[PATCH] drm/xe/d3cold: Set power state to D3Cold during s2idle/s3
Rafael J. Wysocki
rafael at kernel.org
Wed Apr 2 10:19:11 UTC 2025
On Wed, Apr 2, 2025 at 9:31 AM Raag Jadav <raag.jadav at intel.com> wrote:
>
> On Tue, Apr 01, 2025 at 09:35:49PM +0200, Rafael J. Wysocki wrote:
> > On Tue, Apr 1, 2025 at 7:53 PM Raag Jadav <raag.jadav at intel.com> wrote:
>
[skip]
>
> I'm still trying to wrap my head around the total suspend duration argument.
> Do you mean the time taken by xe_pci_suspend() itself can change with
> ->suspend_late()?
No, I don't think so.
> Or are there any other side-effects to it?
Yes, there are.
So currently in the mainline xe_pci_suspend() runs in parallel with
other drivers' ->suspend() callbacks, so the total time taken by all
of them is roughly equal to the time taken by the slowest one.
If xe_pci_suspend() is made run in a different phase of the
transition, it will not run in parallel with all of the other
callbacks, so the time taken by it will become a separate component of
the total suspend time.
More information about the Intel-xe
mailing list