Jump to content

[KERNEL] [Y300/G510] Stock Huawei


Guest

Recommended Posts

Guest fonz93

How can I create my personal version ??

 

Linux and kernel building guides using his sources

Edited by fonz93
Link to comment
Share on other sites

Usually I use to answer many questions and help wherever I can ... as far as I can.

 

But compiling a kernel is a too time consuming topic. It requires some knowledge about Linux, installing required libraries, .... and so on.

 

Sorry when I cannot answer your question blush.png

 

Good luck with your project! wink.png

Link to comment
Share on other sites

 

This should be a good help to succeed in it smile.png

 

I can recommend to install a Virtual-Machine and setup a Xubuntu 14.04 build environment, although it's possible to build kernels with Windows, too, of course.

Edited by Guest
Link to comment
Share on other sites

Guest Syssx420

This should be a good help to succeed in it smile.png

 

I can recommend to install a Virtual-Machine and setup a Xubuntu 14.04 build environment, although it's possible to build kernels with Windows, too, of course.

 

Thanks i will work on it this afternoon

Yeah there is a lot of tutorials on internet but they are mostly aimed to a specific kernel and source. You have adapt that tutorial to the source. Also I've built a kernel with Ubuntu 12/13 and ZorinOS. ZorinOS is the most windows-like linux and it's easy to use maybe easier than Ubuntu. But you have to find many packages because the repos are not the same as Ubuntu. I would recommend you Ubuntu 12 as it has a lot of tutorials how to setup it for android building.

Link to comment
Share on other sites

Guest H3ROS

I tested this against my modded version of an older Synopsis kernel but found that it's write speed to /data is much slower. 90.53MB/s from my modded version vs. just 5.79MB/s with this kernel. And that's after many tests and I've even tried changing the I/O scheduler without any gain. The only plus is that the read speed is a little better (24.84MB/s vs. 32.34MBs/s) but that's not worth the performance hit to writes.

 

https://play.google.com/store/apps/details?id=com.andromeda.androbench2

 

Update: I just tried with an older official version of Synopsis and CeXstel and they're also producing much more than 5.79MB/s.

 

 

UpdateThis was actually a false alarm. This kernel has the correct reading, but older versions of the Synopsis kernel and the CeXstel kernel output an invalid write speed.

Edited by H3ROS
Link to comment
Share on other sites

Thx for your feedback. I'm going to investigate this issue further as soon as I can. This kernel doesn't include any further patches yet, like synopsis did. But maybe the issue is located elsewhere.

 

Do you have any write speed values of other tested kernels for me?

 

Which settings did you use on Androbench?

Edited by Guest
Link to comment
Share on other sites

Guest H3ROS

UpdateThis was actually a false alarm. This kernel has the correct reading, but older versions of the Synopsis kernel and the CeXstel kernel output an invalid write speed.

 

 

On Androbench I simply press all and then let it do it's thing. If you then go to history and tap the latest result you'll see the SEQ RD and SEQ WR results.

 

Here's a result from my modded kernel:





Target: /data (ext4)
Use Buffer: No
SEQ RD: 24.84 MB/s
SEQ WR: 90.53 MB/s
RND RD: 1289.48 IOPS(4K)
RND WR: 24531.33 IOPS(4K)
 
Taget: /data (ext4)
Browser: 246.0 msec
Market: 555.0 msec
Camera: 539.5 msec
Camcorder: 1357.25 msec

And here's a result from this kernel that I've just taken:





Target: /data (ext4)
Use Buffer: No
SEQ RD: 22.24 MB/s
SEQ WR: 1.84 MB/s
RND RD: 1078.69 IOPS(4K)
RND WR: 64.13 IOPS(4K)
 
Taget: /data (ext4)
Browser: 528.75 msec
Market: 722.75 msec
Camera: 631.73 msec
Camcorder: 1454.5 msec
Edited by H3ROS
Link to comment
Share on other sites

Guest H3ROS

Overclocked the GPU and CPU, the same as CeXstel. It's based on your original kernel, but I've also got zram fixed after it stopped working before. Like I've said though, even the CeXstel kernel and the older official version of your kernel score much better.

 

My ROM is currently work in progress. I've been giving some of my fixes to Syssx420 and fonz93 as I work on it.

 

@ fonz93

Changing the governor or the I/O scheduler doesn't seem to improve the result either. Something else has caused it, but I'm not sure what at the moment.

Edited by H3ROS
Link to comment
Share on other sites

I just flashed B199 Stock Kernel...same POOR results like you experienced with this kernel build.

 

Seems to be an issue with huawei's source.

 

Tested with synopsis 17/04...same results.

 

Now i'm confused huh.png

 

Neither this kernel, nor stock b199 or synopsis got different results for me. They are all similar. Sequential write is around 4-5 MB/s and Random Write is below 1 MB/s (~55 IOPS)

 

Can you upload your kernel build please?

 

I tested with Ondemand/CFQ (because they are available at all kernels mentioned)

 

Did you tweak any kernel parameters?

Edited by Guest
Link to comment
Share on other sites

Guest H3ROS

Currently I can't upload the kernel publicly as it's still work in progress. But like I said, Synopsis (BETA 1?) and the CeXstel kernel all score 70MB/s+.

 

I'm actually thinking this may be the other way around. I think older versions of the Synopsis kernel and the CeXstel kernel are giving a false reading. It's just that the read speed should be much faster than the write speed, but here I see the write speed being way more than the read. That never happens in any benchmark software.

 

@ Zola

The app's not to blame. It works fine on my other phones and tablets. The one(s) you linked to actually fail for internal tests as they don't have superuser permissions.

 

Like I've said though, I think it could be Synopsys BETA1 and CeXstel reporting a false reading.

