Re: [PATCH] perf record: Set timestamp boundary for AUX area events

From: Andi Kleen
Date: Mon May 03 2021 - 12:11:39 EST


On Mon, May 03, 2021 at 06:06:38PM +0300, Adrian Hunter wrote:
> On 3/05/21 5:56 pm, Andi Kleen wrote:
> > On Mon, May 03, 2021 at 09:42:22AM +0300, Adrian Hunter wrote:
> >> AUX area data is not processed by 'perf record' and consequently the
> >> --timestamp-boundary option may result in no values for "time of first
> >> sample" and "time of last sample". However there are non-sample events
> >> that can be used instead, namely 'itrace_start' and 'aux'.
>
> "instead" -> "as well"
>
> >> 'itrace_start' is issued before tracing starts, and 'aux' is issued
> >> every time data is ready.
> >
> > Hmm, what happens when some other non PT events are in the same perf record?
>
> The sample timestamps are still processed, so the lowest sample timestamp or
> 'itrace_start' timestamp or 'aux' timestamp is the start. Similarly for the end.
>
> > And those maybe run at different times than PT (e.g. due to some PT specific
> > filter). Does this all work correctly then?
>
> The broadest range is used

Ok makes sense.

Acked-by: Andi Kleen <ak@xxxxxxxxxxxxxxx>

-Andi