Messages in this thread |  | | Date | Mon, 25 Nov 2013 10:42:59 -0300 | From | Arnaldo Carvalho de Melo <> | Subject | Re: [PATCH v2 2/3] perf script: Print comm, fork and exit events also |
| |
Em Tue, Nov 19, 2013 at 09:18:20AM +0900, Namhyung Kim escreveu: > On Mon, 18 Nov 2013 10:57:00 -0300, Arnaldo Carvalho de Melo wrote: > > Em Mon, Nov 18, 2013 at 02:34:53PM +0900, Namhyung Kim escreveu: > >> If --show-task option is given, also print internal COMM, FORK and > >> EXIT events. It would be helpful for debugging.
> > Humm, --show-task-events is clearer albeit long :-\ With shell > > completion shouldn't be an issue tho.
> Okay, will change.
Ok, waiting for a new patchset then.
> > Also, can't we use: perf_event__fprintf_{mmap,comm,task}?
> We can do it. The output will look like below if we use > perf_event__fprintf():
> swapper 0 [009] 3350640.335261: sched:sched_switch: prev_comm=swapper/9 > sleep 9486 [009] 3350640.335509: PERF_RECORD_COMM: sleep:9486 > sleep 9486 [009] 3350640.335806: sched:sched_stat_runtime: comm=sleep pid=9486 > firefox 2635 [003] 3350641.275896: PERF_RECORD_FORK(2635:9487):(2635:2635) > firefox 2635 [003] 3350641.275896: sched:sched_process_fork: comm=firefox pid=2635 > sleep 9486 [009] 3350641.336009: PERF_RECORD_EXIT(9486:9486):(9486:9486)
> But I'd prefer original version below since it shows the changes more > intuitively, no?
Well, by reusing perf_event__fprintf() we will make it clear the kind of that event, i.e. its not a tracepoint, but a sideband record generated by the perf infrastructure.
Or perhaps we can change perf_event__fprintf* to show:
swapper 0 [009] 3350640.335261: sched:sched_switch: prev_comm=swapper/9 sleep 9486 [009] 3350640.335509: perf:comm: sleep:9486 sleep 9486 [009] 3350640.335806: sched:sched_stat_runtime: comm=sleep pid=9486 firefox 2635 [003] 3350641.275896: perf:fork(2635:9487):(2635:2635) firefox 2635 [003] 3350641.275896: sched:sched_process_fork: comm=firefox pid=2635 sleep 9486 [009] 3350641.336009: perf:exit(9486:9486):(9486:9486)
Would that look nicer here and in 'perf report -D'? Ingo?
- Arnaldo
|  |