yank555 wrote:About the wake lag, it's obvious statements are generally inconclusive...
My guess is this :
- wakeup lag only happens when the device is really in deep sleep (sounds obvious, but seems not that obvious to testers as to when a device is in deep sleep, which is not as soon as screen is turned off !)
- testing a new ROM means giving it time (aka. let it settle in) to build up caches, like media scanner, while this runs, phone won't enter deep sleep as it's potentially actively busy doing stuff
- turning screen off, waiting a few seconds, turning it on is not enought to enter deep sleep
- some apps keep wakelocks, thus preventing deep sleep (features like touch2wake do this as well, for as long as you set the delay !!)
- consider that using zram will slowen wakeup, why, because the tasks unneeded during screen off (the interactive ones) will end up getting swapped out, and waking the device will first need to swap back in what's needed for interactive use of the device
- some governor settings will also slow down wakeup, like including a screen off CPU max. frequency, even though zzmoove will kick the CPU's butt, as in all 4 cores go online and freq. goes to the max., as soon as it can on lateresume (=wakeup), we need to get to there and that will take time a little amount of time, as in lag
So to be conclusive, tests need to confirm the device was in actual deep sleep and need to consider other factors's influence on it.
And I stand by my assumption, wakeup lag is inherent to the hardware design of the Exynos platform once it's in actual deep sleep, as I never saw this on a Qualcomm device for instance (my htc Sensation and my Note 3 don't suffer from it).
JP.
The part with deep sleep is right on the money - a lot of people don't realize that sometimes their phone wakes-up immediately since it was not in deep sleep and then some time later no longer wakes-up so fast since that time it went to full deep sleep.
Now regarding the (deep-sleep) cause of the "wakeup lag" - the things that need to be remembered:
- 4.1.2 was also going into some form of deep-sleep and there was almost zero "wakeup lag"
- my personal guess is that now in 4.3 there is extra code executing - for instance to save the GPU state, put the GPU into a deeper sleep and restore the GPU state; since now all display seems to be done with the 3D composer that might add some delay, but IMHO the delay should NOT be anywhere near what we see, so there is still a bug somewhere there
- another factor that might or might not come into play - Samsung might also be making some extra-paranoid security-checks when coming back on the lockscreen (for the Knox part, which was not itself included here, but this attachment point might have been already included somewhere in one of the main Samsung frameworks in the 4.3 ROM)