WebGPU Image PipelineEvidence pending

Canonical Lab · Three.js 0.185.1

WebGPU Image Pipeline

Interactive Three.js WebGPU/TSL demo for WebGPU Image Pipeline, with 1 scenario, 6 mechanisms, and 3 quality tiers and evidence-gated validation.

Evidence status: The implementation is available for technical review, but its native-WebGPU runtime and evidence gates remain incomplete.

Readiness & remaining fixes 3

Declared evidence work remains. Overall implementation status does not establish individual runtime or performance claims.

  1. P0Define and measure the target performance contract

    The full, reduced, debug tiers have no accepted frame target. Record named device/browser/GPU, viewport, DPR, sustained CPU/GPU/presentation p50 and p95, deadline misses, memory, and settled quality. Submission time is not GPU timing.

    tiers[].frameTargetMs
  2. P0Timestamp Query capability proof

    required for performance claims; correctness capture does not accept current-adapter timestamps. Status incomplete (not accepted).

    capabilityRequirements.timestamp-query
  3. P0Current Adapter Timing runtime proof

    Current-adapter GPU timestamps remain INSUFFICIENT_EVIDENCE; correctness capture does not claim performance timing.

    runtimeProof.current-adapter-timing

Use the linked skill and source to inspect route ownership and evidence classification before treating this presentation as an accepted implementation.

Owning skill
Image Pipeline
Published routes
1 scenarios · 6 mechanisms · 3 quality tiers

Supported mechanisms

  • signal-inspector — Signal Inspector
  • ao-bloom-exposure-order — AO Bloom Exposure Order
  • temporal-velocity-and-history — Temporal Velocity And History
  • output-and-color-ownership — Output And Color Ownership
  • resource-budget — Resource Budget
  • diagnostic-extra-pass — Diagnostic Extra Pass
initializing