[BLKMGK]

Sounds like perhaps some ATI users could help out by drop shipping him a new video card

[Harry Muscle]

Unfortunately I can't until the 19th, because I won't have access to ATI cards before then.

Is it possible for you to make the change in the code mentioned by elupus and all of us can do the testing for you? This doesn't happen on all ATI cards, so access to an ATI card might not allow you to reproduce the issue anyway ... just an idea

Thanks,
Harry

[elupus]

Well since most of you compile yourself.. it's a quite easy sourcecode change.

file: xbmc\cores\VideoRenderers\VideoShaders\YUVRGBShader.cpp
CHANGE

string shaderv =
"void main()"
"{"
"gl_TexCoord[0] = gl_TextureMatrix[0] * gl_MultiTexCoord0;"
"gl_TexCoord[1] = gl_TextureMatrix[1] * gl_MultiTexCoord1;"
"gl_TexCoord[2] = gl_TextureMatrix[2] * gl_MultiTexCoord2;"
"gl_TexCoord[3] = gl_TextureMatrix[3] * gl_MultiTexCoord3;"
"gl_Position = ftransform();"
"}";

TO

string shaderv =
"void main()"
"{"
"gl_TexCoord[0] = gl_TextureMatrix[2] * gl_MultiTexCoord2;"
"gl_TexCoord[1] = gl_TextureMatrix[0] * gl_MultiTexCoord0;"
"gl_TexCoord[2] = gl_TextureMatrix[1] * gl_MultiTexCoord1;"
"gl_TexCoord[3] = gl_TextureMatrix[3] * gl_MultiTexCoord3;"
"gl_Position = ftransform();"
"}";

Notice how i've swapped around the indexes of gl_TextureMatrix and gl_MultiTexCoord.

Don't ask me how they managed to be so stupid .

Another fix is to comment out these 3 lines:

if (!glewIsSupported("GL_ARB_texture_non_power_of_two") && glewIsSupported("GL_ARB_texture_rectangle"))
{
m_textureTarget = GL_TEXTURE_RECTANGLE_ARB;
}

in xbmc\cores\VideoRenderers\LinuxRendererGL.cpp

[darkham]

well done! can i've my name in a svn trunk ?

[Wolf on Air]

Well since most of you compile yourself.. it's a quite easy sourcecode change.

file: xbmc\cores\VideoRenderers\VideoShaders\YUVRGBShader.cpp

Making a qualified guess you meant YUV2RGBShader.cpp, anyway, this doesn't help. With both patches applied:

Code:

02:15:01 T:3070683040 M:  8482816  NOTICE: GL: ARB shaders support detected
02:15:01 T:3070683040 M:  8462336  NOTICE: GL: Selecting Single Pass ARB YUV2RGB shader
02:15:01 T:3070683040 M:  8462336  NOTICE: GL: No vertex shader, fixed pipeline in use
02:15:01 T:3070683040 M:  7540736  NOTICE: GL: OpenGL version 1.3 detected
02:15:01 T:3070683040 M:  7540736  NOTICE: GL: NPOT textures are supported through GL_ARB_texture_rectangle extension

Screen still green as all hell. In fact, it looks suspiciously like trying to feed RGB output to a YUV renderer, black is neon green, white is neon purple. I'm not sure where to go from there though, shaders/fragment programs are mostly opaque to me.

Obviously, the patch is only for the GLSL path, and I'm using ARB.
Rv350 Radeon 9600 Mobility, using the Xorg Radeon (R300) DRI driver
Fglrx is not an option for me due to not supporting composite sync output (RGB-SCART VGA cable) and (properly) custom modelines, and my S-video output hardware is rather broken - it jitters horizontally in a most annoying fashion, and doesn't give even close to RGB quality anyway.

Another fix is to comment out these 3 lines:

in xbmc\cores\VideoRenderers\LinuxRendererGL.cpp

I'm not even sure what this is supposed to be doing, and it doesn't change anything noticeably regardless.

In the meantime, I forced RENDER_SW in the relevant spot, and CPU usage still looks unremarkable, and I get usable video output. Still, fixing the hardware acceleration would be a priority.

Thanks for trying, I hope this helps solve this.

[darkham]

a poll: how many peoples are in this situation, with ati graphics?

[elupus]

My fixes wasn't for green screen. It was for the screen shot a few posts earlier. And both fixes at the same time are wrong. The second fix, will break the first fix .

[shagworthy_uk]

Hi, I can confirm that the fix outlined by elupus has resolved my issue on ATI Radeon xpress 1250 using fglrx 8.4 & Hardy. Thank you so much!

[eversmann]

Hi guys!

Being following the software from the xbox versión, now on my linux i'm very excited with it. Congrats fo everyone in the proyect. you're superb guys!! Keep it up!!! ;-)

Have the same issue, Ati Radeon 9800pro, latest drivers, Ubuntu hardy, same green screen/flurred image.

I'll wait for this fix to hit the ppa repository, i'll compile by myself if it takes too long, but wanted to say that, looks like most of the ati users have it.

regards.

[darkham]

It works, yeah finally it works!!! Developpers, what do you wait? this issue isn't properly "minor", most of ati people (enough, i think...) can't watch video correctly, what remains otherwise? Great work, elupus!

[darkham]

Hi guys!

Being following the software from the xbox versión, now on my linux i'm very excited with it. Congrats fo everyone in the proyect. you're superb guys!! Keep it up!!! ;-)

Have the same issue, Ati Radeon 9800pro, latest drivers, Ubuntu hardy, same green screen/flurred image.

I'll wait for this fix to hit the ppa repository, i'll compile by myself if it takes too long, but wanted to say that, looks like most of the ati users have it.

regards.

I've an athlon 3000+ i686,9600pro,ati catalyst 8.5 (8.493.1), hardy heron, compiling an svn trunk needed 40 min, eversmann enjoy, you don't regret!

[Wolf on Air]

My fixes wasn't for green screen. It was for the screen shot a few posts earlier. And both fixes at the same time are wrong. The second fix, will break the first fix .

That does explain some things , I misparsed somewhere and correlated the two issues.

Anyway, as I said, forcing RENDER_SW mode gets me perfect video, so something seems to be wrong in YUV2RGBProgressiveShaderARB and/or its interaction with the R300 driver, I'm not qualified to say which.

Here's an interesting article on how they implemented Xv-to-OpenGL in the very same driver, including YUV-RGB conversion (apparently doable as a native texture format, no shaders required? shouldn't this be easier then?)

[iggy23]

that is awesome. I had almost given up on my current hardware and XBMC, but now video is playing sweet. A bit fuzzy (compared to the CoreAVC on my Windows box) but at least it plays!

Is this fix going to go into the final code, or will it mess up things for Nvidia people?

[elupus]

What remains is to know when to enable the bugfix. It's a utterly silly fix that shouldn't be needed, and it will break gfxs implementing stuff correctly. Thus it's probably be a setting in advancesettings.xml.

[darkham]

What remains is to know when to enable the bugfix. It's a utterly silly fix that shouldn't be needed, and it will break gfxs implementing stuff correctly. Thus it's probably be a setting in advancesettings.xml.

in my post ("this issue isn't properly "minor", most of ati people (enough, i think...) can't watch video correctly, what remains otherwise?") this was referred to ati people, who can do not too much otherwise with XBMC , with this issue (only music for all time annoying a little... ) thanx for explanation elupus, your work have made me, and other ati people i hope, very happy