[devel] [PATCH 0/3] optimize rpmsetcmp()

[Kirill A. Shutemov]

On Fri, Nov 26, 2010 at 12:04:23AM +0200, Kirill A. Shutsemov wrote:
> From: Kirill A. Shutemov <kirill на shutemov.name>
> 
> Tested on Intel Core2 Duo P9500, 3GiB RAM. i586.

apt-shell < /dev/null

rpm-4.0.4-alt100.4:

  Performance counter stats for 'apt-shell' (5 runs):

        7431.705536  task-clock-msecs         #      1.000 CPUs    ( +-   0.037% )
                 39  context-switches         #      0.000 M/sec   ( +-  14.774% )
                 13  CPU-migrations           #      0.000 M/sec   ( +-  27.252% )
              12135  page-faults              #      0.002 M/sec   ( +-   0.003% )
        18654610617  cycles                   #   2510.139 M/sec   ( +-   0.034% )  (scaled from 33.34%)
        17117123518  instructions             #      0.918 IPC     ( +-   0.014% )  (scaled from 50.00%)
         3151838786  branches                 #    424.107 M/sec   ( +-   0.039% )  (scaled from 50.00%)
          558860303  branch-misses            #     17.731 %       ( +-   0.063% )  (scaled from 50.00%)
           52927655  cache-references         #      7.122 M/sec   ( +-   0.316% )  (scaled from 33.34%)
             930111  cache-misses             #      0.125 M/sec   ( +-   1.324% )  (scaled from 33.34%)

        7.434342478  seconds time elapsed   ( +-   0.049% )

rpm-4.0.4-alt100.4 + patches:

  Performance counter stats for 'apt-shell' (5 runs):

        4364.794613  task-clock-msecs         #      1.000 CPUs    ( +-   0.066% )
                 33  context-switches         #      0.000 M/sec   ( +-  20.124% )
                 12  CPU-migrations           #      0.000 M/sec   ( +-   6.780% )
              11681  page-faults              #      0.003 M/sec   ( +-   0.005% )
        10954135791  cycles                   #   2509.657 M/sec   ( +-   0.063% )  (scaled from 33.32%)
        11905100223  instructions             #      1.087 IPC     ( +-   0.020% )  (scaled from 49.99%)
         2181001479  branches                 #    499.680 M/sec   ( +-   0.035% )  (scaled from 49.99%)
          200182757  branch-misses            #      9.178 %       ( +-   0.104% )  (scaled from 50.01%)
           31301034  cache-references         #      7.171 M/sec   ( +-   0.377% )  (scaled from 33.36%)
             940816  cache-misses             #      0.216 M/sec   ( +-   2.534% )  (scaled from 33.34%)

        4.366582986  seconds time elapsed   ( +-   0.067% )

Around 41% of speed up.

-- 
 Kirill A. Shutemov