D1253: dirstate: avoid reading the map when possible (issue5713) (issue5717)

[lothiraldan (Boris Feld)]

lothiraldan added a comment.


  Here are some measures https://phab.mercurial-scm.org/rHG0865d25e8a8ad664e8da31ab1bf12f1fc9bf0c7a:
  
    $ python -m perf command /home/lothiraldan/project/mercurial/mercurial/hg id -r .
    .....................
    command: Mean +- std dev: 109 ms +- 3 ms
    $ python -m perf command /home/lothiraldan/project/mercurial/mercurial/hg id -r .
    .....................
    command: Mean +- std dev: 110 ms +- 2 ms
    $ python -m perf command /home/lothiraldan/project/mercurial/mercurial/hg id -r .
    .....................
    command: Mean +- std dev: 111 ms +- 3 ms
  
  And here are the mesaures on https://phab.mercurial-scm.org/rHGc36c3fa7d35b0aa1c29d316927b21f77ecc809b8:
  
    $ python -m perf command /home/lothiraldan/project/mercurial/mercurial/hg id -r .
    .....................
    command: Mean +- std dev: 166 ms +- 5 ms
    $ python -m perf command /home/lothiraldan/project/mercurial/mercurial/hg id -r .
    .....................
    command: Mean +- std dev: 165 ms +- 3 ms
    $ python -m perf command /home/lothiraldan/project/mercurial/mercurial/hg id -r .
    .....................
    command: Mean +- std dev: 166 ms +- 5 ms
  
  So it's seems that this patch solves most of the performance regression, but with so many changesets since https://phab.mercurial-scm.org/rHGc36c3fa7d35b0aa1c29d316927b21f77ecc809b8 I'm not sure we can do better.

REPOSITORY
  rHG Mercurial

REVISION DETAIL
  https://phab.mercurial-scm.org/D1253

To: durham, #hg-reviewers, lothiraldan
Cc: yuja, lothiraldan, mharbison72, mercurial-devel