From: "Adrián Larumbe" <adrian.larumbe@collabora.com>
To: Steven Price <steven.price@arm.com>
Cc: "Adrián Larumbe" <adrian.larumbe@collabora.com>,
"Boris Brezillon" <boris.brezillon@collabora.com>,
"Rob Herring" <robh@kernel.org>,
"Maarten Lankhorst" <maarten.lankhorst@linux.intel.com>,
"Maxime Ripard" <mripard@kernel.org>,
"Thomas Zimmermann" <tzimmermann@suse.de>,
"David Airlie" <airlied@gmail.com>,
"Simona Vetter" <simona@ffwll.ch>,
"Faith Ekstrand" <faith.ekstrand@collabora.com>,
"Marty E. Plummer" <hanetzer@startmail.com>,
"Tomeu Vizoso" <tomeu@tomeuvizoso.net>,
"Eric Anholt" <eric@anholt.net>,
"Robin Murphy" <robin.murphy@arm.com>,
"Philipp Zabel" <p.zabel@pengutronix.de>,
dri-devel@lists.freedesktop.org, linux-kernel@vger.kernel.org,
"Collabora Kernel Team" <kernel@collabora.com>,
"Neil Armstrong" <neil.armstrong@linaro.org>
Subject: Re: [PATCH v12 12/15] drm/panfrost: Skip cache flush/invalidate when enabling perfcnt
Date: Tue, 06 Oct 2026 17:45:19 +0100 [thread overview]
Message-ID: <179130511905.1018806.3717655233851799587.b4-reply@b4> (raw)
In-Reply-To: <95174a67-0eee-4bf7-b883-9f1138c74a6f@arm.com>
On 2026-10-02 16:14:38+01:00, Steven Price wrote:
> On 29/09/2026 04:44, Adrián Larumbe wrote:
>
> > The GPU cache flush/invalidate operation is unnecessary. First off, the
> > GPU doesn't read off the perfcnt sample buffer, only writes into it, so
>
> I don't think this is entirely true. The GPU performance counter unit
> only writes the counters that are enabled, counters that share a cache
> line but are not enabled are not written by the performance counter
> unit, but if the L2 contains that cache line then the write can hit in
> the L2 and dirty the entire line including stale data where the
> unwritten cache line is.
From the perspective of UM, is it bad if offsets into the perfcnt buffer
that stand for counters they did not enable contain undefined values?
> The upshot is that if the CPU has cleared a block of memory which the
> GPU happens to have cached, then the "unused" counters may end up
> showing the old data before the CPU cleared it (if they share a cache
> line with an active counter).
>
> I have to admit it's probably somewhat academic given that Panfrost
> doesn't expose the ability to control which counters are enabled...
>
> Is there a good reason for this patch (i.e. have you seen a performance
> problem with doing the invalidate)? Otherwise I'd prefer we keep to the
> safe route rather than trying to over optimise cache maintenance.
Like Boris said in a later reply, it was mostly about simplifying error handling
in the perfcnt enable path. We saw this initial clean being done and thought it
wasn't necessary.
> Obviously in the fully coherent case the invalidate could be skipped (as
> in the next patch).
