Messages in this thread | | | Date | Tue, 23 Jun 2015 12:06:47 -0300 | From | Arnaldo Carvalho de Melo <> | Subject | Re: [PATCHv4 00/27] perf stat: Introduce --per-thread option |
| |
Em Tue, Jun 23, 2015 at 05:08:07PM +0300, Adrian Hunter escreveu: > On 23/06/15 17:05, Arnaldo Carvalho de Melo wrote: > > Em Tue, Jun 23, 2015 at 09:22:00AM +0200, Jiri Olsa escreveu: > >> On Mon, Jun 22, 2015 at 08:06:00PM -0300, Arnaldo Carvalho de Melo wrote: > >>> Em Tue, Jun 23, 2015 at 12:36:01AM +0200, Jiri Olsa escreveu: > >>>> adding the possibility to display stat data per thread. > > > >>>> Allowing following commands and output: > > > >>>> $ perf stat -e cycles,instructions --per-thread -p 30190,30242 > > > >>> While testing Adrian's Intel PT patchkit I realised we have --per-thread > >>> in 'record', wonder if using a long option with the exact same name but > >>> different meanings for 'stat' and 'record' would cause confusion... > > > >> I think the name fits for both stat and record.. and both are doing different > > > > For record it is vague, for stat, it seems to fit. > > > > For record it really should be --mmap-per-thread, but then we start > > getting what may seem overly long options, but then, its an oddball > > 'record' option... > > It is not just the mmap, it is also the perf context. So it is central to > the way perf works.
Ok, so probably it should stay like that :-)
- Arnaldo
| |