ThreeDream - GPU particle layer

no adapter 0 particles cpu-upload 0 fps

Probing adapters…

Tier

running on: —
asked for: —
renderer backend: —
device shared: —
instance path: —
blitted / frame: —
feature level: —

Not built yet.

Simulation

radius scale 1.0×

Counters

particles: —
steps / frames: —
steps per frame: —
ms / frame: —
draw calls: —
triangles: —
contacts: —
escaped: —
hash overflow: —
max speed: —
kinetic energy: —
digest: —

Log

The WGSL kernels and the renderer share one GPUDevice, and the instance matrices are expanded on the device straight into the buffer three.js owns — no particle position crosses the bus. ParticleView never writes the field, so replay and training stay independent of this layer; e2e/particles_gpu.spec.ts asserts CPU/GPU parity.