87 replies to this topic

[burstlam]

Blade kernel 2.6.35 for Gingerbread Leak Version(2.3.4)  v2.5.2
2GVMSPLIT - should work on roms base on the zte leak version.

github : here
- Built with GCC 4.5.4 tool-chain  ,G-ziped Zimage
- Overclockable(122-806Hz), AHB overclock enabled
- SWAP enabled
- Support RAMZSWAP & ZRAM compcache module (Information : see )
- Available CPU Governors : Ondemand, InteractiveX, SmartAssV2, Conservative, Performance & MinMax ( Ondemand CPU governor as default )
- I/O Scheduler : NOOP, CFQ, VR, BFQ & SIO
- Built in functions: CIFS -UTF8 - TUN- FUSE- DM-CRYPT - EXT3 - EXT4 - JDB2
- Brain Fxxk Scheduler ( Information )

Quick Reference on diff CPU Governor and IO scheduler : XDA Forum Link
I
Install with ClockWorkMod 3+

Known Issue:
As for the incompetence of hardware driver in the source code itself.
Some of the Blades out there would experience with hardware malfunction,
like proximity sensor and camera not working.  
(Feedbacks  from Chinese Blade Forum draw me a conclusion that machines
with these problems are mostly the White Blade or produce in Spring time this year.
I think they are from the same factory and share the same hardware suppliers
that adjustment  to source code driver is needed even they are using the same chipset.)
*********************************************************************************
24th Sep 2011 - Patcher Version 2.5.2

download link : Press Here
md5 : cec45ac045621df71e4323974e5cda0c
SHA1 : 8408eee9b7474db6d002b65c2e517b2810acdfe5

Change Log:

v2.5.2 (24th Sep 2011)

- Compiled with GCC Arm Toolchain 4.5.4 with optimized Cflags
- More fix picked from CAF
- Drop BFS for now as it breaks multimedia decoding. Rollback to CFS Scheduler (yet cgroups is troubling performance...)
  need to work from scratch again.

v2.5.1 (21st Sep 2011)
- Optimize on VM Read/ Write Speed
- More fixings on Memory service.
- Ondemand Governor Update.

Download Link : Press Here

v.2.5.0 (19th Sep 2011)
- Compiled with GCC Arm Toolchain 4.5.2 with optimized Cflags
- Adopted with Brain Fxxk Scheduler (BFS) as Default. Performance improved.
- Lagfree Cpu Governor, base on Conservative, but have better latency setting.
- Tweaking on Swap memory behavior
  v.2.4.5 (17th Sep 2011)
Thanks to Tomg and wbaw. construct the TSL2771 module with a working init. values.
Now no more calibration is need in most cases.

Download Link : Press Here


v.2.4.4 (15th Sep 2011)
- Compiler Optimization to ARMv6 VFP
- Gzip Image : bigger in size but consume less system resource and release 4mb system ram.

2.4.4 GEN1 BOOT IMAGE : http://www.multiupload.com/B9Z90LUC4F
2.4.4 GEN2 BOOT IMAGE : http://www.multiupload.com/8YUF97803K
v.2.4.3 (14th Sep 2011)
- Use Universal converter Script from KK's. Auto patch the boot image according to Gen1/ Gen2 memory layout.
- Fix Low CPU Floating Point Score issue with Antutu Benchmark
- Add Simple I/O scheduler for testing.  (Base on Deadline Scheduler, It gives better Database I/O performance for emmc/ssd internal storage.)

http://www.multiupload.com/R6J9GCX549

v.2.4.2 (13th Sep 2011)
- Proximity wakelock fix update (Pulled from TomG)
- Accurate Battery information

http://www.multiupload.com/MT6C8XFU70

v.2.4.1 (12th Sep 2011)
- Unique Version : VR- BFQ -CFQ - NOOP all in one version. Just pick it with No Frills CPU Control

v.2.4 (11th Sep 2011)
- add TOMG's Patch on Proximity Sensor, works fine on my OLED SF BLADE with a clean install.
- drop Androko FM driver with RDS support (the battery drainage issue is in a deadend, may be it should be dealt with the Fm apps itself and libaudio.so.which I don't have the resources to do so.)

v.2.3 (8th Sep 2011)
- VR version without RDS support build
        - BFQ vesion

v.2 (4th Sep 2011)
first release in Modaco Forum
  
2.4.3 Download link : Press Here

md5: 61c1af8d0f72ed12e99155a4094844bb
SHA1 : e94f13453bbf1cfdb4c381790d23ab9e930ff047



***********************************************************************************************
VR I/O version - with RDS FM support

Download Link : Press Here

edit : RDS has been dropped in this project. this is the last support version.

MD5: 0f82134b4312d60aa5e1df171bea8520
SHA1 : a8f525fec9629c1f6937be47bd291bf6524b2fc6


**************************************************************************************************************

Edited by burstlam, 23 September 2011 - 06:38 PM.

[t0mm13b]

Any idea would this work in VMSPLIT3G for CM7 rom?

[burstlam]

t0mm13b, on 08 September 2011 - 02:49 PM, said:

Any idea would this work in VMSPLIT3G for CM7 rom?

