Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
From:   John Garry <john.garry@huawei.com>
Subject: perf arm64 --topdown support (was "perf arm64 metricgroup support")
To:     Jiri Olsa <jolsa@redhat.com>, Andi Kleen <ak@linux.intel.com>,
        "Arnaldo Carvalho de Melo" <acme@kernel.org>
CC:     <will@kernel.org>, <mathieu.poirier@linaro.org>,
        <leo.yan@linaro.org>, <peterz@infradead.org>, <mingo@redhat.com>,
        <acme@kernel.org>, <mark.rutland@arm.com>,
        <alexander.shishkin@linux.intel.com>, <namhyung@kernel.org>,
        <irogers@google.com>, <linuxarm@huawei.com>, <kjain@linux.ibm.com>,
        <kan.liang@linux.intel.com>, <linux-kernel@vger.kernel.org>,
        <linux-arm-kernel@lists.infradead.org>,
        <zhangshaokun@hisilicon.com>, <pc@us.ibm.com>,
        <huanglingyan2@huawei.com>
References: <1617791570-165223-1-git-send-email-john.garry@huawei.com>
 <YG7xzXjFzIr9lQlz@krava>
Message-ID: <8b3e6667-ac48-9ca9-6645-b2eebbc828aa@huawei.com>
Date:   Tue, 13 Apr 2021 10:45:44 +0100
User-Agent: Mozilla/5.0 (Windows NT 10.0; WOW64; rv:68.0) Gecko/20100101
 Thunderbird/68.1.2
MIME-Version: 1.0
In-Reply-To: <YG7xzXjFzIr9lQlz@krava>
Content-Type: text/plain; charset="utf-8"; format=flowed
Content-Language: en-US
Content-Transfer-Encoding: 7bit
Precedence: bulk

On 08/04/2021 13:06, Jiri Olsa wrote:
> perf stat --topdown is not supported, as this requires the CPU PMU to
> expose (alias) events for the TopDown L1 metrics from sysfs, which arm
> does not do. To get that to work, we probably need to make perf use the
> pmu-events cpumap to learn about those alias events.

Hi guys,

About supporting --topdown command for other archs apart from x86, it 
seems not possible today. Support there is based on kernel support for 
"topdown" CPU events used in the metric calculations. However, arm64, 
for example, does not support these "topdown" events.

It seems to me that we can change to use pmu-events framework + 
metricgroup support here, rather than hardcoded events - has anyone 
considered this approach previously? Seems a pretty big job, so thought 
I'd ask first ...

Thanks,
John