Linked by diegocg on Sun 13th May 2012 23:48 UTC
Linux Lennart Poettering, the author of systemd, has announced: "I just put a first version of a wiki document together that lists a couple of easy optimizations to get your boot times down to [less than] 2s. It also includes a list of suggested things to hack on to get even quicker boot-ups."
Thread beginning with comment 518000
To view parent comment, click here.
To read all comments associated with this story, please click here.
Alfman
Member since:
2011-01-28

Alot of people here think boot time doesn't matter. I know hibernation makes boot times less frequent and less relevant to some people. However sleep/hibernation modes represent a great deal of complexity between OS/BIOS/hardware, and is frequently the cause of driver bugs. If turning on a computer was as fast as waking it up from sleep, then it might eventually enable a OS to do away with the ugly complexities of hibernation.

In my opinion one should be able to turn on and use a computer much like they turn on and use a TV - only waiting for the display to "warm up".

Reply Parent Score: 2

_txf_ Member since:
2008-03-17

If turning on a computer was as fast as waking it up from sleep, then it might eventually enable a OS to do away with the ugly complexities of hibernation.


Hibernation has other benefits like being able to keep all your programs/documents opened. Even if you have a session manager that saves state on shutdown, it still needs to reopen everything one by one as opposed to loading an image into ram.

Reply Parent Score: 4

Hypnos Member since:
2008-11-19

Does sleep/hibernation support in drivers really add complexity, or just require drivers to be written correctly?

Also, saving the state of the 30+ terminal and app windows I have open is not simple for my desktop -- no desktop I've ever used has done it correctly.

Reply Parent Score: 2

Alfman Member since:
2011-01-28

Hypnos,

"Does sleep/hibernation support in drivers really add complexity, or just require drivers to be written correctly?"

It's both actually.

Unless the system is placed in light sleep where peripherals continue to draw power, they will have to be reinitialized upon restart, but now we need new mechanisms to ensure the OS state that was saved to disk upon hibernation can be fully restored. This kind of state synchronization is far from trivial, especially when there are physical bus changes between hibernation sessions like USB devices being changed around.



"Also, saving the state of the 30+ terminal and app windows I have open is not simple for my desktop -- no desktop I've ever used has done it correctly."

Yeah unfortunately most applications and operating systems weren't designed to enable applications to save and restore their session state. I have read about an OS/API that does it though, despite searching I wasn't able to find it's name again though.

Hibernation adds shutdown delays, although these are less annoying that bootup delays. In theory though, a well tuned normal bootup should beat a hibernation bootup because hibernation saves and restores fragmented ram, which is wasteful.

Reply Parent Score: 2

Delgarde Member since:
2008-08-19

Does sleep/hibernation support in drivers really add complexity, or just require drivers to be written correctly?


Yes, it does add complexity. Because when you have a device completely powered down, it's no longer maintaining state - e.g the graphics chip no longer knows what mode it should be in, the contents of video memory are gone, etc.

Which means that when it wakes up again, the video driver has to put it back into the state it was in before. Re-initialise the hardware (effectively a bootup sequence for the GPU), switch to the right mode, then ensure that userspace does a redraw to finish things.

Reply Parent Score: 3