Recent Posts

Pages: 1 2 [3] 4 5 ... 10
21
English forum / GLSL compiler: NVIDIA vs AMD vs Intel (02)
« Last post by JeGX on November 16, 2014, 12:11:28 PM »
In a recent shader, I wrote this line:
Code: [Select]
float lum =  T * vec4(0.21, 0.72, 0.07, 0);

where T is a vec4.

This line works perfectly on NVIDIA GLSL compiler but generates an error on AMD compiler:
Quote
ERROR: 0:78: '=' :  cannot convert from '4-component vector of float' to 'float' 

The error produced by AMD GLSL compiler is correct and NVIDIA should do the same thing.  Weirdly, the shader works properly and the rendering was  the one expected on NVIDIA GPUs.

The right code is:
Code: [Select]
float lum =  dot(T, vec4(0.21, 0.72, 0.07, 0));

which is ok on all graphics hardware!
22
General Discussion / Re: 3D Algorithms
« Last post by JeGX on November 16, 2014, 11:58:34 AM »
Thanks fro your links. I updated the subject of your topic to be more relevant.
23
It rules!  ;)
24
General Discussion / 3D Algorithms
« Last post by IorPerry on November 16, 2014, 09:35:54 AM »
Hello everybody,
My name is Antonino Perricone, I am a 3D programmer since the year 2000.
I just write the explanation of 2 algorithms that every 3D programmer must know, I think.
You can read it in: http://aperricone.altervista.org/algorithms/
They have an animated area where the algorithm is implemented.
I'd like to know your opinion.
Regards to all,
Perry
25
English forum / GLSL Hacker Blog
« Last post by JeGX on November 14, 2014, 07:49:58 PM »
26
3D-Tech News Around The Web / GPUPI - OpenCL GPU (CPU option)
« Last post by nuninho1980 on November 12, 2014, 10:17:21 PM »
Quote
About two months ago our beloved Turrican passed away in a horrible car accident. He was a big part of our community and is still Austrias only overclocking lengend! Out of my sadness and anger I started to work on this project as a virtual monument, something to honour him in our scene. It's an homage to SuperPI, that Turrican benched on every possible platform, an application that calculates pi completely parallelized on a graphics card. So let's get our gear going and do exactly what our Karl showed us in all his years: Bench the hell out of it!

More info

DOWNLOAD

I got running i5-2500K@4.5GHz and GTX 580@reference-stock 3GB-344.60WHQL:
At 32 millions-digit:
-> GPU

result: 0.681s

-> CPU

result: 17.313s

GPU is 25.4x faster than CPU! :D


GPU - 1 billion-digit: :D

result: 57.095s
27
3D-Tech News Around The Web / Re: NVIDIA Apollo 11 Demo for Maxwell GPUs
« Last post by JeGX on November 12, 2014, 12:09:15 PM »
This requires a GM204 GPU (GTX 980/970) due to the use of the following features: Viewport Multicast, Conservative Raster and Tiled Resources.

Quote
In traditional computer graphics, a light contributes diffuse and specular illumination to any surface that has an unobstructed ray to that light.  The images resulting from this technique are characteristically black in unlit regions because this over-simplification fails to account for light reflected from other surfaces in the scene.

To improve visuals, most games describe indirect illumination by pre-computing lighting and storing the result statically in vertex data or textures.  The results can look very realistic if the scene remains static, but we want to be able to open doors, move lights, and tear down walls.  Clearly games would benefit greatly if we could create a "global illumination" solution that computes direct and indirect lighting in real time.

NVIDIA's VXGI computes indirect light by rendering the scene's lit geometry into a 3D voxel grid, then using that grid as an acceleration structure for computing indirect diffuse light and reflections.  Indirect diffuse light is calculated by tracing broad cones through the voxel grid in the direction of the surface normal and accumulating the light from those voxels.  Reflections are likewise calculated by tracing through the voxel grid in the direction of the reflection vector.

This new technique is made possible from several new features of GeForce GTX 980 including:

"Viewport Multicast" :
  Accelerates the rendering of each triangle into the voxel structure(s)
  by broadcasting it to the 6 directional render targets rather than
  duplicating data.

"Conservative Raster" :
  Ensures that each triangle in a voxel's space can contribute to that
  voxel even if the triangle does not cross that voxel's sample point.

"Tiled Resources" :
  Permits us to create a high resolution 3D Texture but only allocate
  memory for those regions that are occupied by voxels.

NVIDIA can now employ Voxelized Global Illumination on Geforce GTX 980 to test the validity of the alleged moon landing media.
28
3D-Tech News Around The Web / Re: NVIDIA Apollo 11 Demo for Maxwell GPUs
« Last post by oldcrow on November 12, 2014, 03:33:12 AM »
I tried to run it with my EVGA GTX 750 Ti, and a message box was displayed saying that no "Maxwell GPU" could be found in my system....
29
3D-Tech News Around The Web / Re: NVIDIA Apollo 11 Demo for Maxwell GPUs
« Last post by nuninho1980 on November 11, 2014, 10:58:40 PM »
Can GTX 750 series run this demo after Fermi and Kepler failed?
Pages: 1 2 [3] 4 5 ... 10