Re: [PATCH V2 5/6] perf,tools: caculate and save freq/CPU%/CORE_BUSY% in he_stat

[Jiri Olsa]

On Mon, Jul 27, 2015 at 05:19:23PM +0000, Liang, Kan wrote:
> 
> 
> > > -481,6 +486,32 @@ iter_add_next_nop_entry(struct hist_entry_iter
> > *iter
> > > __maybe_unused,  }
> > >
> > >  static int
> > > +iter_add_single_freq_perf_entry(struct hist_entry_iter *iter, struct
> > > +addr_location *al) {
> > > +	struct perf_evsel *evsel = iter->evsel;
> > > +	struct perf_sample *sample = iter->sample;
> > > +	struct hist_entry *he;
> > > +	struct freq_perf_info info = {0};
> > > +	u64 *data = sample->freq_perf_data;
> > > +
> > > +	if (data[FREQ_PERF_REF_CYCLES] > 0)
> > > +		info.freq = (data[FREQ_PERF_CYCLES] * cpu_max_freq) /
> > data[FREQ_PERF_REF_CYCLES];
> > > +	if (data[FREQ_PERF_TSC] > 0)
> > > +		info.cpu_u = (100 * data[FREQ_PERF_REF_CYCLES]) /
> > data[FREQ_PERF_TSC];
> > > +	if (data[FREQ_PERF_MPERF] > 0)
> > > +		info.core_busy = (100 * data[FREQ_PERF_APERF]) /
> > > +data[FREQ_PERF_MPERF];
> > 
> > seems to me the new iterator is too big gun for this, why not initialize
> > 'struct freq_perf_info' in iter_prepare_normal_entry ?
> > 
> 
> Yes, we can initialize it in normal process.
> But we only need the freq_perf_info when "--show-freq-perf" is set.
> For most of the cases, it wastes time to calculate the unused information
> for each sample. That could impact the performance for processing. 

hum, it could fill in freq_perf_info only when --show-freq-perf is set,
other time it'd pass NULL

jirka
--
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/