Shared publicly  - 
 
If you're running CM10 on maguro and are tired of the slowness (because seriously, wtf is wrong with it, it's absurd, we need to fix it, I know) just flash the stock AOSP kernel.

http://android.git.r3pek.org/?p=device/samsung/tuna.git;a=blob_plain;f=kernel;hb=HEAD

I haven't tested how it affects battery life yet but I hope it fixes that (because batterylife sucks atm too, CM10 on gnexus, at least the gsm variant, is broken badly).
44
8
Jonas Hördegård's profile photoSteve C's profile photoDerek Rodriguez's profile photoKeyan Mobli's profile photo
91 comments
 
Does the AOSP kernel fix it? I was trying to do some perf testing between AOSP and CM10 on both maguro and toro and I couldn't find any differences in my tests. I didn't run AOSP long enough to see how it reacted in real-world experiences though.

Toro seems dog slow as well.
 
+Keyan Mobli Let me know how it is with some time. I'm all for stripping the kernel back down to AOSP's current one and only adding the few specific tweaks that we need if that fixes the issues. I never did agree with the massive changes that have been made (like upping the kernel revision WAY beyond what Google has done in AOSP).
 
Before I sold it, my maguro was super slow on CM10. I noticed it was better for a day or two after I'd cleared cache, but then it would slow down again. Let's hope the N4 with CM10 brings better luck.
 
can anyone link the stock kernel for Toro? I want to see if that fixes the slowness on there too.
 
I've also been asking myself wtf was going on. For a moment there I thought it was the manifestation of an underlying urge to upgrade to the Nexus 4. Glad I am not alone.
 
+Miguel Raimundo I had the same concept - but my fiance has been complaining about it too, and she doesn't want to upgrade ;-) +Justin Beeson I don't know what format +Keyan Mobli posted - I'm building my own test from source right now.
 
The kernel is the same for maguro/toro/toroplus? Curious if I'm safe playing guinea pig on a toroplus.
 
+Justin Shidell yes should be.  The current kernel for CM  is the same across the tuna family (maguro/toro/toroplus).
 
Do I flash it as usual via "fastboot flash boot c:\kernel"?
 
I'd like to see a .zip too. Can't fastboot with my busted USB port.
 
So using fastboot to flash +Keyan Mobli's file caused my Toro to get stuck in the bootloader. Tried another boot.img from a stock AOSP 4.1.2 rom and that hung on the boot animation. So anyone have this in a working boot.img format or flashable zip? +Brint Kriebel 
 
I'll have flashable zips of full CM10 systems w/ AOSP kernel for all three devices in a few minutes - just waiting on the last build to finish.
 
+Justin Beeson Ah ok. For future cases, use "fastboot boot" instead of "fastboot flash" to test an image prior to flashing it.
 
My sprint gnexus has been eating performance slugs as well. Acts like I got 512mb ram and single core. 
 
Honestly, rooted stock is my best experience for gnex speed & battery wise.
I only use cm10 on my old Galaxy S now.
 
Since we could probably use some more hands on this to see if there is a noticeable difference:
maguro: http://dropbox.bekit.net/Android/maguro/cm/cm-10-20121109-UNOFFICIAL-maguro.zip
toro: http://dropbox.bekit.net/Android/toro/cm/cm-10-20121109-UNOFFICIAL-toro.zip
toroplus*: http://dropbox.bekit.net/Android/toroplus/cm/cm-10-20121109-UNOFFICIAL-toroplus.zip

These are using CM10 sources synced from about an hour ago and a kernel built from AOSP source. The only change from the released AOSP 4.1.2 version of the kernel is enabling the version name to include "-cyanogenmod". Note: you will lose all ability to overclock, change the color tuning, mount network filesystems, and maybe some other stuff that is in the current CM kernel that I'm not aware of.

I don't know yet if this actually improves performance or not - please let me know what you think. I know for sure that it lowers benchmark scores like Quadrant - but I also know that benchmarks mean jack shit when it comes to actual experience. Please give it some use and let me know if this runs better, worse, or the same as a CM10 nightly from within the last few days.

*I have test flashed toro and maguro on top of recent CM10 nightlies. I don't have a toroplus to test with.
 
So far it has improved my batterly life considerably
 
Yessssss thank you! I was just about to decide to think about looking for a way to do something about this.
 
Can I flash this on Toro, or do I need to do some Google-fu?
 
I'd flash a +Brint Kriebel kernel... didn't bekit work on the 32a HTC Magic? Or am I misremembering my Android history.
 
