[KERNEL] Moxie N7 / M12

Status
Not open for further replies.

bedalus

Senior Member
Jan 2, 2015
330
1,071
0
Chester
Here we go...

Download: here

Version N6 [Nougat]
-Linaro 6.3.1
-Mainline to .48 final
-optimised config (removed a lot of debug stuff)
-a few other minor tweaks, see my git repo

Cheers!

EDIT: Okay, I forgot to mention that BLX is of course included, and Sultan's great tweaks that disable the capacitive buttons while the screen is in use. I haven't posted much recently, but I haven't been idle either. I'll post all my findings once I have some more free time :)
 
Last edited:

Dariusdd

Senior Member
Apr 23, 2011
175
75
48
Bucharest
Okay, I am not sure if I'm doing anything wrong here, but off-screen gestures stopped working after flashing your kernel.
I am not sure if I can provide logs for this? I mean the phone doesn't react. Fingerprint unlock is still working.
This is happening on OOS 4.1.0 rooted with SuperSU SR3.
LE: I took a logcat, but I'm not sure if it's going to help.
Live logcat on my pc didn't seem to record anything happening when I was trying to execute offscreen gestures.
 

Attachments

Last edited:

bedalus

Senior Member
Jan 2, 2015
330
1,071
0
Chester
Okay, I am not sure if I'm doing anything wrong here, but off-screen gestures stopped working after flashing your kernel.
I am not sure if I can provide logs for this? I mean the phone doesn't react. Fingerprint unlock is still working.
This is happening on OOS 4.1.0 rooted with SuperSU SR3.
LE: I took a logcat, but I'm not sure if it's going to help.
Live logcat on my pc didn't seem to record anything happening when I was trying to execute offscreen gestures.
Thanks, I'll investigate
 

bedalus

Senior Member
Jan 2, 2015
330
1,071
0
Chester
Download: here

Version N7
Changed since v6:
-Removed more debug stuff
-a few more minor tweaks, see my git repo
-fixed gestures

The gesture fix actually required a section of code to be commented out. This is why it took me so long to fix. I was actually confused because the last time OnePlus released source with broken gestures it was because there was some code commented out that had to be un-commented out! The opposite! I must have checked the code fifty times, and every time I saw it I thought, 'Oh yeah, I don't have any commented out code, so I'm good, it must be something else...'

Anyway, it wasn't a waste, because I spent a lot of times spiralling around the code and config, trying different possible solutions, and during that cycle a lot of other optimisations happened, so that's good news.

I've got some stats I want to publish too, but I'll get around to that later.
 

Dariusdd

Senior Member
Apr 23, 2011
175
75
48
Bucharest
Colour me surprised, as this may be a one-off score, but running Geekbench 4 on OB13 (even if it's not officially supported) with your kernel results in a higher score than the stock kernel (which I don't remember happening before on Android N).
Does this mean you've found the cause of the lethargic Android N performance you've had issues with in the beginning, or is it related to the stuff you did behind scenes?
 

bedalus

Senior Member
Jan 2, 2015
330
1,071
0
Chester
Colour me surprised, as this may be a one-off score, but running Geekbench 4 on OB13 (even if it's not officially supported) with your kernel results in a higher score than the stock kernel (which I don't remember happening before on Android N).
Does this mean you've found the cause of the lethargic Android N performance you've had issues with in the beginning, or is it related to the stuff you did behind scenes?
I've managed to get a lot of performance back, but I'm really struggling to find the source of the huge drop in stdlib memmove and memcpy operations. Scores are about half that of Marshmallow. At this point I feel like I've tried everything, so perhaps the issue is more to do with the way the new art/dalvic compiler works. At least the scores are equivalent, or slightly improved vs stock. I don't think I'll return to Marshmallow again though. It's not as though I'm noticing any performance problems in day-to-day use.
 

Stroky

Senior Member
Jun 26, 2013
530
180
63
So the configuration of the Kernel like Governor etc are equal to stock?
Will flash the kernel the next days.
Back on marshmallow it was really great.
Thanks for your work!
 
  • Like
Reactions: bedalus

Mostafa Wael

Recognized Contributor
Jan 11, 2013
6,106
5,564
0
22
Gotham
I've managed to get a lot of performance back, but I'm really struggling to find the source of the huge drop in stdlib memmove and memcpy operations. Scores are about half that of Marshmallow. At this point I feel like I've tried everything, so perhaps the issue is more to do with the way the new art/dalvic compiler works...
Exactly. Nothing to do about it.
With OB13 fixing the infamous issue of Touch Latency, I may get back here and give your kernel a spin. Missing it actually.
The thing is VertexOS spoilt me with EAS and every couple of days I witness some huge leaps in how EAS works, making it better and more efficient.
Also curious to know about your BLX
Cheers!

-TDK
 
  • Like
Reactions: Stroky and Miksel12

bedalus

Senior Member
Jan 2, 2015
330
1,071
0
Chester
So the configuration of the Kernel like Governor etc are equal to stock?
Will flash the kernel the next days.
Back on marshmallow it was really great.
Thanks for your work!
Yes mate, just like stock. So far I haven't seen the need to meddle with the CPU governors or hot-plugging.
 
  • Like
Reactions: Stroky

bedalus

Senior Member
Jan 2, 2015
330
1,071
0
Chester
Exactly. Nothing to do about it.
With OB13 fixing the infamous issue of Touch Latency, I may get back here and give your kernel a spin. Missing it actually.
The thing is VertexOS spoilt me with EAS and every couple of days I witness some huge leaps in how EAS works, making it better and more efficient.
Also curious to know about your BLX
Cheers!

-TDK
I'll have to take a look through their commits and see if there's anything I could make use of. Thanks for the heads up.

BLX just limits the charge to a lower level so the battery won't last a long on any given day, but will stay healthier in the long run.
 
  • Like
Reactions: Mostafa Wael

bedalus

Senior Member
Jan 2, 2015
330
1,071
0
Chester
Here's the stats:


I've said before that I've had to run benchmarks many times over to get a reliable average. This time I was very concerned that the stock kernel was still out-performing my own build in stdlib copy operations. I tested stock vs Moxie 30 times. There were I think four or five very odd results (either way too high or way too low). I've removed the outliers and graphed the remaining results from highest to lowest. As you can see, there's no significant difference between the two sets of results.
 
Status
Not open for further replies.