Re: [PATCH 26/31] perf tools: Support perf event alias name

[Wangnan (F)]

On 2015/10/22 15:16, Namhyung Kim wrote:
> Hi,
>
> On Wed, Oct 21, 2015 at 10:00 PM, Wangnan (F) <wangnan0@xxxxxxxxxx> wrote:
> > On 2015/10/21 16:53, Namhyung Kim wrote:
> > > Hi,
> > >
> > > On Wed, Oct 14, 2015 at 12:41:37PM +0000, Wang Nan wrote:
> > > > From: He Kuang <hekuang@xxxxxxxxxx>
> > > >
> > > > This patch adds new bison rules for specifying an alias name to a perf
> > > > event, which allows cmdline refer to previous defined perf event through
> > > > its name. With this patch user can give alias name to a perf event using
> > > > following cmdline:
> > > >
> > > >    # perf record -e mypmu=cycles ...
> > > >
> > > > To allow parser refer to existing event selecter, pass event list to
> > > > 'struct parse_events_evlist'. perf_evlist__find_evsel_by_alias() is
> > > > introduced to get evsel through its alias.
> > > What about using event name directly?  I guess the alias name is used
> > > only to refer an event so it'd be better to use the event name.
> > > Anyway we need alias as well when event has no name or name is complex.
> >
> > It is possible to trigger two perf events with same PMU but
> > different config options:
> >
> >   # perf record -e cycles/period=9999/ -e cycles/period=99999/ -a sleep 1
> >
> > In this case the name of events are:
> >
> > cycles/period=9999/ `
> > cycles/period=99999/
> >
> > Using it in perf cmdline is painful:
> >
> >   # perf record -e cycles/period=9999/ -e cycles/period=99999/ -e
> > bpf_program.c/myevent=cycles/period=9999//...
> I understand the need of using aliases but I think it's more natural
> to use event name for simple cases..

I will consider this. However, if we allow using event name directly 
like this:

 # perf record -e cycles -e test_pmu.c/myevent=cycles/ ...

Then two '-e' seems redundant, right? Why not directly using:

 # perf record -e test_pmu.c/myevent=cycles/ ...

and make perf creates cycles event for test_pmu.c?

We can make syntax like

 # perf record -e test_pmu.c/myevent=cycles/ ...

as a syntax sugar of

 # perf record -e randomname=cycles -e test_pmu.c/myevent=randomname/ ...

and don't need to find evsel through their names if alias not exist.

So this is a new feature, and worth another patch.

Thought?

> Thanks,
> Namhyung


--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/