TomG submit  his .35 kernel along with libsensor  and wifi update here.
I would try my VMSPLIT3G built for CM7 as soon as the repo sync issue  back to nornal  

Edited by burstlam, 08 September 2011 - 03:04 PM.

[brickornot]

THANKS !!

[brickornot]

In short - what is compcache:

Quote

RAM have stuff. More RAM make phone faster. Squish stuff in RAM. RAM have more free space. Phone go faster.

http://forum.xda-dev...785&postcount=9

[wbaw]

brickornot, on 08 September 2011 - 03:22 PM, said:

In short - what is compcache:

http://forum.xda-dev...785&postcount=9

It's usually counter-productive because it takes time & cpu power to compress & decompress stuff. There isn't much of a cpu in the blade. So you'd probably find out that it makes it slower.

[hlubepet]

burstlam, on 08 September 2011 - 03:03 PM, said:

TomG submit  his .35 kernel along with libsensor  and wifi update here.
I would try my VMSPLIT3G built for CM7 as soon as the repo sync issue  back to nornal  

GR8, repo sync was working+there wasnt source for kernel, now there is a kernel source, but repo sync is not working few days , really nice

[mitchde]

I installed on B22 GSF (after wipe darvik cache) - all working so far. THANKS

EDIT: Great RDS radio is working ( using FM RADIO V 0.6)

Edited by mitchde, 08 September 2011 - 04:28 PM.

[brickornot]

Somebody is going to upload gsfb22 with this new kl2.2 ?

sorry
http://android.modac...ost__p__1797900

Edited by brickornot, 08 September 2011 - 06:07 PM.

[FelixL]

Does this work on the OMC-port? I guess it should, since the port afaik uses the same kernel as the leak.

[snowbord]

wbaw, on 08 September 2011 - 03:27 PM, said:

It's usually counter-productive because it takes time & cpu power to compress & decompress stuff. There isn't much of a cpu in the blade. So you'd probably find out that it makes it slower.

I have to agree with wbaw as a whole here.  Great idea but 512mb RAM is plenty for the blade and this exercise is redundant.  wbaw - any chance of disabling this on-the-fly compression / decompression in GSF?  My mission is (optimal) performance and most importantly, maximum battery life.  This compress / decompress will needlessly put a strain on the CPU, hence battery.

Edited by snowbord, 08 September 2011 - 08:25 PM.

[burstlam]

snowbord, on 08 September 2011 - 08:22 PM, said:

I have to agree with wbaw as a whole here.  Great idea but 512mb RAM is plenty for the blade and this exercise is redundant.  wbaw - any chance of disabling this on-the-fly compression / decompression in GSF?  My mission is (optimal) performance and most importantly, maximum battery life.  This compress / decompress will needlessly put a strain on the CPU, hence battery.

let me clarify that the ramswap/zram module won't init. unless u make a script to start. so there is no harm for it compiled in the kernel.
Indeed there are so many fellows in China with a blade running on 256mb system ram only.
The kernel should take care of this. and in real practice.
My test group in china find no major "slow down" or other obvious side effect (yet! of course I hope not)

Edited by burstlam, 08 September 2011 - 09:07 PM.

[snowbord]

burstlam, on 08 September 2011 - 09:04 PM, said:

let me clarify that the ramswap/zram module won't init. unless u make a script to start. so there is no harm for it compiled in the kernel.
Indeed there are so many fellows in China with a blade running on 256mb system ram only.
The kernel should take care of this. and in real practice.
My test group in china find no major "slow down" or other obvious side effect (yet! of course I hope not)

You da man!  If you can squeeze in undervolting (I remember what you said about stability), that would make this kernel killer!
Good work.

[wbaw]

burstlam, on 08 September 2011 - 09:04 PM, said:

let me clarify that the ramswap/zram module won't init. unless u make a script to start. so there is no harm for it compiled in the kernel.
Indeed there are so many fellows in China with a blade running on 256mb system ram only.
The kernel should take care of this. and in real practice.
My test group in china find no major "slow down" or other obvious side effect (yet! of course I hope not)

I only tried it once, a while ago on CM7 & it didn't work well on my CM7 512mb blade, it seemed to have a negative effect. Maybe it's different with your version & on 256mb v880s. I agree there's no harm in compiling it into the kernel, it'll only be used if you have scripts to configure & start it. If it works at all, then it'd probably be better to have a swap partition on the sdcard instead. I don't think we'd see any benefit from either on a 512mb blade though.

[judev]

How do i install this?

[snowbord]

Can someone with the necessary coding skills please enable the ability to have JuiceDefender switch between 2G and 3G (only works on supported ROMs). burstlam?

[fp30]

snowbord, on 08 September 2011 - 09:15 PM, said:

You da man!  If you can squeeze in undervolting (I remember what you said about stability), that would make this kernel killer!
Good work.

I tried (mild) undervolting in my kernel and it was very unstable. Dont know if it will be any different in this kernel but my thoughts are that it wont be stable... worth a shot though if you have time

[brickornot]

judev, on 09 September 2011 - 08:37 AM, said:

How do i install this?

if you have gsf or similar rom, just boot into clockword mod and install zip from sdcard, and select this file, of course you previously copied it on sdcard....
and reboot.

[judev]

Thanks

[hunter`]

is audio routing (for sip calls ) fixed with this kernel?