[Kernel] Netformer...eating all the pies, Testing: 1.6 (15-Aug-2011) for 3.2

Search This thread

axe2

Senior Member
Dec 15, 2009
104
6
Tucson, AZ
Just an FYI... I have been using 1.0 for a couple of days, with max set to a conservative (default) 1GHz, and I've gotten 2 SODs in as many days. I never had them prior to installing this kernel. I'm trying 1.1 now.

That said, is there any way that you can turn on swap? I'm running an Ubuntu chroot, and I'd like to see how it works with heavy usage.

Sent from my Transformer TF101 using Tapatalk
 

unknown_owner

Senior Member
Nov 27, 2009
812
157
Ill give this a go after the speaker issue gets fixed and netflix gets added. Once thats done I'll be on that donate button like white on rice.

Sent from my Dell Streak using fingers

I lost my left speaker a while ago but it's back! I have been flashing Roms all day Sunday and clearing user data and all. I never tested the speakers until late last night and both are at default levels (sounds equal on both sides). I also flashed the netformer experimental kernel 1.0. So I have prime 1.3_us with netformer kernel 1.0. Not sure what fixed it but I will monitor it.

Haven't done 1.1 yet.
 

seshmaru

Senior Member
Aug 22, 2009
1,294
374
I like how I tipped this to them three days ago with photos for proof, credits to netarchy and links to all the sources and they apparently "got it from lilliputing".
 

freeza

Recognized Developer
May 28, 2006
5,416
14,188
Pasadena, CA
www.freeza-inc.com
OnePlus 10 Pro
I like how I tipped this to them three days ago with photos for proof, credits to netarchy and links to all the sources and they apparently "got it from lilliputing".

Engadget sucks. I tipped them about the Google Voice+Sprint integration going live days before they posted it about it and nothing. Maybe once archy drop kicks SoD, then he will be something more than 'someone in XDA' to Engadget.

Anyway, back on topic: 1.4 GHz has been very stable even at v1.0.
 

coolbho3000

Retired Senior Recognized Developer
Dec 26, 2008
897
785
Netarchy, is the kernel for the Transformer based on Google's 2.6.35?

EDIT: NVM, filing through your github now :) I think you could very easily get it up to 1.5GHz. Take a look at the modifications I did on the Xoom (https://github.com/coolbho3k/Xoom-OC/commits/android-tegra-2.6.36-honeycomb/); did you adjust the power regulators for the increased voltage at 1.4? I think 1250 mV is stable at 1.5 for most people on the Xoom.
 
Last edited:

netarchy

Retired Senior Recognized Developer
Jun 16, 2010
1,441
1,170
San Antonio, TX
github.com
Netarchy, is the kernel for the Transformer based on Google's 2.6.35?

EDIT: NVM, filing through your github now :) I think you could very easily get it up to 1.5GHz. Take a look at the modifications I did on the Xoom (https://github.com/coolbho3k/Xoom-OC/commits/android-tegra-2.6.36-honeycomb/); did you adjust the power regulators for the increased voltage at 1.4? I think 1250 mV is stable at 1.5 for most people on the Xoom.

I don't intend for 1.4 to be the max, it's just the current max ;D

Regarding voltages, there is plenty of room for refinement in finding perfect voltages (not to mention some of the current settings are largely for my local testing for other purposes).
 
  • Like
Reactions: Tjwoo

egzthunder1

Admin Emeritus - Spirit of XDA
I like how I tipped this to them three days ago with photos for proof, credits to netarchy and links to all the sources and they apparently "got it from lilliputing".

They may have also gotten it from our Portal, considering that I published this article at noon today and 3 hrs later... poof... it shows up in there.
 

spuniun

Senior Member
Oct 7, 2006
489
68
1.1 seems dramatically more stable for me then 1.0.

Been running it since last night @1.4ghz (interactive) and no issues at all. Great job!

Netarchy: what are your thoughts on adding the 'smart-ass' governor for better battery life?

I haven't had much luck with either versions at 1.4GHz. Is this kernel capable of adjusting vdd levels on the fly?
 

netarchy

Retired Senior Recognized Developer
Jun 16, 2010
1,441
1,170
San Antonio, TX
github.com
Netformer 1.2 up for testing.

Changelog:

Revision 1.2:
- Power regulator bug fixes that may address the "SoD" issue experienced even by users on the stock kernel
(technically not actually a sleep of death)
 
  • Like
Reactions: dirwdirw

funnycreature

Senior Member
Netformer 1.2 up for testing.

Changelog:

Revision 1.2:
- Power regulator bug fixes that may address the "SoD" issue experienced even by users on the stock kernel
(technically not actually a sleep of death)

Downloading now (Kernel Manager). I've had the "SoD" also for the first time when using 1.0 and 1.1. Do you think it can be related to the dock in any way? I noticed a much longer than usual lag between pressing the unlock key and the screen turning on.
 

netarchy

Retired Senior Recognized Developer
Jun 16, 2010
1,441
1,170
San Antonio, TX
github.com
Downloading now (Kernel Manager). I've had the "SoD" also for the first time when using 1.0 and 1.1. Do you think it can be related to the dock in any way? I noticed a much longer than usual lag between pressing the unlock key and the screen turning on.

