forge-point-light-shadows

Omnidirectional point-light shadows in SDL_gpu are implemented by rendering depth from six directions into a TEXTURETYPE_CUBE shadow map, then sampling this cube map in your scene shading pipeline to determine real-time shadow occlusion.

Cube-map shadow mapping for point lights stores linear depth across six faces of a cube texture, capturing depth in all directions around a light source to provide consistent omnidirectional real-time shadows.

You render six faces for a point-light shadow map by creating a TEXTURETYPE_CUBE texture with SDL_CreateGPUTexture, executing per-face render passes using a shared D32_FLOAT depth buffer, and binding the resulting cube map for scene shading.

Yes, SDL_gpu supports real-time shadow rendering by using SDL_CreateGPUTexture with TEXTURETYPE_CUBE to generate six-face depth textures, which are sampled with a NEAREST sampler during the scene shading pipeline passes.

A shared D32_FLOAT depth buffer is used for point-light shadow passes to efficiently render linear depth across all six cube-map faces, ensuring deterministic shadow mapping results while maintaining rendering pipeline performance.

For multiple point lights requiring real-time shadows, the best approach is creating dedicated six-face cube-map shadow textures for each active light, rendering all faces per frame, and binding them for scene shader sampling.