Received-SPF: pass (google.com: domain of linux-kernel+bounces-117088-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:45d1:ec00::1 as permitted sender) client-ip=2604:1380:45d1:ec00::1;
From: =?UTF-8?q?Ilpo=20J=C3=A4rvinen?= <ilpo.jarvinen@linux.intel.com>
Date: Mon, 25 Mar 2024 15:08:09 +0200 (EET)
To: Reinette Chatre <reinette.chatre@intel.com>
cc: linux-kselftest@vger.kernel.org, Shuah Khan <shuah@kernel.org>, 
    Babu Moger <babu.moger@amd.com>, 
    =?ISO-8859-15?Q?Maciej_Wiecz=F3r-Retman?= <maciej.wieczor-retman@intel.com>, 
    Fenghua Yu <fenghua.yu@intel.com>, LKML <linux-kernel@vger.kernel.org>
Subject: Re: [PATCH v2 02/13] selftests/resctrl: Calculate resctrl FS derived
 mem bw over sleep(1) only
In-Reply-To: <0910f9ed-2312-46b7-9c64-2982709da3d8@intel.com>
Message-ID: <10baa8db-3143-fdd4-49a8-0298db90cc4f@linux.intel.com>
References: <20240311135230.7007-1-ilpo.jarvinen@linux.intel.com> <20240311135230.7007-3-ilpo.jarvinen@linux.intel.com> <fe5c0d10-a57c-4a3a-ae30-a7cfa93bc3e8@intel.com> <f214c635-500f-43ea-fce8-0a7083bc1606@linux.intel.com>
 <0910f9ed-2312-46b7-9c64-2982709da3d8@intel.com>
Precedence: bulk
MIME-Version: 1.0
Content-Type: multipart/mixed; boundary="8323328-1681402267-1711372089=:1020"

  This message is in MIME format.  The first part should be readable text,
  while the remaining parts are likely unreadable without MIME-aware tools.

--8323328-1681402267-1711372089=:1020
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: QUOTED-PRINTABLE

On Fri, 22 Mar 2024, Reinette Chatre wrote:
> On 3/22/2024 5:11 AM, Ilpo J=C3=A4rvinen wrote:
> > On Tue, 19 Mar 2024, Reinette Chatre wrote:
> >> On 3/11/2024 6:52 AM, Ilpo J=C3=A4rvinen wrote:

> >>> + * Return: =3D 0 on success. < 0 on failure.
> >>> + */
> >>> +static int get_mem_bw_imc(char *bw_report, float *bw_imc)
> >>> +{
> >>> +=09float reads, writes, of_mul_read, of_mul_write;
> >>> +=09int imc, j;
> >>> +
> >>> +=09/* Start all iMC counters to log values (both read and write) */
> >>> +=09reads =3D 0, writes =3D 0, of_mul_read =3D 1, of_mul_write =3D 1;
> >>> =20
> >>>  =09/*
> >>>  =09 * Get results which are stored in struct type imc_counter_config
> >=20
> >>> @@ -696,7 +725,6 @@ int resctrl_val(const struct resctrl_test *test,
> >>>  =09=09struct resctrl_val_param *param)
> >>>  {
> >>>  =09char *resctrl_val =3D param->resctrl_val;
> >>> -=09unsigned long bw_resc_start =3D 0;
> >>
> >> In the current implementation the first iteration's starting measureme=
nt
> >> is, as seen above, 0 ... which makes the first measurement unreliable
> >> and dropped for both the MBA and MBM tests. In this enhancement, the
> >> first measurement is no longer skewed so much so I wonder if this enha=
ncement
> >> can be expanded to the analysis phase where first measurement no longe=
r
> >> needs to be dropped?
> >=20
> > In ideal world, yes, but I'll have to check the raw numbers. My general=
=20
> > feel is that the numbers tend to converge slowly with more iterations=
=20
> > being run so the first iteration might still be "off" by quite much (th=
is=20
> > is definitely the case with CAT tests iterations but I'm not entirely s=
ure=20
> > any more how it is with other selftests).
>=20
> >From what I can tell the CAT test is not dropping any results. It looks
> to me that any "settling" is and should be handled in the test before
> the data collection starts.

It doesn't, but the "settling" is there in the raw numbers. I've=20
considered adding warm-up test(s) before the actual runs to improve the=20
situation but there's just so many thing still to do...

--=20
 i.

--8323328-1681402267-1711372089=:1020--