[pet] PET: managing resource limits

[Bernd Kiefer]

Hi all,

that is my understanding of Stephan's proposal, too. I'd be happy to
implement it, in fact, i've already produced code that did at least
some of the generalization, but not the multi-stage approach we're
discussing now. It should not be too complicated to add that, though.

So if there's agreement, i'd proceed with what i've started and put it
into the `ng' branch. Could all involved please give me feedback as soon
as possible, to avoid extra work on my side?

Thanks,
		Bernd

On 29.08.2011 20:01, Yi Zhang wrote:
>>
>> ah, sorry, i had maybe oversimplified for exposition: i was assuming the
>> sub-allotment ratio would have to be configurable, e.g. minimally through a
>> command line option or setting that specified a percentage of resources; in
>> principle, one could imagine different percentages for different types of limits
>> (10% of the time but only 5% of passive edges); not sure how complex we
>> need to make things in the beginning, though?
>>
> i see. one percentage distribution over various steps sounds good enough to me for beginning. let me check if i understood correctly:
>
> say a limit of 60s is set for each sentence, and we assume a distribution of 10%, 80%, 10% for the three major steps: (1)-(5), (6), and (7). then the parser should abort the sentence if (5) is not yet complete at the 6s mark. otherwise, the parser continues until the 54s mark and check if (6) is complete. if not, it will abort further forest creation and start unpacking immediately. and whatever number results achieved by 60s mark are given back to the user. this means that the parser actually have at least 80% of the allowed time for forest creation, and at least 10% for unpacking. in case the parser saved up time for completing the previous steps, the saving is loaned to the following step?
>
> yi

-- 
----------------------------------------------------------------------
Bernd Kiefer     DFKI GmbH,  Stuhlsatzenhausweg,  D-66123 Saarbruecken
kiefer at dfki.de   +49-681/85775-5301 (phone)   +49-681/85775-5338 (fax)
----------------------------------------------------------------------
Deutsches Forschungszentrum fuer Kuenstliche Intelligenz GmbH
Trippstadter Strasse 122, D-67663 Kaiserslautern, Germany
Geschaeftsfuehrung: Prof. Dr. Dr. h.c. mult. Wolfgang Wahlster (Vor-
                     sitzender), Dr. Walter Olthoff
Vorsitzender des Aufsichtsrats: Prof. Dr. h.c. Hans A. Aukes
Amtsgericht Kaiserslautern, HRB 2313