Re: [RFC PATCHv3] perf tools: add event grouping capability to "perf stat"

[Corey Ashford <cjashfor@linux.vnet.ibm.com>]

On 11/24/2010 10:32 PM, Peter Zijlstra wrote:
> On Wed, 2010-11-24 at 17:54 -0800, Corey Ashford wrote:
>> Add the ability to create multiple event groups, each with their own leader
>> using the existing "-e<event>[,<event>  ...] [-e<event>[,<event>]]"
>> syntax.  Each additional -e switch creates a new group, and each event
>> listed within a -e switch is within that group.
>>
>> Changes since v1:
>> - Because of a flub, v2 did not contain the changes I had intended to make,
>> and instead, v2 had the same patch contents as v1.
>> - When perf stat is not supplied any events on the command line, put
>> each default event in its own group.
>
> I like this, but could you also extend this to perf-record? its a bit
> odd to diverge between the two.
>
> Using Stephane's latest syntax changes you could actually do something
> like:
>
> perf record -e task-clock:freq=1000,cycles:period=0
>
> Which would create a group with 1 sampling counter and a counting
> counter (at which point we should probably start flipping
> PERF_SAMPLE_READ).

Yes, that would be useful.

>
> Matt was working on supporting that (although not through cmdline
> syntax) and teaching perf-report to cope with such output.

I did briefly consider adding this capability to perf record, but I knew 
it would be a lot more complicated.

This perf stat capability is something we added to an internal version, 
and have been using it for more than 6 months.   It's quite helpful for 
verifying that the kernel code for an arch is implemented correctly.

As an alternative approach, how about if instead of changing the 
existing syntax to perf stat, I instead add a -g/--group option which 
takes groups of events?

That way we won't really be diverging perf record and perf stat; we'll 
just have a feature that can at some point later in time be added to 
perf record when all of the details are worked out.

- Corey