Re: [AD] SF.net SVN: alleg:[11840] allegro/branches/4.9

[Thomas Fjellstrom <tfjellstrom@xxxxxxxxxx>]

On Sat March 14 2009, Peter Wang wrote:
> On 2009-03-14, Thomas Fjellstrom <tfjellstrom@xxxxxxxxxx> wrote:
> > On Sat March 14 2009, Thomas Fjellstrom wrote:
> > > Looks like something here has helped ex_prim in software mode too. Used
> > > to run at 4fps on my Q6600 w 4G ram and a NV 8800GTS. It now runs at
> > > 12fps with blending. Without blending its back up to 60fps, same as
> > > hardware \o/
> > >
> > > I call that a win. Before I couldn't tell what speed it was at without
> > > blending as it would blank the screen of most things... I think if its
> > > possible to work on blending some more, that would rock.
> > >
> > > fwiw, I get "Time = 2.88472 s, FPS = 103.996" from ex_blend_bench.
>
> I think it should be higher. I get "Time = 0.931021 s, FPS = 322.227"
> on a Core2Duo 1.83 GHz, in release mode.

Didn't see that bit before I hit reply.. oops. I'm running in full debug mode. I just retried in RelWithDebInfo, and I get:

Time = 0.788397 s, FPS = 380.519

so much better.

ex_prim is also better, with a total of ~37fps in software mode. somewhat more reasonable. ex_membmp sees a jump in fps as well, 14fps vs 2fps.

> I am running in 32-bit mode.
> I assume you are in 64-bit mode, and the float->int trick might not have
> been enabled. Can you check that by forcing the inline function in
> aintern_float.h?
>
> My results for ex_prim are much worse; ~3.7 FPS in software mode with
> blending.
>
> > Should mention I occasionally get:
> > Assert failed at line 399 of
> > /home/moose/.projects/allegro.git/src/bitmap_new.c
> > Aborted
> >
> > from ex_prim when I switch which primitives its drawing with the arrow
> > keys.
>
> Is that this line in your revision?
>
> ASSERT(height >= 0);
>
> I guess depending on the rotation of the shape and the thickness, etc.
> it's trying to draw outside the bounds. Paul, can you take a look?
>
> Peter
>

--
Thomas Fjellstrom
tfjellstrom@xxxxxxxxxx