+- Forums (https://www.vdrift.net/Forum)
+-- Forum: Project (https://www.vdrift.net/Forum/forumdisplay.php?fid=4)
+--- Forum: Development (https://www.vdrift.net/Forum/forumdisplay.php?fid=9)
+--- Thread: Source file extension changes (cc to cpp) (/showthread.php?tid=718)
- abs1nth - 07-16-2007
modifying the other shaders also makes the shaders compile but it somehow forces vdrift into extremely slow software rendering (although it is pretty ;-)
maybe it is related to apple's glsl implementation, i don't know.
i won't have time to look at it in the next few days ;-(
- joevenzon - 07-17-2007
As far as I know, GLSL is implemented by the graphic card driver, so maybe it's a property of the apple ATI driver...?
- abs1nth - 08-14-2007
abs1nth Wrote:modifying the other shaders also makes the shaders compile but it somehow forces vdrift into extremely slow software rendering (although it is pretty ;-)
something must have changed, it doesn't fall back to software anymore, but it also isn't pretty any more:
![[Image: screenshotjy4.png]](http://img184.imageshack.us/img184/1893/screenshotjy4.png)
btw here is the diff:
Code:
Index: shaders/full/fragment.glsl
===================================================================
--- shaders/full/fragment.glsl (Revision 52)
+++ shaders/full/fragment.glsl (Arbeitskopie)
@@ -10,7 +10,7 @@
//varying vec3 eyecoords;
varying vec3 eyespacenormal;
-uniform vec3 eyelightposition;
+varying vec3 eyelightposition;
varying vec4 ecpos;
varying vec3 viewdir;
Index: shaders/full-nometallic/vertex.glsl
===================================================================
--- shaders/full-nometallic/vertex.glsl (Revision 52)
+++ shaders/full-nometallic/vertex.glsl (Arbeitskopie)
@@ -17,7 +17,7 @@
//set the normal, eyespace normal, and eyespace position
eyespacenormal = normalize(gl_NormalMatrix * gl_Normal);
- normal = (mat3(gl_TextureMatrix[1]) * gl_NormalMatrix) * gl_Normal;
+ normal = (mat3(gl_TextureMatrix[1][0].xyz,gl_TextureMatrix[1][1].xyz,gl_TextureMatrix[1][2].xyz) * gl_NormalMatrix) * gl_Normal;
//normal = vec3(gl_TextureMatrix[1] * (gl_ModelViewProjectionMatrix * gl_Vertex));
//normal = gl_NormalMatrix * gl_Normal;
//normal = gl_Normal;
Index: shaders/full-nospecular/vertex.glsl
===================================================================
--- shaders/full-nospecular/vertex.glsl (Revision 52)
+++ shaders/full-nospecular/vertex.glsl (Arbeitskopie)
@@ -15,6 +15,6 @@
//set the texture coordinates
texcoord_2d = vec2(gl_MultiTexCoord0);
- normal = (mat3(gl_TextureMatrix[1]) * gl_NormalMatrix) * gl_Normal;
+ normal = (mat3(gl_TextureMatrix[1][0].xyz,gl_TextureMatrix[1][1].xyz,gl_TextureMatrix[1][2].xyz) * gl_NormalMatrix) * gl_Normal;
lightposition = normalize(vec3(.1,1.0,.3));
}- joevenzon - 08-14-2007
I don't quite follow, what's the diff for? Do you have the latest SVN code and data? The artifacts look like shadow aliasing, which shouldn't be there because I have generous polygon offsets set when rendering the shadow pass. What kind of 3D card do you have?
- abs1nth - 08-15-2007
joevenzon Wrote:I don't quite follow, what's the diff for? Do you have the latest SVN code and data?yes
the diff is to to get the shaders compiling on mac os x. remember this?
joevenzon Wrote:but I only modified the full shader. If this compiles for you and makes things work, go ahead and copy & paste to the other shaders and check in (or I can).i posted the remaining modifications to get the shaders to compile here.
joevenzon Wrote:The artifacts look like shadow aliasing, which shouldn't be there because I have generous polygon offsets set when rendering the shadow pass. What kind of 3D card do you have?ATI Radeon X1900 XT 512MB
- joevenzon - 08-15-2007
According to some google hits, ATI cards have issues with the glPolygonOffset function. If you tweak the values in line 1240 (increase both numbers) do the artifacts eventually go away?