Edited by H3ROS
Link to comment
Share on other sites

Currently I can't upload the kernel publicly as it's still work in progress. But like I said, Synopsys (BETA 1?) and the CeXstel kernel all score 70MB/s+.

 

I'm actually thinking this may be the other way around. I think older versions of the Synopsys kernel and the CeXstel kernel are giving a false reading. It's just that the read speed should be much faster than the write speed, but here I see the write speed being way more than the read. That never happens in any benchmark software.

 

That's exactly what i thought, too. The writing is always slower than reading.

 

I can confirm, that CeXtels v0.62 kernel behaves like you described before: sequential read (23 MB/s) - sequential write (63.45 mb/s)

 

All other tested Kernels (B172, B199, Synopsis 17/04 and the kernel of this topic) got like: sequential read (25-30 MB/s) - sequential write (3-5 MB/s)

 

I assume that's some kind of bug and to be honest I don't trust on the results of one single benchmark which got updated the last time on 2012. This needs definitely further investigation and testing.

Link to comment
Share on other sites

Guest H3ROS

It's a bug for sure, but I'm 99% sure it's kernel related as the software works for all of my other phones and tablets. I actually just tested with one of the original versions of Synopsis from back when I repacked it so that people could install it onto FusionX. That version also scores less than 6MB/s for write, but everything after that scores much more. So one of the cherry picks must of caused it, but which one I've no idea.

Link to comment
Share on other sites

To summon it up: this kernel might not be optimized regarding writing performance, yet, but also doesn't have any significant issues about it? Agree? :)

Link to comment
Share on other sites

Guest ZolaIII

@ h3ros

I did told your to use SDS not a SD2! & run I is for internal!

One of the reasons why this one is slower is that asynchronous is not enabled!

Edited by ZolaIII
Link to comment
Share on other sites

Guest H3ROS

I agree yes, it was a false alarm. I should of checked the stock kernels, but I'd gone right from my modded kernel to this one and I seen the difference so I said something right away. Sorry about that.

 

As the stock kernels also report the same write speed I'll go back to my original post and add an update, just in case someone reads that. I guess I'll also need to rebase my kernel on your latest work if I want dynamic fsync and GPU/CPU overclocks, that is unless I want to waste a few hours trying to locate the bad commit that caused this.

 

@ Zolla

SDS is the first thing I used and it throws errors. It says some status, but there's errors there. What it does display is no different than AndroBench though because it's not an app issue. The issue is with a commit made to older versions of Synopsis and the CeXstel kernel. Which commit I'd like to know as it may cause other currently unknown issues.

Edited by H3ROS
Link to comment
Share on other sites

I'm OK with your report, don't worry. I'm glad whenever someone is figuring out potential issues which needs to be solved.

 

I agree that finding the bad commit might take a lot of time though. As far as I remember comparing BETA2 and BETA1 i only know what BETA2 did NOT have: fsync, dynamic fsync, timer slack controller, frandom (features cexstel's kernel got, too)

 

By the way: PassMark Benchmark is reporting "reasonable" values for read/write. I tested this APP on Cexstels v0.62 kernel and this one. The values for read/write speed were close to each other. Both kernels had faster reading than writing.

 

EDIT: you should consider testing other storage related benchmarks too and analyze the results

Edited by Guest
Link to comment
Share on other sites

Guest ZolaIII

This are my results with runl & a deadline & it work's OK!

Android DriveSpeed Benchmark 1.1X 18-maj-2014 00.09

Internal Drive Data Cached

MBytes/Second

MB Write1 Write2 Write3 Read1 Read2 Read3

8 138.3 140.1 136.1 386.3 402.4 393.7

16 135.8 136.8 131.0 391.6 394.2 385.3

Cached

8 58.0 57.1 10.3 258.5 355.0 394.1

Random Write Read

From MB 4 8 16 4 8 16

msecs 163.28 43.71 40.10 0.01 0.00 0.01

200 Files Write Read Delete

File KB 4 8 16 4 8 16 secs

MB/sec 30.30 18.98 36.18 106.00 200.34 284.23

msecs 0.14 0.43 0.45 0.04 0.04 0.06 0.024

Files Deleted

Total Elapsed Time 26,5 seconds

Path Used /data/data/com.drivespeed/files/

System Information

Internal Drive MB 1135 Free 367

SD Card MB 29971 Free 3622

Screen pixels w x h 480 x 800

Android Build Version 4.1.1

Processor : ARMv7 Processor rev 1 (v7l)

processor : 0

BogoMIPS : 668.86

processor : 1

BogoMIPS : 668.86

Features : swp half thumb fastmult vfp edsp thumbee neon vfpv3 tls vfpv4

CPU implementer : 0x41

CPU architecture: 7

CPU variant : 0x0

CPU part : 0xc05

CPU revision : 1

Hardware : MSM8x25 U8833 BOARD

Revision : 0000

Serial : 0000000000000000

Linux version 3.4.0-perf-gada7609-dirty (android@localhost) (gcc-Version 4.8.3 20140401 (prerelease) (crosstool-NG linaro-1.13.1-4.8-2014.04) ) #1 SMP PREEMPT Fri May 16 23:44:19 CEST 2014

Device Y300

Something is wrong bat it's much deeper than any commit you merged.

Hire their are no adds to Huaweis source...

I will try to find a NAND chips declaration!

Pass Mark is the worse test for R/W performance, try to run individual tests in it. ;)

Edited by ZolaIII
Link to comment
Share on other sites

Please sign in to comment

You will be able to leave a comment after signing in



Sign In Now
×
×
  • Create New...

Important Information

By using this site, you agree to our Terms of Use.