Messages in this thread | | | Date | Mon, 29 Jun 2020 23:26:52 -0400 | From | Steven Rostedt <> | Subject | Re: [RFC][PATCH] ring-buffer: Have nested events still record running time stamp |
| |
On Mon, 29 Jun 2020 23:13:32 -0400 (EDT) Mathieu Desnoyers <mathieu.desnoyers@efficios.com> wrote:
> -> This configuration's avg varies between 239 and 256 ns based on your benchmark methodology. The std. dev. varies between > 138 and 201 ns. > > In the benchmark results you collected, config #1 appeared to have little delta between xadd and cmpxchg, whereas > config #2 and #3 had some avg difference. Did you try running this test many times on each machine to see if the > result was indeed stable ?
Yes, but as we discussed on IRC, my machines are much smaller than this one, and that makes it a bit more deterministic. I also run them with distro configs. When there's debugging enabled, I get much more erratic results.
-- Steve
| |