[Bug 14038] Max Payne 2: bullet time screen filled with solid color

wine-bugs at winehq.org wine-bugs at winehq.org
Tue Jul 29 12:27:10 CDT 2008


http://bugs.winehq.org/show_bug.cgi?id=14038





--- Comment #65 from Tobias Jakobi <liquid.acid at gmx.net>  2008-07-29 12:27:09 ---
(In reply to comment #64)
> I can reproduce this problem on my card (bullet time filled with solid color)
> with git by commenting out the line with GL_ARB_texture_non_power_of_two in
> wined3d/directx.c or doing 
> 
> gl_info->supported[ARB_TEXTURE_NON_POWER_OF_TWO] = FALSE;
> 
> Also lots of these messages appear:
> 
> fixme:d3d:transform_texture Non-power2 texture being used with generated
> texture coords
> 
> Not sure if it means anything :/. Considering that it's not supposedly fixed on
> geforce5 that doesn't have this extension, this confuses me.
> 

Hmm, wasn't there an issue with non_power_of_two textures on the FX? I think it
was not explicitly advertised in the GL extensions string, but the OpenGL
version reported by the driver indicated that the extension should be supported
(since it's core in this version).
However the FX only supports the texture_rectangle with good performance.

Does this seem correct? At least that's what I read from Stefan Dösinger in
comment #28.

-------------------------------------------------

Like already explained in bug #14686 I could test this issue today on a Radeon
card (thanks to Jan again!):

Graphics card: ATI Mobility Radeon HD2600
Drivers: ati-drivers-8.512 (=8.7, =8.51.3) (fglrx)

Test was done with latest wine git master (snapshot from today): last commit =
6a21ab270b2a9dc04ad38714f410e1ab27872f02

ORM is fbo, GLSL is disabled for the application.

Ingame settings:
PPE=medium
pixel shader skins=off

I tested all resolutions available (640x480, 800x600, 1024x768, 1280x800) and
all of them work (no crop issue), BUT:
(i) Performance is slower than backbuffer
(ii) Console is flooded with FIXMEs (*)

This doesn't change when switching PPE to high. Performance still is worse and
FIXMEs are still there (and the same).

(*):
fixme:d3d_surface:read_from_framebuffer_texture >>>>>>>>>>>>>>>>>
GL_INVALID_ENUM (0x500) from glReadBuffer @ surface.c / 920

Since we didn't have that much time I only made one performance comparison.

ORM=fbo, PPE=high, res=1024x768:
normal mode: around 19 fps
bullettime mode: around 13 fps

OK, that's with fbo mode and is already kinda slow (in bt mode).
Switching to backbuffer:

ORM=backbuffer, PPE=high, res=1024x768:
normal mode: around 21 fps
bullettime mode: around 21 fps

As you can see switching to bt mode with backbuffer ORM doesn't affect the
performance in any way. Consider now that fbo should deliver an improvement to
the performance.

Test was done from nearly the same ingame position. And even if it was not
exactly the same: it's the difference between the fps values.

In fbo mode the FB processing increases frame rendering time from 52.6 to 76.9
ms. That's a plus of 24.3 ms, a lot if you ask me.

Greets,
Tobias


-- 
Configure bugmail: http://bugs.winehq.org/userprefs.cgi?tab=email
Do not reply to this email, post in Bugzilla using the
above URL to reply.
------- You are receiving this mail because: -------
You are watching all bug changes.


More information about the wine-bugs mailing list