|
Unreal Tournament III Tips
This information is provided by DGUnreal. Please do not copy this and post it
elsewhere.
StaticMesh Lightmaps versus Vertex-Lighting
It is a common misconception that Vertex-Lighting is a no-cost method of providing shadows
on StaticMeshes, and should therefore be used to improve map performance and decrease file
size. This is false on both accounts.
Lightmaps use a flattened UV set for the mesh triangles similar to a skin texture, which
can be a variable size such as 64x64 as defined by the level designer. This provides
a per-pixel style of shadow data.
Vertex-Lighting uses a data structure that contains light brightness and color information
for each mesh vertex. This data structure consumes a specific non-variable amount of
memory as determined by the mesh vertex count.
In most cases, StaticMeshes should be set to use Lightmaps, as this provides the best
visual quality, the best performance, and can also consume less memory than
Vertex-Lighting.
A comparison of Lightmaps versus Vertex-Lighting provides the following advantages:
- Lightmaps reduce both CPU and GPU load.
- Lightmaps require fewer light-object interactions to be taken into account by the CPU.
- Lightmaps don't have to be rendered in multiple passes by the GPU.
- The Lightmap pass is integrated into the Emissive pass thereby reducing rendering time.
- Lightmaps can reproduce complex per-pixel shadows across each StaticMesh triangle,
whereas Vertex-Lighting can only produce a constant shadow gradient from vertex to vertex.
- StaticMeshes designed to use Lightmaps can be optimized for the minimum number of
triangles, resulting in additional improved performance. StaticMeshes designed for
Vertex-Lighting are usually highly tesselated to increase the number of vertices in an
attempt to improve the shadowing quality, however this has the negative side-effect of
increasing the total object triangle count and impacting rendering performance.
- Lightmaps can be set to very small sizes such as 16x16 pixels or 32x32 pixels to
conserve on texture memory. This is useful for StaticMeshes that are outside of the
main play area, and for StaticMeshes whose shadowing is relatively even across their
surface. Vertex-Lighting cannot be optimized in this manner and always consumes the
same amount of memory to hold the data structure for the total number of StaticMesh
vertices.
- Lightmaps can have UV layouts that are optimized to provide the best shadowing quality.
For example, a spherical rock can have all of the underside triangle UV sizes set
very small so that they don't use very much of the lightmap texture area, which results in
a larger texture area for the visible top-side triangles and higher quality shadows on
that section of the StaticMesh. Vertex-Lighting is always constant for the
StaticMesh's vertex count and relative 3D spacing and cannot be optimized.
A very low triangle and low vertex count StaticMesh set to Vertex-Lighting may consume
less memory than a small Lightmap version, however, the Lightmap version will definitely
look better and will have better performance. Using Lightmaps also allows the level
designer to optimize the shadow quality and memory requirements.
Lightmaps are clearly a better performance choice over Vertex-Lighting.
An example:
UT3 StaticMesh HU_Deco_Pipes.SM.Mesh.S_HU_Deco_Pipes_SM_Pipe01.
This mesh has 2555 triangles and 2393 vertices.
420 instances set to Vertex-Lighting consumes 11MB of lightmap memory.
420 instances set to Lightmap of 32 consumes 850kb of lightmap memory.
420 instances set to Lightmap of 64 consumes 3.3MB of lightmap memory.
The required memory amount is also reflected in the size of the map file.
A percentage of the StaticMeshes in this example can even be increased from Lightmaps of
64 to Lightmaps of 128 or higher for top-quality shadowing and still require less memory
than the Vertex-Lit versions. The Lightmap versions of the StaticMesh also render 8
to 10% faster than the Vertex-Lit versions.
|
|
