flowchart TB
  A["ExecuteIndirect: argument buffer<br/>per command: root args + DISPATCH_RAYS_DESC"] --> E["expand_indirect kernel<br/>on the GPU, per command"]
  C["count buffer"] --> E
  T["host template:<br/>function table, root buffers"] --> E
  E -- "copies the desc's 25 words" --> R["per-command record:<br/>root block + mtl12_rt_dispatch"]
  E -- "W·H or W·H·D over 2^30" --> Z["0 threadgroups:<br/>command dropped"]
  E -- "raygen size 0" --> N["raygen address 0:<br/>rays call nothing"]
  R --> K["mtl12_dispatch_rays<br/>ceil(W/8) × ceil(H/8) × D threadgroups"]
  K --> S["shader table walk<br/>(raygen, miss, hit records)"]