Latency and per-call overhead. These cases are dominated by launch, dispatch and synchronization, not by device compute. Read them as API overhead, never as GPU/CPU throughput.
8fec1ca70852No latency and per-call overhead cases.
cpu/fft (collection date unrecorded)cpu/overhead_smoke (2026-10-28)cpu/session_matrix (collection date unrecorded)cpu/small_work (collection date unrecorded)