Re: [gtk-list] Benchmarking glib (atleast GList) (was: how can I trust glib when it has so many memleaks?)

[Damon Chaplin <damon karuna freeserve co uk>]

Steven Rostedt wrote:


> These were my results:
> 
> glibtest:
>   utime = 4:530000
>   stime = 0:160000
>   shared mem size = 0
>   unshared data size = 0
>   unshared stack size = 0
>   page reclaims = 746
>   page faults = 378
>   swaps = 0
>   block input = 0
>   block output = 0
> 
> noglibtest:
>   utime = 1:290000
>   stime = 1:180000
>   shared mem size = 0
>   unshared data size = 0
>   unshared stack size = 0
>   page reclaims = 30910
>   page faults = 67
>   swaps = 0
>   block input = 0
>   block output = 0

You cheated! Your own g_list_append() function really does a g_list_prepend(),
which is much faster.

If I use the real g_list_prepend() the results look quite different:

glibtest: 
utime = 1:950000
stime = 0:40000
shared mem size = 0
unshared data size = 0
unshared stack size = 0
page reclaims = 730
page faults = 175
swaps = 0
block input = 0
block output = 0

noglibtest:
utime = 3:90000
stime = 2:240000
shared mem size = 0
unshared data size = 0
unshared stack size = 0
page reclaims = 30687
page faults = 64
swaps = 0
block input = 0
block output = 0


So GMemChunk stays ;-)

Damon