Skip to main content
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
Re: [sumo-user] Different results on different machines
Hi Karl,
we are aware of platform-specific behavior (windows / linux) in regards to floating point operations. Another way to get differences is by using slightly different versions of third party libraries on different platforms.  The proj library that handles coordinate conversions behaves slightly different between versions and this can affect the simulation. Also, there are differences in generated random numbers between python2 and python3 which might play a role when running traci scripts.

regards,
Jakob

Am Mo., 23. März 2020 um 13:17 Uhr schrieb Schrab, Karl <karl.schrab@xxxxxxxxxxxxxxxxxxx>:

Hi Jakob,

 

Thanks! Anyhow, I doubt this is the main problem, as I add all vehicles via TraCI.  Also, I always get the same results on the same machine, but different results on a second machine, so I would exclude the RNG as a culprit.

I’m not that deep into C++ and compiler stuff, but could precision / rounding / comparison of doubles play a role and result in different behavior on different machines?

 

Kind regards,

Karl

 

 

Von: sumo-user-bounces@xxxxxxxxxxx <sumo-user-bounces@xxxxxxxxxxx> Im Auftrag von Jakob Erdmann
Gesendet: Montag, 23. März 2020 12:05
An: Sumo project User discussions <sumo-user@xxxxxxxxxxx>
Betreff: Re: [sumo-user] Different results on different machines

 

All RNGs should be initialized now (https://github.com/eclipse/sumo/issues/6830) - (downloads not updated until tomorrow).

The most likely culprit in your case was the RNG that determined insertion of flows with attribute 'probability'.

 

regards,

Jakob

 

Am Mo., 23. März 2020 um 11:11 Uhr schrieb Richter Gerald <Gerald.Richter@xxxxxxxxx>:

Hi,

 

a.f.a.i.k. not ALL random number generators (i think there are 4 in SUMO)
are initialized deterministically by the random seed.

 

regards,
Gerald

Von: sumo-user-bounces@xxxxxxxxxxx <sumo-user-bounces@xxxxxxxxxxx> im Auftrag von Schrab, Karl <karl.schrab@xxxxxxxxxxxxxxxxxxx>
Gesendet: Montag, 23. März 2020 10:58:11
An: Sumo project User discussions
Betreff: [sumo-user] Different results on different machines

 

Hello,

 

We are currently experiencing a very weird issue related to counting vehicles within lane area detectors (E2). On different machines we receive slightly different results in the counts of vehicles (+/-1 vehicle). Both machines uses the same version of SUMO and the same random seed. Has anyone of you experienced such behavior in the past, can give me an explanation, or has any other idea?

 

Kind regards,

Karl

_______________________________________________
sumo-user mailing list
sumo-user@xxxxxxxxxxx
To unsubscribe from this list, visit https://www.eclipse.org/mailman/listinfo/sumo-user

_______________________________________________
sumo-user mailing list
sumo-user@xxxxxxxxxxx
To unsubscribe from this list, visit https://www.eclipse.org/mailman/listinfo/sumo-user

Back to the top