Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Message-ID: <ef780335-7dc6-3a23-54a8-b6fc9c8a2ed3@intel.com>
Date:   Wed, 16 Mar 2022 13:19:46 +0200
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:91.0) Gecko/20100101
 Firefox/91.0 Thunderbird/91.5.0
Subject: Re: [PATCH] perf intel-pt: Synthesize cycle events
Content-Language: en-US
To:     "Steinar H. Gunderson" <sesse@google.com>
Cc:     Peter Zijlstra <peterz@infradead.org>,
        Ingo Molnar <mingo@redhat.com>,
        Arnaldo Carvalho de Melo <acme@kernel.org>,
        Alexander Shishkin <alexander.shishkin@linux.intel.com>,
        Jiri Olsa <jolsa@redhat.com>,
        Namhyung Kim <namhyung@kernel.org>,
        linux-perf-users@vger.kernel.org, linux-kernel@vger.kernel.org
References: <20220310093844.982656-1-sesse@google.com>
 <586de5fc-858b-2693-1986-5c77e8c0e3d0@intel.com>
 <YiuKAk7SaXP7B7Ee@google.com>
 <ba2c49da-22c5-06ea-e953-82211b953ca8@intel.com>
 <YjBnet2813sEGjZY@google.com>
 <c50bb87d-9fee-c4f9-a350-8729e503e43a@intel.com>
 <YjDUO6bbyfGw/u0C@google.com>
 <52903e58-e74c-5ea0-36b4-277ea3610af4@intel.com>
 <YjGdoGy4Z2UUG9S9@google.com>
From:   Adrian Hunter <adrian.hunter@intel.com>
Organization: Intel Finland Oy, Registered Address: PL 281, 00181 Helsinki,
 Business Identity Code: 0357606 - 4, Domiciled in Helsinki
In-Reply-To: <YjGdoGy4Z2UUG9S9@google.com>
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 7bit
Precedence: bulk

On 16.3.2022 10.19, Steinar H. Gunderson wrote:
> On Tue, Mar 15, 2022 at 10:11:54PM +0200, Adrian Hunter wrote:
>> Sorry, my first suggestion has issues, but the second is better.
> 
> I tried your second one, but can't see much difference. The original and
> your patched version still differ by a lot, and I still see various
> non-branch instructions getting very tiny fractions. (PSB packets
> _might_ be it, as there are ~1000 of them in the 165M-cycle trace.)
> 
> I guess the good news is that the perf report coming out of your version
> looks more likely to me; I have some functions that are around 1% that
> shouldn't intuitively be that much (and, if I write some Perl to sum up
> the cycles from the IPC lines in perf script, are more around 0.1%).
> So perhaps we should stop chasing the difference? I don't know.

That doesn't sound right.  I will look at it more closely in the next few days.