Measured on this device
Performance lab
Build, move, inspect, and group repeatable seeded workloads. Results come from this browser—not a marketing estimate.
Measure the full path
Construction, completed rendering, picking, CPU worker force, and the GPU force path use one repeatable workload.
No benchmark has run on this device yet.
Timing is device-specific. Memory is allocated graph typed arrays, not total heap. Reports include browser/GPU identifiers; review before sharing.
Move a large network
Arrange the same graph with the CPU worker or GPU path. GPU is not automatically faster: setup and readback can outweigh parallel work on smaller networks.
CPU worker · not run
GPU path · not run
Motion not run
Single-run wall time; use Measure for repeated evidence.
Filter without rebuilding
Filter the typed graph in place. Hover any visible node to isolate its immediate neighbourhood.
Filter time —
Change a grouped network
Contain up to 150 spatial groups, add a node and edge to every group, or grow only the busiest group while the wider network keeps its context.
Backend CPU main thread
Last change not run