@group(N) slots.
You typically don’t construct one directly. Two paths:
- Auto (default) — omit
bindGroupLayoutsonGPUPipeline.newand retrieve viapipeline:getBindGroupLayout(N)forGPUBindGroup.new. Mirrors WebGPU’slayout: 'auto'. - Explicit — call
GPUBindGroupLayout.newand pass the layout to N pipelines viabindGroupLayouts = { sharedLayout }. ONE BindGroup then binds across all those pipelines (e.g. a per-frame camera/lights group reused by many materials), avoiding per- pipeline BindGroup duplication and reducing descriptor traffic.