Keep the GPU-Memory Running
===========================
Pro:
* Only need to set a kernel module parameter
Con:
* Only works in S2
* Uses more power than necessary (at least if more than 256 MB graphic memory are used)
* Does not work with suspend-to-disk (at least if I understood things right)
Copy Memory to tmpfs
====================
Dump the graphic-memory into tmpfs. I wouldn't suggest using /run, but creating a driver-specific tmpfs that is large enough to hold the graphic memory.
Pro:
* Just works, as the kernel takes care of paging the tmpfs to disk in case of stand-by or suspend-to-disk.
Con:
* Delayed suspend and wakeup
* Unexpected swap-usage -> OOM-Killer might kick in on suspend if there is not enough or no swap.
Get everyone to work around it
==============================
Get all the affected applications to update their video-memory after suspend-to-X.
Pro:
* Does not have any of the Cons of the other solutions
Con:
* Probably very tedious.
Keep it as it is
================
Pro:
* Minimal effort
* Kind of works most of the time.
I think we have some options to fix this. Yet, which would be a solution that could be packaged? If I understand https:/ /download. nvidia. com/XFree86/ Linux-x86_ 64/510. 54/README/ powermanagement .html correctly there are the following options:
Keep the GPU-Memory Running ======= ======= ======
=======
Pro:
* Only need to set a kernel module parameter
Con:
* Only works in S2
* Uses more power than necessary (at least if more than 256 MB graphic memory are used)
* Does not work with suspend-to-disk (at least if I understood things right)
Copy Memory to tmpfs ======= ======
=======
Dump the graphic-memory into tmpfs. I wouldn't suggest using /run, but creating a driver-specific tmpfs that is large enough to hold the graphic memory.
Pro:
* Just works, as the kernel takes care of paging the tmpfs to disk in case of stand-by or suspend-to-disk.
Con:
* Delayed suspend and wakeup
* Unexpected swap-usage -> OOM-Killer might kick in on suspend if there is not enough or no swap.
Get everyone to work around it ======= ======= ======= ==
=======
Get all the affected applications to update their video-memory after suspend-to-X.
Pro:
* Does not have any of the Cons of the other solutions
Con:
* Probably very tedious.
Keep it as it is
================
Pro:
* Minimal effort
* Kind of works most of the time.
Con:
* Nobody is really happy.
Comments?