Re: [AD] stretch_blit bug

[ Thread Index | Date Index | More lists.liballeg.org/allegro-developers Archives ]


Oops, I didn't mean to include all those comments... also I did more
benchmarking and it turns out it's only faster in 24 bit. That's my
desktop color depth but I don't think that's the reason it's faster.
These are the numbers I get at 32 bit.

mm stretch_blit took 283 millis
mm my_stretch took 422 millis
mv stretch_blit took 3764 millis
mv my_stretch took 3855 millis
vm stretch_blit took 309 millis
vm my_stretch took 426 millis

On Wed, 2007-04-25 at 17:30 -0600, Trent Gamblin wrote:
> OK, I've managed to get it as fast if not a little faster than the
> Allegro version. These are the results after compiling with
> -mtune=pentium -O2 -funroll-loops -ffast-math -fomit-frame-pointer.
> 
> mm stretch_blit took 658 millis
> mm my_stretch took 652 millis
> mv stretch_blit took 5724 millis
> mv my_stretch took 5688 millis
> vm stretch_blit took 682 millis
> vm my_stretch took 655 millis
> 
> It is also a lot simpler. If this would be committed to svn I could add
> masked and 8 bit support. New version attached.






Mail converted by MHonArc 2.6.19+ http://listengine.tuxfamily.org/