On Wed, Dec 14, 2016 at 4:12 AM, John Ferlan <jferlan(a)redhat.com> wrote:
On 12/13/2016 11:01 AM, Nitesh Konkar wrote:
> Add support and documentation for generic
> perf events like bus_cycles, stalled_cycles_frontend,
> stalled_cycles_backend and ref_cpu_cycles.
>
> Nitesh Konkar (4):
> perf: add bus_cycles perf event support
> perf: add stalled_cycles_frontend perf event support
> perf: add stalled_cycles_backend perf event support
> perf: add ref_cpu_cycles perf event support
>
> docs/formatdomain.html.in | 24 +++++++++++++++++
> docs/news.html.in | 6 +++--
> docs/schemas/domaincommon.rng | 4 +++
> include/libvirt/libvirt-domain.h | 40
+++++++++++++++++++++++++++++
> src/libvirt-domain.c | 11 ++++++++
> src/qemu/qemu_driver.c | 4 +++
> src/util/virperf.c | 16 +++++++++++-
> src/util/virperf.h | 4 +++
> tests/genericxml2xmlindata/generic-perf.xml | 4 +++
> tools/virsh.pod | 16 +++++++++++-
> 10 files changed, 125 insertions(+), 4 deletions(-)
>
As long as you're OK with my word-smithing in patches 2 & 3 regarding a
longer description for those cycles values... Then these are fine and I
can push them once I get the OK.
Makes sense. I m OK with it. Thanks!
If you can provide a better explanation
for ref_cpu_cycles that would be fantastic.
Based on the below two links:
http://lkml.iu.edu/hypermail/linux/kernel/1112.1/01029.html
http://man7.org/linux/man-pages/man2/perf_event_open.2.html
I have come up with the following two explanations.
The count of total cpu cycles not affected by CPU frequency
scaling by applications running on the platform
OR
The count of the number of reference core cpu cycles
by applications running on the platform
Tks -
John