[darcs-users] benchmarking darcs-1

[Petr Rockai]

Eric Kow <kowey at darcs.net> writes:
> ==============  =============   ===============  ===============
>                   old-f 1.0.9       darcs-2.3.1   darcs-2.3.99.2
> ==============  =============   ===============  ===============
>     get (full)          51.2s      231.9s d=3.5     231.8s d=7.9
>     get (lazy)              -       38.0s d=0.8      35.6s d=0.5
>       pull 100         116.8s      139.2s d=1.4     115.0s d=1.1
>             wh         1206ms     343.2ms d=0.0   1790.8ms d=0.0
>         wh mod         1208ms     342.7ms d=0.0   1783.2ms d=0.0
>          wh -l         7920ms   22976.3ms d=0.2   4335.8ms d=0.0
>     record mod         2670ms   57570.6ms d=0.1  33440.5ms d=0.2
>     revert mod         2348ms   28263.7ms d=0.2   6829.0ms d=0.1
> (un)revert mod         4920ms   91301.9ms d=0.4  24173.4ms d=0.2
>          check         189.0s     994.4s d=89.3     592.9s d=9.1
>         repair         166.7s      940.3s d=2.8    584.6s d=14.8
>       annotate              -                 -                -
>      pull 1000         301.9s      258.0s d=2.3     228.0s d=4.6
> ==============  =============    ===============  ===============
This looks fairly reasonable now. Nathan, if I may bother you further,
would it be possible to also run 2.3.99.2 on a hashed repository where
you ran darcs(-2.3.99.2) optimize --pristine? I think that should cut
down the times further, and bring the whatsnew times back into line with
2.3.1. Thanks a lot!

Yours,
   Petr.

PS: At least the record performance is a victim of issue1106. This is
something that will be (hopefully) addressed in darcs 2.5.