Re: [PATCH] perf: create a helper structure perf_event_context

[Arnaldo Carvalho de Melo]

Em Tue, Aug 21, 2012 at 11:59:20AM +0200, Ingo Molnar escreveu:
> * Andrew Vagin <avagin@xxxxxxxxxx> wrote:
> >  18 files changed, 222 insertions(+), 255 deletions(-)
> > -static int process_sample_event(struct perf_tool *tool,
> > -				union perf_event *event,
> > -				struct perf_sample *sample,
> > -				struct perf_evsel *evsel,
> > -				struct machine *machine)
> > +static int process_sample_event(struct perf_event_context *ectx)
> >  {
> > -	struct perf_annotate *ann = container_of(tool, struct perf_annotate, tool);
> > +	struct perf_sample *sample = ectx->sample;
> > +	struct perf_annotate *ann = container_of(ectx->tool, struct perf_annotate, tool);
> >  	struct addr_location al;
> 
> that looks like a nice cleanliness win.
> 
> Acked-by: Ingo Molnar <mingo@xxxxxxxxxx>

I don't like it, in most cases it just moves variables from the function
signature to the start of the function.

I do agree tho that we should reduce the number of parameters as much as
we can, as I did removing that pevent one that we used to look up
event_format from the sample ID, that just was not natural as we already
had the evsel and did the lookup when reading the perf.data headers, so
it was just a matter of linking evsel to event_format at that point and
remove the pevent parameter and eliminate the needless relookups, see:

 "perf evsel: Cache associated event_format"
 fcf65bf149afa91b875ffde4455967cb63ee0be9

In the above function signature there is already room for tool specific
state in perf_tool/container_of, the other parameters are per event
natural objects.

- Arnaldo
--
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/