PDA

View Full Version : nivida drivers linux



sivlardemalle
06-01-2005, 02:08 AM
Hi everyone,

I'm writing a raytracer on a gpu

The problem is I want to use FBO's. I am a registered developer, but no registered workstation developer (I've had no response on that application). I would like to use the latest nvidia drivers (>76), but I haven't encountered them anywhere ...

Does anybody here have any idea on where to encounter them?

Thanx in advance,
rob(dot)van(dot)dyck(ad)gmail(dot)com

sivlardemalle
06-01-2005, 12:14 PM
A link to my current output :
http://lumumba.luc.ac.be/sivlardemalle/bugatti800_600.png

Thanx in advance,
Rob.

stanlylee
06-01-2005, 04:44 PM
Did you get the driver now ?
If you got it . please share it with me . I have seeked for her long long time ~_~
email : stanlylee@126.com

stanlylee
06-01-2005, 04:56 PM
Nvidia has release their new driver for linux: 76.74

sivlardemalle
06-01-2005, 10:38 PM
Hi,

No, I haven't been able to find them, I google for them every day ...

Regards,
Rob.

stefan
06-01-2005, 10:43 PM
Cool, they released it yesterday.
http://www.nvidia.com/object/linux_display_ia32_1.0-7664.html

stanlylee
06-02-2005, 12:09 AM
I visit the Nvidia Website everyday . but yesterdaty :(

namespace
06-07-2005, 11:43 AM
Hello!

This driver seems to have a GLSL bug.

When I use a GLSL shader with the discard command in it
my renderwindow will stay black and my app seems to be "trapped" in the driver. I have to kill it manually.
The shader compiled, linked and was validated without a problem or any warnings.

Could someone try to reproduce this bug to make sure that it is not my machine / fault.

Thx.

Edit: the previous driver worked fine

Robert Osfield
06-08-2005, 04:31 AM
I have latest NVidia drivers installed under Linux
and shaders with the discard command work fine for me. The test example is from the osgshaders example in the OpenSceneGraph-0.9.9 distrbution.

Running glxinfo on my systems gives me:

OpenGL renderer string: GeForce FX 5900XT/AGP/SSE2/3DNOW!
OpenGL version string: 2.0.0 NVIDIA 76.64


Originally posted by namespace:
Hello!

This driver seems to have a GLSL bug.

When I use a GLSL shader with the discard command in it
my renderwindow will stay black and my app seems to be "trapped" in the driver. I have to kill it manually.
The shader compiled, linked and was validated without a problem or any warnings.

Could someone try to reproduce this bug to make sure that it is not my machine / fault.

Thx.

Edit: the previous driver worked fine

namespace
06-08-2005, 07:18 AM
OpenGL renderer string: GeForce 6800 GT/AGP/SSE2
OpenGL version string: 2.0.0 NVIDIA 76.64

This is getting really strange. Here is my shader (simple parallaxmapping):

uniform sampler2D t_base; // basemap, color(rgb), specularmak(a)
uniform sampler2D t_bump; // bumpmap, normal(rgb), height(a)
uniform sampler2D t_attenuation_xz; // xz-attenuation map
uniform sampler2D t_attenuation_y; // y-attenuation map

uniform float u_specular_exp; // als texcoord für t_spec
uniform float u_parallax_scale; // heightmap scale für parallax offset
uniform vec4 u_light_color;

varying vec3 v_to_viewer; // vektor zum viewer (tangentspace)
varying vec3 v_to_light; // vektor zum light vector (tangentspace)
varying vec3 v_half; // half vector (tangentspace)

varying vec3 v_attenuation_sts;

void main()
{
float attenuation = texture2D(t_attenuation_xz, v_attenuation_sts.xz).x;
attenuation *= texture2D(t_attenuation_y, vec2(v_attenuation_sts.y, 0)).x;

//if(attenuation <= 0.0)
// discard;

// height aus bumpmap laden
float height = texture2D(t_bump, gl_TexCoord[0].xy).a;

vec2 texcoords = gl_TexCoord[0].xy;

// inputvektoren normalisieren
vec3 to_viewer = normalize(v_to_viewer);
vec3 to_light = normalize(v_to_light);
vec3 half = normalize(v_half);

// parallax offset berechnen
height = height * u_parallax_scale - u_parallax_scale * 0.5; // scale und bias
texcoords.x += height * to_viewer.x; // offset x
texcoords.y += height * to_viewer.y; // offset y

vec3 normal = texture2D(t_bump, texcoords).rgb;

// farb- und specularwerte aus der basemap laden
vec4 color = texture2D(t_base, texcoords);

// normals von [0;1] nach [-1;1]
normal -= 0.5;
normal *= 2.0;

// renormalisierung
normal = normalize(normal);
//normal.y = -normal.y; doom3 benutzt y-flip

// phong
const float diffuse = max(dot(normal, to_light), 0.0); // N dot L, diffuse term
const float specular = max(dot(normal, half), 0.0); // N dot H, specular term

color.rgb *= diffuse;
color.rgb *= u_light_color.rgb;
color.rgb += pow(specular, u_specular_exp);
color.rgb *= attenuation;

gl_FragColor = color;
}I have tested this shader in my engine and the shader designer from typhoonlabs (the linux version), both show the
same behaviour.
As long as the

if(attenuation <= 0.0)
discard;

is in comments everything works fine.
If you remove the comments and recompile, the application stops rendering and I have to restart the shaderdesigner / my engine to make it respond again.
But thats not all. If you comment out the

color.rgb *= diffuse;

instruction the shader works again even with the discard command.
I tried to replace the diffuse-variable with the instructions
which calculate its value, but no luck.

I'm really confused. :confused:

namespace
06-08-2005, 07:24 AM
I almost forgot:

The discard-shader ("lattice") which ships with the shader designer works fine, even with the new driver.

fahlen
06-09-2005, 11:23 PM
I experience similar problems with the new NVIDIA graphics driver for Linux, but I haven't been able to pinpoint the problem. Does anyone at NVIDIA know of any bug in the driver that might cause this? Being a driver issue, I don't really know how one would go about debugging to find what the problem is.

mogumbo
06-10-2005, 12:28 PM
I, too, had a problem with one of my shader demos freezing while using the newest NVidia Linux driver. I haven't found the exact problem yet.

namespace
06-10-2005, 11:23 PM
Do you both use GLSL or are the "old" arb_program extensions affected too?

mogumbo
06-11-2005, 10:38 AM
I have only observed a problem with one app that makes heavy use of GLSL, and no arb_vp/arb_fp. Still no luck tracking down the exact problem....

fahlen
06-13-2005, 07:00 AM
Same here.