LeanProfiler

3.2. Capture totals🔗

trace_window_ns runs from the earliest retained start to the latest retained end.

recorded_thread_time_ns is the sum of every event's self time. Parallel threads can make this larger than the trace window. It is not CPU utilization.

recorded_heartbeats is the matching sum of self heartbeats. Each row's share_permille is:

1000\cdot \frac{\text{row self time}}{\text{recorded thread time}}

The implementation uses integer division, so displayed percentages can lose a tenth of a percent and need not sum to exactly 100%.