4
\$\begingroup\$

Usually fastest code challenges have the challenge author time the results on his/her machine.

For old questions for which the challenge author may not be active any more (ex. those by user Lembik), how can the timings be compared fairly? Is there a community accepted reference machine that can be used? (Related: Is TIO acceptable for fastest-code questions?)

\$\endgroup\$
5
  • 2
    \$\begingroup\$ The short answer to the title is "they simply cannot be judged, in general", since a fastest-code usually contains the spec of the machine the submissions will be run on. But someone determined enough could always build an identical machine and judge submissions on it, so the OP being unavailable does not make it off topic by itself. \$\endgroup\$
    – Bubbler
    May 10 at 4:36
  • 1
    \$\begingroup\$ That said, it might be helpful to have an explicit policy that allows user B to take over A's challenge and score submissions on a machine available to B. (Would it make the scoring criterion non-objective?) \$\endgroup\$
    – Bubbler
    May 10 at 4:37
  • 1
    \$\begingroup\$ I imagine a system where if the original author is gone, a mod or well-established user can take over timing on their system. \$\endgroup\$
    – qwr
    May 10 at 4:45
  • 3
    \$\begingroup\$ I think the performance would generally scale, assuming everyone is running on a standard desktop CPU. So an answer that takes 1 minute won't suddenly beat a 1 second answer. Though maybe stuff like cores and memory size can make a big difference in timings. \$\endgroup\$
    – qwr
    May 10 at 4:46
  • 1
    \$\begingroup\$ It's more complicated than that, especially when people start doing micro-optimizations via threads, vector instructions (different CPUs have different optimal number of threads and different vector instruction support), and probably even GPU and external storage for huge lookup tables. \$\endgroup\$
    – Bubbler
    May 11 at 2:22

You must log in to answer this question.

Browse other questions tagged .