[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <2d693635-9697-2cf5-54dc-b91da4dfd14f@linux.intel.com>
Date: Tue, 28 May 2019 14:24:56 -0400
From: "Liang, Kan" <kan.liang@...ux.intel.com>
To: Peter Zijlstra <peterz@...radead.org>
Cc: acme@...nel.org, mingo@...hat.com, linux-kernel@...r.kernel.org,
tglx@...utronix.de, jolsa@...nel.org, eranian@...gle.com,
alexander.shishkin@...ux.intel.com, ak@...ux.intel.com
Subject: Re: [PATCH 4/9] perf/x86/intel: Support hardware TopDown metrics
On 5/28/2019 9:48 AM, Peter Zijlstra wrote:
> On Tue, May 21, 2019 at 02:40:50PM -0700, kan.liang@...ux.intel.com wrote:
>> diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h
>> index b980b9e95d2a..0d7081434d1d 100644
>> --- a/include/linux/perf_event.h
>> +++ b/include/linux/perf_event.h
>> @@ -133,6 +133,11 @@ struct hw_perf_event {
>>
>> struct hw_perf_event_extra extra_reg;
>> struct hw_perf_event_extra branch_reg;
>> +
>> + u64 saved_metric;
>> + u64 saved_slots;
>> + u64 last_slots;
>> + u64 last_metric;
>
> This is really sad, and I'm thinking much of that really isn't needed
> anyway, due to how you're not using some of the other fields.
If we don't cache the value, we have to update all metrics events when
reading any metrics event. I think that could bring high overhead.
Thanks,
Kan
>
>> };
>> struct { /* software */
>> struct hrtimer hrtimer;
>> --
>> 2.14.5
>>
Powered by blists - more mailing lists