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:   Song Liu <songliubraving@fb.com>
To:     Arnaldo Carvalho de Melo <acme@kernel.org>
CC:     Namhyung Kim <namhyung@kernel.org>,
        linux-kernel <linux-kernel@vger.kernel.org>,
        Kernel Team <Kernel-team@fb.com>,
        "Arnaldo Carvalho de Melo" <acme@redhat.com>,
        Jiri Olsa <jolsa@kernel.org>
Subject: Re: [PATCH v2 0/3] perf-stat: share hardware PMCs with BPF
Thread-Topic: [PATCH v2 0/3] perf-stat: share hardware PMCs with BPF
Thread-Index: AQHXGqn3m2pRtlCATE6KZ0qUM3kt26qHp6sAgACBEACAAPZFAA==
Date:   Thu, 18 Mar 2021 03:52:51 +0000
Message-ID: <7D48A756-C253-48DE-B536-826314778404@fb.com>
References: <20210316211837.910506-1-songliubraving@fb.com>
 <CAM9d7ci=hfFjq3+XuBcCZ0TUJxv6AmdFk0dkHFQD3wx27aJMjA@mail.gmail.com>
 <YFH//FRPvrPswhld@kernel.org>
In-Reply-To: <YFH//FRPvrPswhld@kernel.org>
Accept-Language: en-US
Content-Language: en-US
Content-Type: text/plain; charset="us-ascii"
Content-ID: <D6E86289470C1140B35F0A8A9A01A299@namprd15.prod.outlook.com>
Content-Transfer-Encoding: quoted-printable
MIME-Version: 1.0
X-MS-Exchange-CrossTenant-Network-Message-Id: 9fd62e1e-9ee4-4a07-c221-08d8e9c14d9f
X-MS-Exchange-CrossTenant-originalarrivaltime: 18 Mar 2021 03:52:51.1981
 (UTC)
X-MS-Exchange-CrossTenant-fromentityheader: Hosted
X-MS-Exchange-CrossTenant-id: 8ae927fe-1255-47a7-a2af-5f3a069daaa2
X-MS-Exchange-CrossTenant-mailboxtype: HOSTED
X-MS-Exchange-CrossTenant-userprincipalname: sDFgBVGcuCCQRVjthtEIuoCq8wMayuTPEiIWWRSeGgMEQX0SchnpFdHs/0ZvpODiuQazWof7Rpfh+VVZZNYs3Q==
X-MS-Exchange-Transport-CrossTenantHeadersStamped: BYAPR15MB2405
X-OriginatorOrg: fb.com
X-Proofpoint-Virus-Version: vendor=fsecure engine=2.50.10434:6.0.369,18.0.761
 definitions=2021-03-18_01:2021-03-17,2021-03-18 signatures=0
X-Proofpoint-Spam-Details: rule=fb_default_notspam policy=fb_default score=0 phishscore=0
 mlxlogscore=999 suspectscore=0 malwarescore=0 impostorscore=0 mlxscore=0
 spamscore=0 clxscore=1015 bulkscore=0 adultscore=0 priorityscore=1501
 lowpriorityscore=0 classifier=spam adjust=0 reason=mlx scancount=1
 engine=8.12.0-2009150000 definitions=main-2103180028
X-FB-Internal: deliver
Precedence: bulk
List-ID: <linux-kernel.vger.kernel.org>
X-Mailing-List: linux-kernel@vger.kernel.org


> On Mar 17, 2021, at 6:11 AM, Arnaldo Carvalho de Melo <acme@kernel.org> w=
rote:
>=20
> Em Wed, Mar 17, 2021 at 02:29:28PM +0900, Namhyung Kim escreveu:
>> Hi Song,
>>=20
>> On Wed, Mar 17, 2021 at 6:18 AM Song Liu <songliubraving@fb.com> wrote:
>>>=20
>>> perf uses performance monitoring counters (PMCs) to monitor system
>>> performance. The PMCs are limited hardware resources. For example,
>>> Intel CPUs have 3x fixed PMCs and 4x programmable PMCs per cpu.
>>>=20
>>> Modern data center systems use these PMCs in many different ways:
>>> system level monitoring, (maybe nested) container level monitoring, per
>>> process monitoring, profiling (in sample mode), etc. In some cases,
>>> there are more active perf_events than available hardware PMCs. To allo=
w
>>> all perf_events to have a chance to run, it is necessary to do expensiv=
e
>>> time multiplexing of events.
>>>=20
>>> On the other hand, many monitoring tools count the common metrics (cycl=
es,
>>> instructions). It is a waste to have multiple tools create multiple
>>> perf_events of "cycles" and occupy multiple PMCs.
>>=20
>> Right, it'd be really helpful when the PMCs are frequently or mostly sha=
red.
>> But it'd also increase the overhead for uncontended cases as BPF program=
s
>> need to run on every context switch.  Depending on the workload, it may
>> cause a non-negligible performance impact.  So users should be aware of =
it.
>=20
> Would be interesting to, humm, measure both cases to have a firm number
> of the impact, how many instructions are added when sharing using
> --bpf-counters?
>=20
> I.e. compare the "expensive time multiplexing of events" with its
> avoidance by using --bpf-counters.
>=20
> Song, have you perfmormed such measurements?

I have got some measurements with perf-bench-sched-messaging:

The system: x86_64 with 23 cores (46 HT)

The perf-stat command:
perf stat -e cycles,cycles,instructions,instructions,ref-cycles,ref-cycles =
<target, etc.>

The benchmark command and output:
./perf bench sched messaging -g 40 -l 50000 -t
# Running 'sched/messaging' benchmark:
# 20 sender and receiver threads per group
# 40 groups =3D=3D 1600 threads run
     Total time: 10X.XXX [sec]


I use the "Total time" as measurement, so smaller number is better.=20

For each condition, I run the command 5 times, and took the median of=20
"Total time".=20

Baseline (no perf-stat)			104.873 [sec]
# global
perf stat -a				107.887 [sec]
perf stat -a --bpf-counters		106.071 [sec]
# per task
perf stat 				106.314 [sec]
perf stat --bpf-counters 		105.965 [sec]
# per cpu
perf stat -C 1,3,5 			107.063 [sec]
perf stat -C 1,3,5 --bpf-counters 	106.406 [sec]

From the data, --bpf-counters is slightly better than the regular event
for all targets. I noticed that the results are not very stable. There=20
are a couple 108.xx runs in some of the conditions (w/ and w/o=20
--bpf-counters).


I also measured the average runtime of the BPF programs, with=20

	sysctl kernel.bpf_stats_enabled=3D1

For each event, if we have one leader and two followers, the total run=20
time is about 340ns. IOW, 340ns for two perf-stat reading instructions,=20
340ns for two perf-stat reading cycles, etc.=20

Thanks,
Song=