If my suspicion is correct, the sleep of death isn't actually a sleep of death, it's a glitch in the power regulator that can be triggered when the tablet changes from a very low voltage to a high voltage (such as jumping from 216mhz to 1ghz). When the glitch is triggered, the result is a complete system lockup.

It stands to reason that a big frequency jump like this would be likely to happen when waking up the device from sleep (ie; screen is off, and you try to turn the screen on), it could be interpreted as a sleep of death.

The problem is currently unfixed in the official ASUS kernel sources, which (assuming this is the source of the problem) is likely why you can find reports of people running stock configurations experiencing the problem.
 
  • Like
Reactions: Shinzul

Top Liked Posts

  • There are no posts matching your filters.
  • 68
    So, who has cajones that clank?

    Current things of interest:
    - Overclock up to 1.6ghz [Use a tool such as Setcpu to overclock]
    - 2.6.36.4 kernel base
    - CIFS included
    - TUN built-in
    - NTFS support
    - Voodoo Sound

    All sources can be found on my github (see signature)

    These are designed to be flashed via clockwork, or you can flash them via Shinzul's Kernel Manager (Available on the android market)

    Disclaimer: You flash these at your own risk, and assume any and all potential consequences, including (but not limited to), loss of data, meltdown of your device, world destruction, injury of assorted puppies, spoiling of your glass of milk, revocation of chocolate privileges...etc. In all seriousness, it's an experimental kernel, so treat it as such, and enjoy.

    Overclocking disclaimer: Overclocking is never guaranteed to be stable (that's simply common sense). If an overclock speed is not stable, your device may not be able to handle it and you should try a lower speed.

    BE SAFE!!! MAKE A NANDROID BEFORE FLASHING SO YOU HAVE SOMETHING TO FALL BACK ON

    Current version: Netformer 1.6 (For Honeycomb 3.2) [Also available in Kernel Manager, from the Android Market]
    Netformer 1.6

    Tested with Prime 1.7

    md5sum
    3223bb177a86434837f5fe8630315a7d netformer-1.6-signed.zip
    9
    I'm apparently one of those who doesn't have any issues; just flashed Netformer 1.5 over new Prime 1.6 again, no issues whatsoever.

    Question to netarchy if he ever checks out the thread:

    Do you think we can add some other improvements?

    - ext4 no journal filesystem options
    - noatime, nodiratime (is it already there? not sure)
    - jhash3 instead of jhash2
    - HZ value (what's the current one I wonder? 500Hz was the best for SGS, made things look much smoother)
    - what codesourcery are you using? 2010.09 seems the best
    - new tiny RCU
    - remove debug logging

    Just throwing in some ideas.

    EDIT: noauto_da_alloc is a good idea too for ext4. Also, no journaling at all might be a bit of overkill but data=writeback instead of default data=ordered should speed it up by quite some value.

    The ext4 mount stuff is already a default even in stock.
    Last I looked tinyrcu did not play well with smp
    Debug stuff is already removed
    jhash3 is planned, I just haven't gotten around to adding it
    HZ is currently at default (100), a brief trial at 1000 revealed problems I haven't had a chance to correct before playing with the config option again.
    When I use codesourcery, I use a newer build (based on gcc 4.5.2) than what you reference (which I believe is a gcc 4.5.1 release), but usually I use a toolchain I built on my machine.

    I know there appears to be a lack of attention from me lately in the tf forums, but I assure you that I'm still watching and I have plans for the kernel (I do love my TF after all ;D).
    8
    Hey Netarchy,

    Would it be possible to see a Netformer version with updated Voodoo Sound? Last version of voodoo control app says the kernel module is outdated. Thanks!

    Soon™

    Especially with the 3.2 update hitting now, no doubt there will be some small delay before asus drops updated kernel sources, and then balls to the wall. ;D
    5
    Code:
    Changelog:
    Revision 1.6:
    - Updated for Honeycomb 3.2
    - Due to some rather large changes by ASUS in the kernel, the tap-to-click disabler app currently will not function in this latest kernel. If it is possible to bring the functionality back, it shall be done.
    
    Revision 1.5:
    - Updated to support android 3.1
    - Added Voodoo Sound
    *Your quadrant epeen may not be quite so large (though they may be more accurate) due to some bug fixes from upstream, if there is room for improvement, it will be explored*
    
    Revision 1.4r2:
    - Fix a bug causing wifi to break when the tablet sleeps
    
    Revision 1.4:
    - Add SLQB memory allocator
    - Set timer to 1000hz
    - Add 1.504 and 1.6ghz frequency steps
    - Add preliminary toggle interface to allow tap-to-click to be disabled (See OP for a nice gui front-end for the toggle, courtesy of vividboarder)
    
    Revision 1.3:
    - Assorted bugfixes, including a possible fix for a bug that could lead a cpu core to 'forget' it's speed policy and wander out of sync
    
    Revision 1.2:
    - Power regulator bug fixes that may address the "SoD" issue experienced even by users on the stock kernel 
       (technically not actually a sleep of death)
    
    Revision 1.1:
    - Adjusted voltage levels for stability
    - Added NTFS support
    - Assorted bugfixes
    5
    New build up for testing: 1.5

    Changelog:
    Revision 1.5:
    - Updated to support android 3.1
    - Added Voodoo Sound
    *Your quadrant epeen may not be quite so large (though they may be more accurate) due to some bug fixes from upstream, if there is room for improvement, it will be explored*