Depending on the architecture, it can take between 80ns to 2us to write an event. So it's very much possible that your configuration takes about 400ns to write an event. This has little to do with clock source precision, but it is rather related to the runtime overhead of the tracer.
Thanks, Mathieu ----- On Jun 20, 2017, at 1:05 PM, Tom Deneau [email protected] wrote: > Is there a write-up on lttng timestamp precision? > > I have noticed on a 4.10 kernel with a mix of kernel and usermode events > that the minimum time between events on a single software thread seems to be > about 400 ns. > > I noticed on this same system that clock_getres(CLOCK_MONOTONIC) reports 1 ns > resolution. > > -- Tom Deneau > > _______________________________________________ > lttng-dev mailing list > [email protected] > https://lists.lttng.org/cgi-bin/mailman/listinfo/lttng-dev -- Mathieu Desnoyers EfficiOS Inc. http://www.efficios.com _______________________________________________ lttng-dev mailing list [email protected] https://lists.lttng.org/cgi-bin/mailman/listinfo/lttng-dev
