Module gpu 

Thin helpers over Godot’s main RenderingDevice.

Every function here must run on the render thread (the buffer RIDs we touch belong to the global device) — callers go through RenderingServer::call_on_render_thread, CEL-58 style.

Modules

chunk_gpu

All RD resources for one chunk multimesh: the ChunkRealize compute pipeline + the shared vertex pool (positions buffer + attribute texture) + the descriptor/params buffers + uniform set + an indirect MultiMesh. Render-thread only. The chunk realize is a single vertex pass: it reads chunks/params and writes the vertex pool — it does NOT bind the multimesh, so the compute path is multimesh-independent.

device

Pipeline / buffer / texture creation helpers (render-thread only).

owned

RAII ownership of Godot RenderingDevice RIDs (CEL-91).

Constants

ATTR_TEX_WIDTH

Width (texels) of the GPU attribute/atlas textures: chunk vertex pools and per-chunk detail atlases are 1-D runs wrapped at this width. Shaders read it as attr_w to recover 2-D texel coords from a linear index.