Re: [PATCH V2 07/14] perf tools: Add evsel->own_cpus

[Arnaldo Carvalho de Melo <acme@redhat.com>]

Em Tue, Sep 08, 2015 at 09:38:54AM +0200, Jiri Olsa escreveu:
> On Tue, Sep 08, 2015 at 10:09:10AM +0300, Adrian Hunter wrote:
> > On 08/09/15 09:45, Jiri Olsa wrote:
> > > On Mon, Sep 07, 2015 at 05:27:49PM +0300, Adrian Hunter wrote:
> > >> +++ b/tools/perf/util/evlist.c
> > >> @@ -1111,9 +1111,12 @@ static void perf_evlist__propagate_maps(struct perf_evlist *evlist)
> > >>  		 * We already have cpus for evsel (via PMU sysfs) so
> > >>  		 * keep it, if there's no target cpu list defined.
> > >>  		 */
> > >> -		if (!evsel->cpus || evlist->has_user_cpus) {
> > >> +		if (!evsel->own_cpus || evlist->has_user_cpus) {
> > >>  			cpu_map__put(evsel->cpus);
> > >>  			evsel->cpus = cpu_map__get(evlist->cpus);
> > >> +		} else if (evsel->cpus != evsel->own_cpus) {
> > >> +			cpu_map__put(evsel->cpus);
> > >> +			evsel->cpus = cpu_map__get(evsel->own_cpus);
> > > 
> > > hum, so (evsel->cpus != evsel->own_cpus) could happen only when:
> > >   - evsel->own_cpus != NULL
> > >   - we overloaded evsel->cpus with evlist->cpus via perf_evlist__propagate_maps
> > >   - we changed evlist->has_user_cpus = false
> > >   - we recall perf_evlist__propagate_maps

> > > I'm missing usecase for that, or something else ;-)

> > That's true but the idea is to establish rules (invariants) that are always
> > true.  Like an evsel either has its own cpu map or the same as the evlist.
 
> ook, jirka

Yup, I haven't even looked at the code that much, but at least Adrian's
explanation matches my expectations 8-)

- Arnaldo