Date: Sun, 19 Jul 2015 21:13:22 +0200
From: Jiri Olsa <jolsa@redhat.com>
To: Namhyung Kim <namhyung@kernel.org>
Cc: kan.liang@intel.com, acme@kernel.org, jolsa@kernel.org, ak@linux.intel.com,
        linux-kernel@vger.kernel.org
Subject: Re: [PATCH RFC V5 2/4] perf,tool: per-event time support
Message-ID: <20150719191322.GA27893@krava.brq.redhat.com>
References: <1437132655-15883-1-git-send-email-kan.liang@intel.com>
 <1437132655-15883-3-git-send-email-kan.liang@intel.com>
 <20150718124547.GA3759@krava.brq.redhat.com>
 <20150719032128.GA24219@danjae.kornet>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20150719032128.GA24219@danjae.kornet>
User-Agent: Mutt/1.5.23 (2014-03-12)
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 3181
Lines: 77

On Sun, Jul 19, 2015 at 12:21:28PM +0900, Namhyung Kim wrote:

SNIP

> > 
> > [jolsa@krava perf]$ ./perf record   -e 'cpu/cpu-cycles/,cpu/instructions,time/' kill
> > ...
> > [jolsa@krava perf]$ ./perf evlist -v 
> > cpu/cpu-cycles/: type: 4, size: 112, config: 0x3c, { sample_period, sample_freq }: 4000, sample_type: IP|TID|PERIOD|IDENTIFIER, read_format: ID, disabled: 1, inherit: 1, mmap: 1, comm: 1, freq: 1, enable_on_exec: 1, task: 1, sample_id_all: 1, exclude_guest: 1, mmap2: 1, comm_exec: 1
> > cpu/instructions,time/: type: 4, size: 112, config: 0xc0, { sample_period, sample_freq }: 4000, sample_type: IP|TID|TIME|PERIOD|IDENTIFIER, read_format: ID, disabled: 1, inherit: 1, freq: 1, enable_on_exec: 1, sample_id_all: 1, exclude_guest: 1
> 
> What about this case?
> 
>   $ perf record -e 'cpu/cpu-cycles/,cpu/instructions,time=0/' kill

right.. hum, we need somehow separate/pospone the users term application
to the final perf_event_attr.. spent some time on it today, but could not
find any nice solution so far.. will try tomorrow ;-)

SNIP

> > +	 * If time (-T) is not set and we have events with TIME sample_type
> > +	 * set (tracepoints or events with time term), disable timestamp for
> > +	 * the rest of the events.
> > +	 */
> > +	if (!opts->sample_time_set && perf_evlist__has_time(evlist))
> > +		opts->sample_time = false;
> 
> I think it'd be better if the -T/--timestamp option gives the default
> TIME sample_type value but it can be overridden by per-event setting.
> 
> I mean:
> 
>   (per-event 'time' setting is meaningless here)
>   $ perf record  -e 'cpu/cpu-cycles/,cpu/instructions,time/' kill
> 
>   $ perf evlist -v 
>   cpu/cpu-cycles/: ..., sample_type: IP|TID|TIME|PERIOD|IDENTIFIER, ...
>   cpu/instructions,time/: ..., sample_type: IP|TID|TIME|PERIOD|IDENTIFIER, ...
> 
> 
>   (adding -T option, same as the default)
>   $ perf record -T -e 'cpu/cpu-cycles/,cpu/instructions,time/' kill
> 
>   $ perf evlist -v 
>   cpu/cpu-cycles/: ..., sample_type: IP|TID|TIME|PERIOD|IDENTIFIER, ...
>   cpu/instructions,time/: ..., sample_type: IP|TID|TIME|PERIOD|IDENTIFIER, ...
> 
>   $ perf record -T -e 'cpu/cpu-cycles/,cpu/instructions,time=0/' kill
> 
>   $ perf evlist -v 
>   cpu/cpu-cycles/: ..., sample_type: IP|TID|TIME|PERIOD|IDENTIFIER, ...
>   cpu/instructions,time/: ..., sample_type: IP|TID|PERIOD|IDENTIFIER, ...
> 
> 
>   (adding --no-timestamp option)
>   $ perf record --no-timestamp -e 'cpu/cpu-cycles/,cpu/instructions/' kill
> 
>   $ perf evlist -v 
>   cpu/cpu-cycles/: ..., sample_type: IP|TID|PERIOD|IDENTIFIER, ...
>   cpu/instructions,time/: ..., sample_type: IP|TID|PERIOD|IDENTIFIER, ...
> 
>   $ perf record --no-timestamp -e 'cpu/cpu-cycles/,cpu/instructions,time=1/' kill
> 
>   $ perf evlist -v 
>   cpu/cpu-cycles/: ..., sample_type: IP|TID|PERIOD|IDENTIFIER, ...
>   cpu/instructions,time/: ..., sample_type: IP|TID|TIME|PERIOD|IDENTIFIER, ...

agreed

thanks,
jirka
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/