Dynamic Triple Buffering Patch Updated For GNOME 45

[phoronix]

Phoronix: Dynamic Triple Buffering Patch Updated For GNOME 45

GNOME 45 released last week and while it has many interesting desktop improvements, a feature still not found upstream is the Canonical-led work on dynamic triple buffering for Mutter...

https://www.phoronix.com/news/GNOME-45-Triple-Buffering

[binarybanana]

This could be interesting if implemented at the driver level. Double buffer by default and switch to triple buffer when the hardware can't keep up with display refresh. Probably comes with its own can of worms for clients that care about exact presentation timing, though. But maybe not that many in practice.

[szymon_g]

it's hard to measure it, but the ubuntu's version of gnome that include that patch is indeed more responsive when compared to the vanilla version. i hope it will get upstreamed quickly

[drake23]

Not invented here syndrome :/

[timofonic]

GNOME and their eternally long bureaucratic processes are amazing! Unless you are from IBMHat, of course.

[mxan]

Knew it wouldn’t be too long before a comment mentioning IBM / Red Hat showed up… truth is, this patch hasn’t made it in because 1) it’s a workaround for a bug in Intel’s driver. Why should GNOME be responsible for making Intel’s driver work? 2) it introduces multiple bugs, just check the merge request thread.

It has absolutely nothing to do with the fact it comes from a Canonical dev. Daniel has many other patches in GNOME that have been accepted without trouble.

By the way, most of the visible contributions to GNOME lately have come from just about everyone *except* Red Hat. Purism, Endless, Collabora, Igalia etc… but of course, Phoronix’s usual trolls won’t stop their ignorance

[Myownfriend]

Not invented here syndrome :/

That's not remotely what's happening. The same guy has successful merged other patches in Gnome. I suggest actually following the discussions in the Mutter merge request instead of jumping to conclusions without any knowledge of the situation.

[Myownfriend]

GNOME and their eternally long bureaucratic processes are amazing! Unless you are from IBMHat, of course.

The most recent bug reported in the MR thread was that it causes the whole desktop to freeze when watching a full screen video in Firefox and it doesn't look to be resolved yet and the user isn't even seeing anything in the logs that could help. Gnome's process isn't the reason the patch is buggy...and bugs do come up often. Daniel himself brings up most of them.

[user1]

it's hard to measure it, but the ubuntu's version of gnome that include that patch is indeed more responsive when compared to the vanilla version. i hope it will get upstreamed quickly

Do you have Intel iGPU? Because I have a discrete AMD GPU and I notice zero difference between having this patch and not having it. In order to make Gnome animations 100% smooth, in both cases I need to set "performance" power profile.

[You-]

This could be interesting if implemented at the driver level.

Its a workaround to get around a driver-level problem, so it wont be fixed like this in the driver.

The patch is to force the driver to do more work than is necessary to stop it going into a too low power state because it thinks it doesnt have enough work to continue current power state.

The driver level fix would be to simply NOT power down that far, not implement triple buffering.