+Nathan Olsen Original Droid, not the HTC Magic ;-)

And I can't take credit for this one - as I said, it's pure AOSP at the moment.
 
Man, I was wondering what was going on. Glad to see I wasn't the only one with sluggishisness. Thanks +Brint Kriebel for the new build. I'll test it out as well on my toroplus.
 
+Brint Kriebel l sorry, left this post open for a little bit before i commented... i thought i was post #5
 
+Dave Knepper make sure to report back. No changes have been made to the official source. I'll make them there only if we get consistent reports of improvement.
 
+Brint Kriebel no report on battery life just yet but overall smoothness seems to be more than just a placebo after the initial boot. Will try to remember to monitor the battery over the next day
 
+Brint Kriebel running really smooth so far. flashed right after you posted so its been an hour or so and usually i already have a bit of slow down already.
 
+Brint Kriebel Just fired it up and so far it IS quicker. Will report on battery tomorrow. Thanks for building this so we can test it. 
 
I flashed toro. Will report back tomorrow re:battery.
 
This comment is just me subscribing for future reference. Keep scrolling. 
 
Thx +Brint Kriebel. For future releases of CM10: Can someone create a flashable CWM zip of the aosp kernel? Then we can update on-the-run and reflash the kernel.
 
So, I've been running CM10 w/ AOSP-kernel (from the post by Brit) and my phone is way snappier now!
 
+Keyan Mobli That's good! The only thing I miss rom CM-kernel is color tuning, but I happily leave that for a faster phone!
 
+Keyan Mobli I had to rebase my GN kernel and start from scratch. No matter what, my old kernel ended up sluggish and drained the hell out of the battery. Starting from scratch and adding in certain things (color control, updating to 3.0.51, etc) helped it ten fold. Maybe you guys might consider that?
 
+Keyan Mobli Didn't know if you guys planned on just debugging and figuring out the cause or not. I know how much of a pain starting over can be haha
 
It seems faster. I'll comment more after it runs for awhile.
 
Running about 12hrs now and it's definitely smoother. Battery is at least the same if not better but need some more time running to be sure. I'm not entirely sure, but I was have poor WiFi signal issues and that seems to be better too. Can that be kernel related?
 
Definitely faster and doesn't slow down once it's been on for awhile. Battery seems about the same after about a half day of usage. (Toro)
 
I think it's kind of a mixed bag for me. It seems smoother in the UI, but I'm not 100% sure that it actually is.
I'm surprised that the quadrant scores are as far below "stock" Galaxy Nexus as they are...which is on both CM & AOSP, but I hadn't been paying any attention to those for quite a while. (I find the GNex to be fast enough for my needs) I also use the NOOP scheduler.
Well, I guess I'm looking forward to seeing the rebased kernel show up in the nightlies.
 
+Nathan Olsen Yeah, the quadrant scores are weird and, IMHO, worthless. I'm just ignoring them.

We've begun the process of switching the CM10 source over to use this base. We're adding back in things like the color adjustment options. Hopefully we'll see everything up and running on this new base within the next couple of nightlies.
 
Were y'all going to use the suggestion from +Derek Rodriguez and advance the kernel version and rebase at the same time?
I'm just curious if there are any particular kernel advances that are worth keeping an eye on as well.
 
+Nathan Olsen What I'm pushing for is starting over at the 4.1.2 AOSP kernel and only adding in specific commits for enhancements and fixes. No advancing of the kernel version unless there is an absolute reason for it.
 
Absolute { functionality, enhancements } || Absolute { functionality }

Sorry, that was stupid: Absolute meaning functionality alone or functionality and enhancements?
 
Absolute - mostly only provable enhancements or provable security flaws. In my opinion - most likely no version updates at all until Google does it first.
 
Anyone pushing for a branch that could evaluate kernel improvements in re: enhancements? 
Where have all the kernel devs gone? (the real ones)
 
Are we talking about the same thing? I already said that enhancements like color hack, etc. will be made. I just don't want us trying to up the version number from some kernel.org release for no provable reason.

And folks are welcome to make add-on kernels if they wish - but base CM should be stable, fast, reliable and as close to AOSP as possible while still providing some added benefit.
 
no, no, no... this isn't some kind of attack or push for anything specific, I like the direction you're encouraging. (my comment about real kernel devs is that some people like to tick every single check box and market it as some kind of "demon kernel" that will give ponies)

I was just wondering about how to evaluate whether a particular version tick is a true enhancement or not.

