Expand description
2026-09-26: L2 scheduler and L3 cache instruments (queue depths, batch
shape, loop phase timings, blocking syncs and D2H copies, CUDA-graph
replays and captures, KV-block and SSM-slot occupancy) and the L4
SpecMatrix.
Fed through the scheduler’s TelemetryIo router and the GPU backend’s
copy, sync and graph entry points. The prefix-cache hit and miss counters
live in crate::run_metrics, not here.
Owner: telemetry. Invariants: none beyond the types.
Structs§
- Sched
Instruments - Sched
Shape - 2026-09-26: One tick’s scheduler shape, as the scheduler publishes it.
- Spec
Matrix
Constants§
- MAX_
DRAFTS - 2026-09-26: The widest verify step
SpecMatrixplaces. - MAX_
PHASES - 2026-09-26: The most loop phases
Telemetry::set_phase_nameskeeps.