From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <igt-dev-bounces@lists.freedesktop.org>
Received: from mga18.intel.com (mga18.intel.com [134.134.136.126])
 by gabe.freedesktop.org (Postfix) with ESMTPS id 29BDC10E88C
 for <igt-dev@lists.freedesktop.org>; Thu, 19 Jan 2023 03:32:22 +0000 (UTC)
Date: Wed, 18 Jan 2023 19:32:19 -0800
Message-ID: <87o7qvurlo.wl-ashutosh.dixit@intel.com>
From: "Dixit, Ashutosh" <ashutosh.dixit@intel.com>
To: "Tauro, Riana" <riana.tauro@intel.com>
In-Reply-To: <bddffb35-4a95-58c5-473d-5904fb1dfc84@intel.com>
References: <20230117190317.17728-1-ashutosh.dixit@intel.com>
 <CY5PR11MB621110BD11315EEA8F6F69BC95C79@CY5PR11MB6211.namprd11.prod.outlook.com>
 <bddffb35-4a95-58c5-473d-5904fb1dfc84@intel.com>
MIME-Version: 1.0 (generated by SEMI-EPG 1.14.7 - "Harue")
Content-Type: text/plain; charset=US-ASCII
Subject: Re: [igt-dev] [PATCH i-g-t] i915/i915_power: Measure GPU idle/busy
 power
List-Unsubscribe: <https://lists.freedesktop.org/mailman/options/igt-dev>,
 <mailto:igt-dev-request@lists.freedesktop.org?subject=unsubscribe>
List-Archive: <https://lists.freedesktop.org/archives/igt-dev>
List-Post: <mailto:igt-dev@lists.freedesktop.org>
List-Help: <mailto:igt-dev-request@lists.freedesktop.org?subject=help>
List-Subscribe: <https://lists.freedesktop.org/mailman/listinfo/igt-dev>,
 <mailto:igt-dev-request@lists.freedesktop.org?subject=subscribe>
Cc: "igt-dev@lists.freedesktop.org" <igt-dev@lists.freedesktop.org>, "Nilawar,
 Badal" <badal.nilawar@intel.com>
Errors-To: igt-dev-bounces@lists.freedesktop.org
Sender: "igt-dev" <igt-dev-bounces@lists.freedesktop.org>
List-ID: <igt-dev@lists.freedesktop.org>

On Tue, 17 Jan 2023 23:22:46 -0800, Tauro, Riana wrote:
>
> Hi Ashutosh
>

Hi Riana,

> Energy measurement when idle and with workload is also a part of rc6-idle
> test.  Why do we need new tests?

We want to be able to measure max/peak power.

rc6-idle has this:

static void bg_load()
{
	...
		/* aim for ~1% busy */

Therefore it measures this:

[adixit@DUT1220-DG2SVC igt-gpu-tools]$ sudo ./build/tests/i915_pm_rc6_residency --r rc6-idle
IGT-Version: 1.27-g369bb6fb0f (x86_64) (Linux: 5.15.47+ x86_64)
Opened device: /dev/dri/card0
Starting subtest: rc6-idle
Starting dynamic subtest: gt0-rcs0
Total energy used while idle: 85825.3mJ (28606.5mW)
Total energy used for normal: 86731.6mJ (28908.6mW)
Total energy used for boost: 86992.8mJ (28995.8mW)
Total energy used for once: 86677.9mJ (28890.5mW)
Dynamic subtest gt0-rcs0: SUCCESS (12.235s)

The new test measures this:

[adixit@DUT1220-DG2SVC igt-gpu-tools]$ sudo ./build/tests/i915_power
IGT-Version: 1.27-g369bb6fb0f (x86_64) (Linux: 5.15.47+ x86_64)
Opened device: /dev/dri/card0
Starting subtest: gpu-idle
Measured power: 28535.5 mW
Subtest gpu-idle: SUCCESS (3.002s)
Starting subtest: gpu-busy
Measured power: 50331.9 mW
Subtest gpu-busy: SUCCESS (3.008s)

So you can see rc6-idle basically measures idle power (28 W instead of 50
W), not power under load.

Other tests such as gem_exec_schedule also have igt_require etc. which skip
in all sorts of situations. That is why I was forced to write these simple
new tests.

Thanks.
--
Ashutosh