Re: [PATCH 02/14] perf/x86: output NMI overhead

[Peter Zijlstra]

On Wed, Nov 23, 2016 at 04:44:40AM -0500, kan.liang@xxxxxxxxx wrote:
> From: Kan Liang <kan.liang@xxxxxxxxx>
> 
> NMI handler is one of the most important part which brings overhead.
> 
> There are lots of NMI during sampling. It's very expensive to log each
> NMI. So the accumulated time and NMI# will be output when event is going
> to be disabled or task is scheduling out.
> The newly introduced flag PERF_EF_LOG indicate to output the overhead
> log.
> 
> Signed-off-by: Kan Liang <kan.liang@xxxxxxxxx>
> ---
>  arch/x86/events/core.c          | 19 ++++++++++++++-
>  arch/x86/events/perf_event.h    |  2 ++
>  include/linux/perf_event.h      |  1 +
>  include/uapi/linux/perf_event.h |  2 ++
>  kernel/events/core.c            | 54 ++++++++++++++++++++++-------------------
>  5 files changed, 52 insertions(+), 26 deletions(-)

The subject says this patch is x86 specific, but most changes are in
core code.