[darcs-devel] [issue1731] performance regression in check/repair (2.4.x)

[Petr Ročkai]

Petr Ročkai <me at mornfall.net> added the comment:

=== darcs ===   
                
                || darcs-2.3        | darcs
================++==================+=============
     get (full) ||      2.3s  11.0M |  2.3s  10.0M
get (lazy, x10) ||      2.5s   3.0M |  2.8s   3.0M
       pull 100 ||      1.6s  25.0M |  1.5s  27.0M
       annotate ||     13.7s 181.0M | 14.9s 182.0M
         wh x50 ||      0.4s   0.0M |  1.1s   0.0M
     wh mod x50 ||      2.8s   2.0M |  3.4s   2.0M
      wh -l x20 ||      1.8s   5.0M |  0.9s   2.0M
          check ||      9.0s  27.0M | 11.6s  27.0M
         repair ||      9.0s  26.0M | 11.6s  27.0M
      pull 1000 ||      9.0s  40.0M |  8.2s  37.0M

=== ghc-hashed ===
                
                || darcs-2.3        |  darcs
================++==================+==============
     get (full) ||     10.5s  71.0M |  10.1s  71.0M
get (lazy, x10) ||      9.5s   7.0M |  10.3s   8.0M
       pull 100 ||      7.0s  36.0M |   6.4s  65.0M
       annotate ||         -      - |      -      -
         wh x50 ||      0.8s   0.0M |   2.4s   0.0M
     wh mod x50 ||      3.8s   3.0M |   5.3s   3.0M
      wh -l x20 ||      4.8s   9.0M |   2.2s   3.0M
          check ||    121.5s 152.0M | 121.8s 169.0M
         repair ||    121.2s 152.0M | 121.3s 177.0M
      pull 1000 ||     27.4s 201.0M |  25.2s 197.0M

Please note that check/repair in HEAD do more work than in 2.3 (that is, 
they also check the index which was not done in previous versions). So I 
think this is more than reasonable fix for the speed regression.

Now the question stands what about whatsnew... (but that's for another 
bug, unless people observe that hashed-storage HEAD is substantially 
slower on whatsnew than 0.4.6, that is).

__________________________________
Darcs bug tracker <bugs at darcs.net>
<http://bugs.darcs.net/issue1731>
__________________________________