Re: [PATCH RFC V5 2/4] perf,tool: per-event time support

[Namhyung Kim <namhyung@kernel.org>]

Hi,

On Tue, Jul 21, 2015 at 03:39:17AM +0000, Liang, Kan wrote:
> > On Mon, Jul 20, 2015 at 03:04:20PM +0000, Liang, Kan wrote:
> > 
> > SNIP
> > 
> > >  		break;
> > > +	case PARSE_EVENTS__TERM_TYPE_TIME:
> > > +		if (time_set)
> > > +			*time_set = true;
> > > +		CHECK_TYPE_VAL(NUM);
> > > +		if (term->val.num > 1)
> > > +			return -EINVAL;
> > > +		if (term->val.num == 1)
> > > +			attr->sample_type |= PERF_SAMPLE_TIME;
> > > +		 break;
> > >  	case PARSE_EVENTS__TERM_TYPE_NAME:
> > >  		CHECK_TYPE_VAL(STR);
> > >  		break;
> > > @@ -611,12 +621,13 @@ do {
> > 			   \
> > >
> > >  static int config_attr(struct perf_event_attr *attr,
> > >  		       struct list_head *head,
> > > -		       struct parse_events_error *err)
> > > +		       struct parse_events_error *err,
> > > +		       bool *time_set)
> > >  {
> > >  	struct parse_events_term *term;
> > >
> > >  	list_for_each_entry(term, head, list)
> > > -		if (config_term(attr, term, err))
> > > +		if (config_term(attr, term, err, time_set))
> > 
> > I think we should rather have some generic way to mart event specific
> > settings otherwise this function prototype will grow wild ;-)
> > 
> 
> I agree.
> 
> > how about posponing static terms configuration after global config was set..
> > like in attached change
> > 
> > 
> > works for period now:
> > 
> > [jolsa@krava perf]$ ./perf record -e 'cpu/instructions,period=20000/' -c
> > 1000 sleep 1 [ perf record: Woken up 1 times to write data ] /proc/kcore
> > requires CAP_SYS_RAWIO capability to access.
> > [ perf record: Captured and wrote 0.015 MB perf.data (35 samples) ]
> > [jolsa@krava perf]$ ./perf evlist -vv
> > cpu/instructions,period=20000/: type: 4, size: 112, config: 0xc0,
> > { sample_period, sample_freq }: 20000
> > 
> 
> So you are going to change current behavior?
> The current behavior is "global opts setting" > "per_event settring" > default
> You are going to change it to "per_event settring" > "global opts setting" >
> Default. Right?

Hmm.. I agree that changing current behavior is not good.  But I think
it makes more sense to prefer per-event settings over global settings
in general.

Thanks,
Namhyung