Re: [AD] About the speed of flat polygon primitives

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


> $ ./misc/cmplog.pl tests/noaccel.log tests/accel.log
>   (note the order)
>         before -> after
>         putpixel()                       = 96% -> 95%
>         hline()                          = 41% -> 42%
>         vline()                          = 69% -> 78%
>         line()                           = 16% -> 16%
>         rectfill()                       = 273% -> 271%
>         circle()                         = 21% -> 21%
>         circlefill()                     = 22% -> 51%
>         ellipse()                        = 28% -> 28%
>         ellipsefill()                    = 20% -> 49%
>         arc()                            = 20% -> 21%
>         triangle()                       = 18% -> 46%
>
> Improvements for circlefill, ellipsefill and triangle,

Yes, roughly double the speed. I'm going to commit the patch.
But the performances are still cut off by 50% compared with the software
driver.

> but otherwise no effect.

It wasn't meant to have any effect on other primitives.

The remaining gap is purely an internal problem of the DGA2 driver: an extra
layer of synchronization is added by the accel driver even if it reverts to
software drawing for line(), circle(), circlefill(), ellipse(),
ellipsefill(), arc() and triangle().

--
Eric Botcazou
ebotcazou@xxxxxxxxxx



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