This is the line to tweak:
Code:
glPolygonOffset(1.0,1.0);Also the full-no* shader files aren't used anymore; I've just removed them from SVN. Also, I'm not sure why you had to change that uniform to varying; it's a uniform! Were you getting any useful error, or just some generic compile error?
- abs1nth - 09-13-2007
joevenzon Wrote:According to some google hits, ATI cards have issues with the glPolygonOffset function. If you tweak the values in line 1240 (increase both numbers) do the artifacts eventually go away?nah that didn't work :-(
This is the line to tweak:
Code:glPolygonOffset(1.0,1.0);
joevenzon Wrote:Also, I'm not sure why you had to change that uniform to varying; it's a uniform! Were you getting any useful error, or just some generic compile error?
Code:
----- Start Shader Log for full -----
INTERNAL ERROR: Symbol 'eyelightposition' is bound as uniform but token type is not local param.
INTERNAL ERROR: Mismatch between compiler's and linker's count of vertex shader local params.
----- End Shader Log for full ------ joevenzon - 10-16-2007
This is an old thread, but I finally got around to fixing the varying/uniform bug. I was wrong, it was varying. Whoops!
- joevenzon - 12-27-2007
abs1nth, can you see if the problem shown in the screenshot above is fixed in the latest SVN (and data R139)?
- abs1nth - 12-28-2007
i would, but data revision 138 broke shaders on the mac again ;-)
Code:
shaders/depthgen/fragment.glsl:
Error on line 13: assigning non-constant to 'const 4-component vector of float' (hint: '=')
const vec4 tu0color = texture2D(tu0_2D, texcoord);EDIT:
but if i remove the "const" to get rid of this error AND remove the line:
Code:
lightdotnorm = max(dot(lightposition,normal),0.0);(in shaders/depthgen/vertex.glsl) to get rid of this warning:
Code:
"WARNING: vertex shader writes varying 'lightdotnorm' which is not active. "the shaders work fine and the problem (from the screenshot) goes away.
btw. i don't think a warning should disable shader usage.
- joevenzon - 12-28-2007
I made the changes above in R140 (better double check).
I closed this guy:
http://sourceforge.net/tracker/index.php?func=detail&aid=1837684&group_id=137283&atid=1015284
Quote:i don't think a warning should disable shader usage
I agree with you. Fixed in SVN R1904.
- abs1nth - 12-29-2007
hm that bug seemed to be a metabug for all shader problems on macosx, so i don't know if we should close it.
the misrendering issue (10.4) is fixed.
the software fallback issue might exist on some hardware configurations, but i guess there is no way to know, other than making a release with shaderdefault = on and waiting if people report abysmal performance.
the shaders with shadows causing crashes or kernel panics (10.5) is still relevant. btw. apple closed my bugreport about the kernelpanics - they can't reproduce it.
- joevenzon - 12-29-2007
Okay, opened it back up. Any idea how we could get to the bottom of the kernel panics?
- abs1nth - 12-29-2007
no.
(they were willing to investigate further but i wasn't able to do two-machine kernel debugging due to lack of a second machine...)
i'd say don't bother, apple should fix their glsl stuff (even if they closed this bug, the 10.5 opengl stuff WILL improve over time). the best we could do would try to debug it on a machine where it just crashes instead of panics (like kidrock reported) - however since it works fine everywhere else (10.4, linux, etc)...
shadows aren't turned on by default anyway, are they?
- joevenzon - 12-30-2007
They are enabled by default, but if you change data/settings/VDrift.config to say "car_shadows=off" instead of "on", then they will be disabled by default, which I guess would need to be the case for the OS X release.