forge-instanced-rendering

Instanced rendering reduces draw calls by streaming per-instance transforms through a dedicated vertex buffer, allowing the GPU to render many identical meshes in a single call. The shader reconstructs the model matrix on the GPU from per-instance columns.

To implement instancing, bind a per-vertex buffer and a per-instance buffer containing your InstanceData. Upload the instance transforms to the GPU, bind both buffers, and issue an instanced draw call to render all copies.

Use per-instance transforms for scenes with repeated geometry, such as forests, crowds, particle systems, or city blocks. It is ideal when you need to draw a single mesh multiple times with varying spatial transforms efficiently.

Yes, reconstructing the model matrix on the GPU is required. The shader logic reads the per-instance columns from the instance buffer to build the transform matrix directly, eliminating the need to send pre-calculated matrices from the CPU.

The best way is using an instanced draw with a two-buffer pipeline. You define an InstanceData structure, upload an array of transforms to a GPU buffer, and issue a single draw call to render all instances with varying transforms.