[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
Re: [p2-dev] MISC 2010 News

For those who wonder what paranoid means, it is the name of one of the optimisation function that was defined by the organisers of the competition.
The competition also defined a second optimisation function called trendy. Here is the definition of those two functions as provided by the organisers.


Optimization critera:
---------------------

Two fixed optimization criterias will be used in this version of the
competition: they are both lexicographic combinations of four simple
integer valued utility functions of a solution, which we describe
below.

Let's call U the universe of a CUDF, which contains the description of
all available packages, and the field Installed: which is set to 1 for
packages installed on the user machine, and 0 otherwise.

Let's call S a solution to the user request computed by your solver.

We write V(X,name) the set of versions in which name (the name of a
package) is installed in X, where X may be U or S. That set may be
empty (name is not installed), contain one element (name is installed
in exactely that version), or even contain multiple elements in case a
package is installed in multiple versions.

We write

- #removed(U,S): is the number of packages REMOVED in the proposed
  solution S w.r.t. the original installation U :

  #removed(U,S) = #{name | V(U,name) nonempty and V(S,name) empty}

- #new(U,S): is the number of NEW packages in the proposed solution S  
w.r.t. the original installation U :

  #new(U,S) = #{name| V(U,name) empty and V(S,name) nonempy}

- #changed(U,S): is the number of packages with a MODIFIED (set of)  
version(s) in the proposed solution S w.r.t. the original installation U :

  #changed(U,S) = #{name | V(U,name) different to V(S,name) }

- #notuptodate(U,S): is the number of installed packages but NOT in the
  latest available version :

  #notuptodate(U,S) = #{name| V(S,name) nonempty and does not contain  
the
                              most recent version of name in S}

The two optimization criteria are

 PARANOID: we want to answer the user request, minimizing the number
           of packages removed in the solution, and also the packages
           changed by the solution; the optimization criterion is

                lex( min #removed, min #changed)

           Hence, two solutions S1 and S2 will be compared as follows:

           i) compute ri = #removed(U,Si), ci = #changed(U,Si)

           ii) S1 is better than S2 iff r1 < r2 or (r1=r2 and c1<c2)


TRENDY: we want to answer the user request, minimizing the number of
         packages removed in the solution, minimising the number of
         outdated packages in the solution, and minimizing the
         number of extra packages installed; the optimization
         criterion is

                lex( min #removed, min #notuptodate, min #new)

           Hence, two solutions S1 and S2 will be compared as follows:

           i) compute ri = #removed(U,Si), ui = #notuptodate(U,Si),
              ni = #new(U,Si)

           ii) S1 is better than S2 iff
               r1 < r2 or (r1=r2 and (u1<u2 or (u1=u2 and n1<n2)))






Inactive hide details for Daniel Le Berre ---09/01/2010 03:55:47 AM---preliminary raw results are available here:Daniel Le Berre ---09/01/2010 03:55:47 AM---preliminary raw results are available here:


From:

Daniel Le Berre <leberre@xxxxxxxxxxxxxxxxxxx>

To:

P2 developer discussions <p2-dev@xxxxxxxxxxx>

Date:

09/01/2010 03:55 AM

Subject:

Re: [p2-dev] MISC 2010 News




preliminary raw results are available here:
http://www.mancoosi.org/~zack/misc2010/results/

The archives contains the logs of the solvers.
More user friendly results should be available by the end of next week.

Daniel

Daniel Le Berre a écrit :
> Not for the moment.
>
> We are supposed to dig the results tomorrow morning, so I guess they will be available by then.
>
> I will let you know ASA they are available.
>
> Daniel
>
> Le 7 janv. 2010 à 21:13, Henrik Lindberg a écrit :
>
>> Cool !
>> Are the results available somewhere?  The links does not point to any result afaikt.
>>
>> Regards
>> Henrik Lindberg
>> henrik.lindberg@xxxxxxxxxxxxxx
>>
>>
>>
>> On Jan 7, 2010, at 7:22 PM, Daniel Le Berre wrote:
>>
>>> Dear all,
>>>
>>> I am currently attending a Mancoosi [1] meeting in Nice.
>>> The result of the internal competition of resolvers (MISC) [2] have just been disclosed.
>>>
>>> The p2 based resolver we submitted with Pascal got first place in one of the track, the paranoid one.
>>>
>>> The interesting point here is that we submitted exactly the configuration that is currently shipping with Eclipse 3.6 milestones.
>>>
>>> Congrats to all p2 contributors !
>>>
>>> Daniel
>>>
>>> [1]
http://www.mancoosi.org/
>>> [2]
http://www.mancoosi.org/~zack/misc2010/_______________________________________________
>>> p2-dev mailing list
>>> p2-dev@xxxxxxxxxxx
>>>
https://dev.eclipse.org/mailman/listinfo/p2-dev
>> _______________________________________________
>> p2-dev mailing list
>> p2-dev@xxxxxxxxxxx
>>
https://dev.eclipse.org/mailman/listinfo/p2-dev
>>
>
> _______________________________________________
> p2-dev mailing list
> p2-dev@xxxxxxxxxxx
>
https://dev.eclipse.org/mailman/listinfo/p2-dev
>
>


--
            Daniel Le Berre
mailto:leberre@xxxxxxxxxxxxxxxxxxx
            MCF,    CRIL-CNRS UMR 8188,    Universite d'Artois
           
http://www.cril.univ-artois.fr/~leberre
_______________________________________________
p2-dev mailing list
p2-dev@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/p2-dev


GIF image

GIF image