Sebastian Jaszczur added a comment - 2016-04-13 16:15 I won't create a doc; comment here, along with attachment (runnigtimechart.png) should be enough. I've run (manually) 9 different solutions for different problems on all problems' tests. It looks promising - timelimits will probably stay the same (see the line "y = x"). But I've encountered one solution (for problem Kurierzy) when szkopuł showed much larger running times than oitimetool - but relationship is still linear. Probably in this problem cache is more important (and oitimetool assumes infinite cache). It's worth noticing that even if we'd apply y=x scaling to Kurierzy problem - we'd effectively make tests easier. That means that still every golden solution will pass - and we'd have problems only with sub-optimal solutions getting too many points. But that's relatively small problem and it could be even acceptable to leave scaling as y=x. We'll create a script for running one golden solution from every public problem just to check how many anomalies are out there. But it seems like linear scaling will be OK (or, really - no scaling at all). But that's another ticket.

Piotr Wiśniewski added a comment - 2016-04-13 17:17 +1