Deano Calver
Newcomer
As a mentioned them thought I'd throw up the abstracts, I'll see about getting the real things posted here soon (assuming legals o.k. and Dave doesn't mind).
Deferred Lighting on PS 3.0 with High Dynamic Range
Deferred lighting is a technique increasingly being used for real-time rendering. With the new capabilities of PS 3.0 and high precision blending we can easily move to high dynamic range.
This article also covers high dynamic range tone-mapping for PS 3.0 in detail, using a GPU efficient method. The technique is independent of the method of creating the buffer that needs to be tone-mapped, so can be used in a conventional non-deferred renderer.
Accessing and modifying topology on the GPU.
A vertex shader can only access a single vertex at a time; the hardware has fixed function indexed triangle model that doesn't allow a vertex shader to manually do a vertex indirection. This seems to limit the extent that topology can be accessed and modified on the GPU, this article introduces several techniques that essentially bypass this limitation allowing complex operations that involve accessing and in some cases modifying topology on the GPU.
The two major techniques are the related ideas of topology and vertex maps, texture maps that hold the mesh itself. The ability of vertex shaders to access video memory surfaces (textures) is the heart of the system and limits this technique to vertex shader 3.0 or higher. The article also uses the render to vertex map approaches to optimise many operations by caching data across shader executions. In some cases the vertex maps themselves are created on the GPU just prior to being used to generate the visuals.
Deferred Lighting on PS 3.0 with High Dynamic Range
Deferred lighting is a technique increasingly being used for real-time rendering. With the new capabilities of PS 3.0 and high precision blending we can easily move to high dynamic range.
This article also covers high dynamic range tone-mapping for PS 3.0 in detail, using a GPU efficient method. The technique is independent of the method of creating the buffer that needs to be tone-mapped, so can be used in a conventional non-deferred renderer.
Accessing and modifying topology on the GPU.
A vertex shader can only access a single vertex at a time; the hardware has fixed function indexed triangle model that doesn't allow a vertex shader to manually do a vertex indirection. This seems to limit the extent that topology can be accessed and modified on the GPU, this article introduces several techniques that essentially bypass this limitation allowing complex operations that involve accessing and in some cases modifying topology on the GPU.
The two major techniques are the related ideas of topology and vertex maps, texture maps that hold the mesh itself. The ability of vertex shaders to access video memory surfaces (textures) is the heart of the system and limits this technique to vertex shader 3.0 or higher. The article also uses the render to vertex map approaches to optimise many operations by caching data across shader executions. In some cases the vertex maps themselves are created on the GPU just prior to being used to generate the visuals.
