Re: [glob2-devel] To Erik: My Profiling Information

[Erik Søe Sørensen]

Kieran P skrev:

> Hey Erik,
>
> I saw your changes in the source code. Nice that you are submitting it 
> now :D

You mean instead of posting patches?
Well, now that I have repository access I might as well do something 
serious...
Running valgrind today paid off, I think - one significant (i.e. 
memory-corrupting) bug and one memory leak (0.4MB leaked in a relatively 
short run) found & fixed :-)

> I ran another test (cloned, compiled, run, won, quit, and run the two 
> codes (one with -q and one without).
>
> Results: http://dl.sv.nongnu.org/releases/glob2/profiling.tar.bz2
>
> I'll do this every day or two when you make commits if you want.
>
> I have another Linux virtualisation (Fedora), which I will also run 
> these tests on at some stage.

Well, thank you for the effort, but I'm not sure these virtualized 
timings say much. They won't correspond to running the game on a 
machine's 'real' OS... the virtualization software influences the 
measurements.

And in your timings, the three most dominant (by far) functions are:

29.70% GAGCore::DrawableSurface::drawSurface(int, int, 
GAGCore::DrawableSurface*, int, int, int, int, unsigned char)
17.85% GAGCore::DrawableSurface::drawFilledRect(int, int, int, int, 
GAGCore::Color const&)
15.60% GAGCore::Sprite::getW(int)

- i.e. two drawing operations, which I doubt we could  do anything to 
optimize, and this mysterious getW, which may be an artefact of the 
virtualization (I have no clue why else it shows up so prominently in 
your data)...

/Erik