>
> Thanks,
> Steve
next prev parent reply other threads:[~2026-10-06 16:46 UTC|newest]
Thread overview: 48+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-09-29 3:44 [PATCH v12 00/15] Collection of fixes for Panfrost: Perfcnt, RPM, refactorings Adrián Larumbe
2026-09-29 3:44 ` [PATCH v12 01/15] drm/panfrost: Move shrinker initialization and unplug one level down Adrián Larumbe
2026-10-02 13:58 ` Steven Price
2026-09-29 3:44 ` [PATCH v12 02/15] drm/panfrost: Move lock and modparam initialisations into their subsystems Adrián Larumbe
2026-10-02 14:06 ` Steven Price
2026-09-29 3:44 ` [PATCH v12 03/15] drm/panfrost: Move debugfs initialisation to relevant subsystems Adrián Larumbe
2026-10-02 14:10 ` Steven Price
2026-10-06 0:50 ` Adrián Larumbe
2026-09-29 3:44 ` [PATCH v12 04/15] drm/panfrost: Skip NULL checks for clock enable/disabling Adrián Larumbe
2026-10-02 14:14 ` Steven Price
2026-09-29 3:44 ` [PATCH v12 05/15] drm/panfrost: Consolidate device clock management and reset Adrián Larumbe
2026-10-02 14:20 ` Steven Price
2026-09-29 3:44 ` [PATCH v12 06/15] drm/panfrost: Fix PM refcnt and autosuspend issues at device probe/remove Adrián Larumbe
2026-10-02 14:21 ` Steven Price
2026-09-29 3:44 ` [PATCH v12 07/15] drm/panfrost: Explicitly enable MMU interrupts at device init Adrián Larumbe
2026-10-02 14:26 ` Steven Price
2026-09-29 3:44 ` [PATCH v12 08/15] drm/panfrost: Move all DRM device initialisation into device_init() Adrián Larumbe
2026-10-02 14:37 ` Steven Price
2026-09-29 3:44 ` [PATCH v12 09/15] drm/panfrost: Add warning messages to fatal error conditions Adrián Larumbe
2026-10-02 14:59 ` Steven Price
2026-10-06 15:09 ` Adrián Larumbe
2026-10-07 12:51 ` Steven Price
2026-09-29 3:44 ` [PATCH v12 10/15] drm/panfrost: Add debugfs knob for manually triggering a GPU reset Adrián Larumbe
2026-10-02 15:10 ` Steven Price
2026-10-06 13:38 ` Adrián Larumbe
2026-10-07 12:54 ` Steven Price
2026-09-29 3:44 ` [PATCH v12 11/15] drm/panfrost: Move perfcnt GPU disable sequence into a helper Adrián Larumbe
2026-09-29 3:44 ` [PATCH v12 12/15] drm/panfrost: Skip cache flush/invalidate when enabling perfcnt Adrián Larumbe
2026-10-02 15:14 ` Steven Price
2026-10-05 8:38 ` Boris Brezillon
2026-10-05 15:06 ` Steven Price
2026-10-05 15:37 ` Boris Brezillon
2026-10-05 16:05 ` Steven Price
2026-10-05 16:57 ` Boris Brezillon
2026-10-06 17:42 ` Adrián Larumbe
2026-10-06 17:35 ` Adrián Larumbe
2026-10-07 13:28 ` Steven Price
2026-10-06 17:18 ` Adrián Larumbe
2026-10-07 13:21 ` Steven Price
2026-10-06 16:45 ` Adrián Larumbe [this message]
2026-10-07 13:09 ` Steven Price
2026-09-29 3:44 ` [PATCH v12 13/15] drm/panfrost: Avoid cache flush after perfcnt sample in fully coherent systems Adrián Larumbe
2026-10-02 15:28 ` Steven Price
2026-09-29 3:44 ` [PATCH v12 14/15] drm/panfrost: Introduce a reset lock Adrián Larumbe
2026-10-02 15:34 ` Steven Price
2026-09-29 3:44 ` [PATCH v12 15/15] drm/panfrost: Fix races between perfcnt and reset sequence Adrián Larumbe
2026-10-02 15:44 ` Steven Price
2026-10-06 14:14 ` Adrián Larumbe
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=179130511905.1018806.3717655233851799587.b4-reply@b4 \
--to=adrian.larumbe@collabora.com \
--cc=airlied@gmail.com \
--cc=boris.brezillon@collabora.com \
--cc=dri-devel@lists.freedesktop.org \
--cc=eric@anholt.net \
--cc=faith.ekstrand@collabora.com \
--cc=hanetzer@startmail.com \
--cc=kernel@collabora.com \
--cc=linux-kernel@vger.kernel.org \
--cc=maarten.lankhorst@linux.intel.com \
--cc=mripard@kernel.org \
--cc=neil.armstrong@linaro.org \
--cc=p.zabel@pengutronix.de \
--cc=robh@kernel.org \
--cc=robin.murphy@arm.com \
--cc=simona@ffwll.ch \
--cc=steven.price@arm.com \
--cc=tomeu@tomeuvizoso.net \
--cc=tzimmermann@suse.de \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox
all inboxes | Powered by JetHome®