Changes between Version 1 and Version 2 of Ticket #1923, comment 16


Ignore:
Timestamp:
May 14, 2013, 7:10:39 PM (11 years ago)
Author:
tuan kuranes

Legend:

Unmodified
Added
Removed
Modified
  • Ticket #1923, comment 16

    v1 v2  
    99On the "time" performance, they're not representative at all on pathfinding patch, as in order to keep "save" compatibility, serialisation does some vector to map/ptr_map translation and deserialisation each time (which happens a lot during replay apparently), which surely gives wrong timing there. (the entity map was used for find and loop, whereas a vector for loop and an entityid indexed vector gives ).  Testable with a non-tested hack with pathfind_test, just for perf, which gives 10% on that replay
    1010
    11 I'll look for a workaround for both problems (hash and serialisation overhead breaking benchmarks (perhas related)).
     11I'll look for a workaround for both problems (hash and serialisation overhead breaking benchmarks (perhaps related)).