Color depth more than 24(32)bits. Is it possible?
Author
I need to blend a lot of primitives (let's say >> 256).
In standard situation buffer has 8 bits / channel (r,g,b,a).
Because of this a color [component] less than 1/255 is rounded to zero (if I used greater (multiplied) color I would receive overflow).
Is there a way to render and blend in OpenGL with better color accuracy? ("ACCUM" buffer has 16bits/channel, but it is not proper solution because of several reasons).
Is there a way to do this with any other API than OpenGL?
2D is also ok for me :)
Thank you in advance for your suggestions.
Can't the Matrox Parhelia do 10 bits/channel?
I remember a highly amusing ad on their website with two images, supposedly one 8 bits/channel and one 10 bits/channel, next to each other with the caption "See the difference!". Of course, unless you already had one both pictures would obviously be displayed on an 8 bits/channel monitor... :P
Otherwise, can't some new hardware do floating-point higher-dynamic-range pixel buffers?
I remember a highly amusing ad on their website with two images, supposedly one 8 bits/channel and one 10 bits/channel, next to each other with the caption "See the difference!". Of course, unless you already had one both pictures would obviously be displayed on an 8 bits/channel monitor... :P
Otherwise, can't some new hardware do floating-point higher-dynamic-range pixel buffers?
You can use a 128 bit floating point render-target on newer cards (Radeon 9500+, GeforceFX). Unfortunately, there is no good way of doing this in OpenGL atm. However, if you are using NVIDIA hardware you can use pbuffers + some NV-specific extensions (for rect textures and floating point formats), otherwise you'll have to wait for the upcomming EXT/ARB_render_target~ extension.
floating point pbuffers combined with floating point textures as render targets will currently allow you to do it, its a tad clunky however (apprently ATI dont have a specific extension for it, it just... works).
However, it will be remapped into the 8bit/channel range when displayed as currently there isnt a way to support > 32bit color on screen (the GF6800 has the abilty, but i dont belive NV have exposed it yet via OpenGL)
However, it will be remapped into the 8bit/channel range when displayed as currently there isnt a way to support > 32bit color on screen (the GF6800 has the abilty, but i dont belive NV have exposed it yet via OpenGL)
Author
Quote:You can use a 128 bit floating point render-target on newer cards (Radeon 9500+, GeforceFX). Unfortunately, there is no good way of doing this in OpenGL atm.
How can I do it in any "usual" way? DirectX? Or this is is just a declared possibility that can be used only in very specific way?
Can you please give me any links or tips to find necessary documents?
Thnks.
Quote:Original post by Cro
How can I do it in any "usual" way? DirectX? Or this is is just a declared possibility that can be used only in very specific way?
Can you please give me any links or tips to find necessary documents?
Thnks.
Yeah, in Direct3D this is easy to do. Anyway, here are some links that tell you how to set up a float render-target on NVIDIA hardware.
http://developer.nvidia.com/attach/6725
http://oss.sgi.com/projects/ogl-sample/registry/NV/render_texture_rectangle.txt - non-power of two render targets
http://oss.sgi.com/projects/ogl-sample/registry/NV/float_buffer.txt - float render target, 64/128 bits per pixel
http://www.google.com
It's perfectly possible to render to a floating point render target in OpenGL on ATI hardware. Check out WGL_ATI_pixel_format_float and ATI_EXT_texture_float. FWIW, general floating point support is likely to be included in the OpenGL 2.0 spec.
Quote:Original post by GameCat
It's perfectly possible to render to a floating point render target in OpenGL on ATI hardware. Check out WGL_ATI_pixel_format_float and ATI_EXT_texture_float. FWIW, general floating point support is likely to be included in the OpenGL 2.0 spec.
Yes, if you limit yourself to power-of-two render targets.
Author
Thank you all, I am reading suggested documents.
> FWIW, general floating point support is likely to be included in the OpenGL 2.0 spec
And when it's going to happen? :)
> FWIW, general floating point support is likely to be included in the OpenGL 2.0 spec
And when it's going to happen? :)
This topic is closed to new replies.
Advertisement
Popular Topics
Advertisement
