On Wed, May 13, 2020 at 10:15:25AM +0100, Emil Velikov wrote:
On Tue, 12 May 2020 at 19:47, Thomas Zimmermann tzimmermann@suse.de wrote:
Hi
Am 12.05.20 um 12:14 schrieb Emil Velikov:
Hi Thomas,
On Tue, 12 May 2020 at 09:43, Thomas Zimmermann tzimmermann@suse.de wrote:
The suspend/resume helpers are unused. Also remove associated state from struct mga_device.
Although DPMS in it's traditional sense is no longer a thing, would it make sense to keep this around for documentation purposes? In particular, the pci magic and associated PLL/DAC/pixel clock could be used for dynamic PM.
That patch is not about DPMS. The DPMS code is still there. The suspend/resume helpers were outcommented and I don't know if they ever worked. Let's remove them.
Seems like the idea is to suspend/resume the device on DPMS off/on. A rather moot point IMHO. As the DPMS semantics and the whole modeset, got more subtle with atomic modeset, the idea gets even more moot.
With atomic it's actually a lot easier to do runtime pm in your modesetting code, since the flow is a lot more structured. There's even a helper function for that with drm_atomic_helper_commit_tail_rpm, since the default is optimized for max compatability with old legacy helpers. Maybe we should switch that around actually. -Daniel
If the documentation has that process - sure nuke it. Although for dynPM, this code is essential. Considering a) one has interest in dynPM and b) the code is (close to) working.
The last two are very big ifs so I'll leave it there.
-Emil