[Cbc] Convergence problem with miplib2010/30n20b8

[Yann COLLETTE]

On miplib2010/30n20b8.mps.gz, the thing I noticed is that master version 
of cbc is not able to compute "best solution". The value stays at 1e50.
And with version 2.8.9, the "best solution" is quickly set at 653.

As far as I understand, this "best solution" is used to cut part of the 
tree ?

Le 27/06/2017 à 16:36, Yann COLLETTE a écrit :
> Thanks.
> I will add the Easy / Medium / Hard tags to this test set and just 
> restart the Easy set.
>
> Le 27/06/2017 à 16:35, Stefan Vigerske a écrit :
>> https://urldefense.proofpoint.com/v2/url?u=http-3A__miplib.zib.de_CHANGELOG&d=DwIGaQ&c=Ngd-ta5yRYsqeUsEDgxhcqsYYY1Xs5ogLxWPA_2Wlc4&r=kY-tTJAFuHScYZHjojaaKVZZOEYVVkPpGiWUPr3h80U&m=ZLboDToYNeQTxBmzszcr2h0nVC4tWHbVwCBwZsdHDTg&s=JSR7d6xnCbNL6q-IeVs-PenpwLQQzYhtyRXa_RXcmEI&e= 
>> The instances in MIPLIB do not change.
>>
>> The newer Cbc just has less luck on this instance than the older 
>> version.
>>
>> On 06/27/2017 04:32 PM, Yann COLLETTE wrote:
>>> The other problem is that the miplib test set has been updated since 
>>> the last time I performed the test.
>>> It was miplib2010-1.1.2 and now, it's 1.1.3.
>>> I am not able to find what was updated.
>>>
>>>
>>>
>>
>>
>
> _______________________________________________
> Cbc mailing list
> Cbc at list.coin-or.org
> https://urldefense.proofpoint.com/v2/url?u=https-3A__list.coin-2Dor.org_mailman_listinfo_cbc&d=DwIGaQ&c=Ngd-ta5yRYsqeUsEDgxhcqsYYY1Xs5ogLxWPA_2Wlc4&r=kY-tTJAFuHScYZHjojaaKVZZOEYVVkPpGiWUPr3h80U&m=ZLboDToYNeQTxBmzszcr2h0nVC4tWHbVwCBwZsdHDTg&s=7oO1SNoy-Vv-7ICZSOjJlabKPFjv0IcLsx6vMlLwdTU&e=