LAB 156 / GPU COMPACTION

Compact dispatch

Mark active particles, compact their IDs into a dense storage list, and let a GPU-written indirect buffer determine how many workgroups the next kernel receives.

—Preparing

Studying GPU-generated dispatch metadata.

LIVE PARAMETERS

Shape active work

Time scale 1.00 Activity threshold 0.52 Motion field 0.42 Pause marking and compaction

The GPU compaction pass is intentionally serial over a small pedagogical sample; the following motion kernel is dispatched through a GPU-written indirect workgroup buffer. This exposes the contract, not a universal prefix-sum implementation.

Source particles—

Active list—

Workgroups—

Dispatches—

Draw calls—

Storage—

The route will identify whether the current renderer is using GPU-generated indirect dispatch or the CPU reference.

Benchmark: not run

RUNTIME ERROR

Compact dispatch

Shape active work

The compaction study could not start.