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

["Eric Botcazou" <ebotcazou@xxxxxxxxxx>]

> DGA2 ./misc/cmplog.pl noaccel.log accel.log

You swapped the files :-)

> DRAW_MODE_SOLID results:
>
>         putpixel()                       = 98%
>         hline()                          = 41%
>         vline()                          = 82%
>         line()                           = 13%
>         rectfill()                       = 414%
>         circle()                         = 21%
>         circlefill()                     = 17%
>         ellipse()                        = 25%
>         ellipsefill()                    = 15%
>         arc()                            = 17%
>         triangle()                       = 14%
>
>
> Accel:   hline() - 188961, textout() - 1116
> Unaccel: hline() - 454380, textout() - 42359
>
> I would say not all that good :(

Yes, nearly a disaster :-(

There seems to be something flawed in the accelerated DGA2 driver: from what
I read in the code, it accelerates the same primitives as Windows
(hline, vline, rectfill, clear, blit_to_self, draw_sprite, masked_blit)
so why the hell does it perform so badly with line(), circle(), ellipse(),
arc() and textout(), given that it reverts to software drawing for them ?

Under Windows, there is parity for these primitives between the
hardware-accelerated and the software driver.

> Was tested at 640x480x16 DGA2 (Un)Accelerated.
> (MMX, 3DNow was enabled)

Thanks for testing.

--
Eric Botcazou
ebotcazou@xxxxxxxxxx