indirect-first-instance

GPU-driven pipelines build their draw lists on the GPU: a culling compute pass writes indirect draw commands, and firstInstance is the natural slot for which per-object data a draw uses. Without this feature, indirect draws must keep firstInstance at zero and route object indices through other channels (instance step-mode vertex buffers or extra bindings).

Engines doing large-scene culling, LOD selection, or visibility buffers on the GPU care most. The workaround is well known but costs an indirection; availability data tells you how often you would actually take the slow path.