yade-users team mailing list archive

[Bruno Chareyre <bruno.chareyre@xxxxxxxxxxx>]

Jiao: there is nothing to fix, indeterminism is not bad. :)

Cross-linking is always a good idea Jerome. It would be nice to add the
links. OTOH, indeterminism is a topic in itself, not strictly related to
multithreading.
What people sometimes don't realize, is that the one-thread result is
not better than multithread.
j>1 reveals the indeterminism, j=1 hides it, but it is still
indeterminate in both cases.

Reordering interactions in a one thread simulation would give different
results, and I don't think Yade is magicaly choosing a "best" ordering.
Consequently, j=1 should be seen like running a random number generator
with always the same seed, while j>1 initializes the seed with the
computer's clock (i.e. "randomly").

Bruno


On 10/05/12 11:28, Jérôme Duriez wrote:
> Hello Liqing
>
> - please think to https://answers.launchpad.net/yade/+addquestion to
> ask question ;-)
>
> - second, give a look to
> https://yade-dem.org/doc/formulation.html#result-indeterminism
> concerning your problem.
>
> This was several times discussed on the lists, but I understand you
> were not aware of this, I had myself difficulties to find this link.
> Perhaps would it be particularly useful to put this link in
> https://www.yade-dem.org/doc/prog.html#parallel-execution or
> https://www.yade-dem.org/doc/user.html#controlling-parallel-compuation
> for example ?
>
> Jerome
>
> Le 10/05/2012 11:08, Liqing Jiao a écrit :
>> Dear all,
>>
>> When the same script runs on the same version of yade, but one is
>> single core, another one is multi-cores, the results are different. I
>> use yade trunk version, which is recommended by Anton. It's good. But
>> when I run the same script, the result is different. You can try the
>> concrete uniax.py, which is in examples folder, you will see the
>> different lines printed out on the screen. And I tried my model both
>> using parallel and non parallel running, the results are also
>> different. Thank you for your attention. Could you please reply me if
>> you know how to fix it? Thank you in advance.
>>
>> Cheers, Liqing
>>
>>
>>
>>
>> _______________________________________________
>> Mailing list: https://launchpad.net/~yade-users
>> Post to     : yade-users@xxxxxxxxxxxxxxxxxxx
>> Unsubscribe : https://launchpad.net/~yade-users
>> More help   : https://help.launchpad.net/ListHelp
>


-- 
_______________
Bruno Chareyre
Associate Professor
ENSE³ - Grenoble INP
11, rue des Mathématiques
BP 46
38402 St Martin d'Hères, France
Tél : +33 4 56 52 86 21
Fax : +33 4 76 82 70 43
________________