with your code reading data from wrong locations 
(usually then you'd see major differences)
since the DLLs are loaded in different place everytime (although [at least 
in Delphi] you can pass info to the compiler on where to load the DLL 
[offset relative to the virual base address of the .EXE] - that is a hint to 
the OS of course since many DLLs may be specifying overlapping offsets)


>I have now tested to run the same simulation on one computer. This gives 
>the same result every time, until I restart the computer. Then it starts to 
>give a different result. Altough I haven't looked at it in detail, it looks 
>like the divergence starts happening after the first collision (as Victor 
>mentioned in the other thread). Now the question is what may change after a 
>computer is restarted? Perhaps some RNG is seeded with the start time of 
>the machine. Perhaps it has something to do with when DLL-files are loaded 
>(do they stay in memory between program runs?).
>