...and I can't remember who the dev is that used to do kernels for the G2 that were superior in power consumption, etc, but were usually just a minor tick above the official. (edit:@pershoot)

I'm going to back out now for a bit, I think my comments got a bit off-track.
 
My responsiveness and battery life have been much better after using this kernel all day. My friend's still on the CM10 kernel and his battery is around 50%, while mine is at 80%. We both have maguro.
Steve C
 
a CWM flashable of the AOSP Kernel would be terrific! Thanks to all who are helping out with this post!
 
+Steve C, they're working it into the nightlies. If you can't wait, +Brint Kriebel posted a link in this thread to a CWM flashable unofficial build of CM10 with the AOSP kernel included.
 
+Nathan Olsen upping to the mainline kernel can provide instabilities and all sorts of issues. Our kernel source is based on an altered version of the Linux kernel. Google, IMO, stopped at 3.0.31 for a reason (I believe 3.0.34 was the latest when 4.1 pre-release was pushed at I/O at the end of June). I upped to 3.0.51 mainly to see if there was something that caused my battery drain and instability from patching in mainline (had a similar problem when 3.0.18 came out, had to revert a commit).
 
Definitely faster and better battery life. maguro
 
The latest CM10 nightly (20121111) now includes the rebased kernel with some of the features added back in (colour tuning ability, etc.). Give it a shot and let us know how it works!
 
The toro 20121110 nightly has the update, too.
 
Don't thank me for those, that is all other people's work. I just coordinated the rebase. :-)
 
Does the nightly from today also include the stock kernel? I just ask cause the name differs from yesterday...
 
+Daniel Emile Quandel the kernel base has been reverted back to the original stock and then ONLY the few tweaks have been added back in. Namely the color modifications. My understanding is it will stay this way until more extensive testing is done on newer kernels. Future nightlies (at least for now) will have that AOSP kernel above.
 
+Daniel Emile Quandel The name will differ as we put some patches on the kernel. The kernel version is what shouldn't change too often (3.0.31). We won't be upping that version unless Google does, or unless there is a critical need to.
 
I still felt a little sluggish on my GNex with my upgrade to 10.0 Stable, since the stock kernel I found didn't seem to boot (30+ minutes on the boot animation) so I went with Franco's kernel and I think I'm seeing a marked boost in performance. I was hanging on certain actions, like coming out of sleep, unlocking, and even opening the app drawer.. I'm not seeing these hangups on Franco's kernel after letting it settle for 10 minutes, rebooting, and using it for 40 minutes or so.


 
+Beau Briggs If you are getting 30+ minutes on the boot animation you have some serious issues going on. Did you do a data wipe?
 
Yeah, those are not normal results +Beau Briggs. If you didn't data wipe as +Brint Kriebel mentioned make sure you give that a try. Very strange issues like those generally are solved best by starting clean.
 
+Lucas Robb CM10 stable and any current nightlies all contain the new kernel. I think the overwhelming opinion is that things have been dramatically faster since that point. If you haven't, I would recommend doing a data wipe (especially if you haven't since we went from 4.1.1. to 4.1.2). It is also possible that there is a specific app causing issues for you.

If anyone else is seeing slowness problems still, let me know.
 
+Lucas Robb You mean the CyanogenMod updater? It is actually perfectly fine to use most of the time (it's what I use). I tend to do a clean wipe every so often, though - especially when versions jump.
 
The 30 minutes on the boot animation (which I never let finish) was only when I tried a kernel labeled as a Stock Galaxy Nexus kernel... I figured that it was broken so I restored my nandroid backup and it cleared my boot animation problem, but I was still laggy in certain situations, about the same lagginess on 10.0 as the there was on M2...

After flashing Franco's kernel my phone feels much faster and doesn't hang on the places where my phone was hanging before. I read that you should wipe cache and dalvic when installing a new kernel, and that's all I've wiped since moving from stock-CM10nightlies-M1-M2 and now 10.0 stable. I'll try to wipe and see if that fixes my lag issues with the CM kernel. I don't know why Franco's kernel had the effect that I've observed these last couple days, should a cache wipe clear up some hanging issues with a kernel?
 
I don't think cache wiping should make much of a difference. If there is any repeatable process to produce lag, please let me know about it. It is certainly possible that Franco's kernel will perform faster for some instances - it has a bunch of changes including overclocking. The base CM kernel should not have those kinds of changes made to it, especially on AOSP supported devices. The primary goal of CM is stability, and users always have the option of using a third party kernel if they want to go that route.
Add a comment...