Re: [PATCH 07/10] perf, tool: Add support to specify hw event as pmuevent term

From: Stephane Eranian
Date: Thu Jul 05 2012 - 21:08:37 EST


On Wed, Jul 4, 2012 at 2:13 PM, Peter Zijlstra <a.p.zijlstra@xxxxxxxxx> wrote:
> On Wed, 2012-07-04 at 14:00 +0200, Jiri Olsa wrote:
>> On Wed, Jul 04, 2012 at 12:39:20PM +0200, Peter Zijlstra wrote:
>> > On Wed, 2012-07-04 at 00:00 +0200, Jiri Olsa wrote:
>> > >
>> > > It's possible to use sysfs attribute name as 'event' term value,
>> > > or any possible alias for hw event, e.g.:
>> > > cycles OR cpu-cycles
>> > > branch_misses OR branch-misses
>> > > bus_cycles OR bus-cycles
>> > > cache_misses OR cache-misses
>> > > cache_references OR cache-references
>> > > ref_cycles OR ref-cycles
>> > > stalled_cycles_backend OR stalled-cycles-backend
>> > > stalled_cycles_frontend OR stalled-cycles-frontend
>> >
>> > Do we really want to do that?
>>
>> well, the aliasing works on sysfs filename base.. so when you specify
>> event=str
>>
>> 'str' file is looked up in 'events' dir.
>
> Right..
>
>> We use '_' in file names and '-' in event symbol names. I thought it might be
>> confusing allowing just '_' so I added also the '-' version.
>>
What's wrong with keeping the - in the filenames as well? That would help
keep things simple. That would avoid yet another layer of aliases.


>> I could keep just the '-' version.. with some more work on kernel ATTR
>> function names ;) but it seems like common sysfs practise to use '_'.
>
> We could map '-' to '_', but I'm not sure about the aliases in general.
> I mean if the sysfs event name is 'cycles' also allowing 'cpu_cycles' is
> somewhat confusing and arbitrary.
>
> I think it would be best to stick to discoverable names only. Anybody?
>
>
--
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/