E17 benchmarks

[Olivier Fourdan]

On Thu, 2005-06-09 at 14:34 +0200, samuel verstraete wrote:
> ok i went a little bit fast through the options... i agree :) i was	
> talking about the "Windows snapping" thing... that is not exactly the
> same off course... but I thought it would be the same... but it seems
> like you can't turn it off completely... 

Windows snapping has nothing to do with mapping, it will just speed up
the moves.

> Look, i'm not saying this *is* important... i'm just thinking it's
> interesting... if you don't agree with me that's ok :) i don't mind...
> i think it tests a very limited thing of the window manager and it has
> *nothing* to do with user experience... (I just like benches :p )

Ok, so here come some results. It clearly shows that the bottleneck is
from the smart placement code, as the "map_throughput" ix approx. 10
times faster without that part. We can remove the smart placement and
map window in the center of screen (which seems surprisingly usable and
'natural somehow to get the newly created windows centered at first).

E17
Test:        map_response
             MIN: 0.006172s, MAX: 0.014244, AVG: 0.008894
Test:        map_throughput
             WIN/SEC 33.868340

xfwm4 - 4.3.0 without smart placement
Test:        map_response
             MIN: 0.017219s, MAX: 0.183254, AVG: 0.048186
Test:        map_throughput
             WIN/SEC 31.628390

metacity - 2.8.3
Test:        map_response
             MIN: 0.008220s, MAX: 0.089859, AVG: 0.046456
Test:        map_throughput
             WIN/SEC 25.596825

kwin - 3.0
Test:        map_response
             MIN: 0.017280s, MAX: 0.076186, AVG: 0.042384
Test:        map_throughput
             WIN/SEC 19.295727

xfwm4 - 4.3.0 with smart placement (current code)
Test:        map_response
             MIN: 0.026628s, MAX: 1.144981, AVG: 0.102753
Test:        map_throughput
             WIN/SEC 3.146941

Cheers,
Olivier.