• After 15+ years, we've made a big change: Android Forums is now Early Bird Club. Learn more here.

Root [KERNEL] MiRaGe - for HTC stock ICS ROM - 3.0.101 - 10222013

I have a test build below that is compiled with the latest linaro-2013.05-gcc-4.8. It also has an insignificant bluetooth stack update from villec2 3.4 kernel which I think gives a little better sound over the bluetooth headphones. It is not a good idea to test two things at once but I am not expecting any problems with the bluetooth stack update. So here it is. My initial testing didn't have any problems, please let me know if you see any problems with this kernel.

MiRaGe_shooter_test_kernel_802.zip
 
The new build of MiRaGe is available in the OP with the following changes

- switched to Linaro gcc 4.8.1 compiler
- updated bluetooth stack from Ville C2 3.4 kernel
- updated net/wireless from Google common kernel
- updated bcmdhd driver from CAF msm-3.0 kernel (not relevant for ICS stock ROM)
- updated camera gemini driver from Ville C2 kernel, thanks to johnnyslt
- merged several fixes for camera and serial/tty driver from CAF, thanks to johnnyslt
- disabled gentle fair sleepers that should improve the GUI responsiveness
- added a patch from Intel to improve RWSEM performance

Enjoy!
 
Sorry, there was a freeze/stuttering problem with the rwsem commit that I missed in my testing.

I have reverted and rebuilt the kernel. The fixed build is available in the OP. I would appreciate if you can confirm the fix.

Edit: I have just updated today's build again after rebuilding with gcc 4.6.3. It is not worth risking stability for a supposedly infinitesimal gain in performance. I will switch to gcc 4.8 when it is labeled stable.
 
The problem was most probably due to the compiler flags. I was micromanaging the compiler flags but some of the flags change from one toolchain to another. I have now simplified the compiler flags into a common smaller set. The rest of the flags will be determined by the toolchain based on the optimization level chosen (-O2 currently) and the architecture set. This is a safer way and there is no noticeable performance loss.

I am now testing the 4.7.4 stable toolchain. If everything goes well, I will be using this toolchain in the future. I have applied the rwsem patch again and the kernel built with linaro 4.7.4 using the reduced cflags is working excellent so far.

On a different note:
I have been searching for a good sound processor that works well with the updated sound driver in the kernel. I think I found it.
http://forum.xda-developers.com/showthread.php?t=2168756

Just freeze Music and Music Enhancer in /system/app (I use link2sd for this). And flash the file "srs_mod_for_nxt.zip" in the above link via recovery. I use Google Play Music. Sound settings will appear in settings/Equalizer of Google Play Music with the same interface you can find in JB roms.

There are many other options I know. But this one works best for me.
 
mrg, the best sound processor out there is Nooxoide. It does not have to be flashed and it is free. It does not work with the HTC music player, but it works with most other players.
 
ron scubadiver said:
mrg, the best sound processor out there is Nooxoide. It does not have to be flashed and it is free. It does not work with the HTC music player, but it works with most other players.
Click to expand...

I tried it. But there were three problems for me. Too many settings for a sound noob like me. Some settings seem to be ineffective (they specify 50% compatibility with Qualcomm). Interface was a little out of place with the rest of the system. But you are right it seem to have probably the most sophisticated sound engine.
 
The most important setting in Noox is natural sound stage, and it is subtle. Some of the settings are not obvious unless set to maximum, or without extended listening. A number of ROM developers are combining it with Beats, but I really think Beats sounds dry.
 
It is the short of read-write semaphore for controlling the ordered access of threads for reading and locking the resource when there is a write request. The performance of rwsem is important due to its central role in the kernel. Here is a more detailed article
http://lwn.net/Articles/124747/
 
Here is a test kernel after applying many patches from Linux 3.10-rc5. I have picked all of the rwsem patches and thanks to johnnyslt for the heads up. I have also added an important patch that accelerates the tlb code significantly. I am quoting Peter Zijlstra:
"Since the introduction of preemptible mmu_gather TLB fast mode has been broken. TLB fast mode relies on there being absolutely no concurrency; it frees pages first and invalidates TLBs later. However now we can get concurrency and stuff goes *bang*."

My testing gives me the impression that both johnnyslt and Peter Zijlstra are right. Check it out and please post if you have any feedback.

http://www.mediafire.com/download/g4cb34rtgf4r4la/MiRaGe_shooter_test_kernel_813.zip

Edit: I have updated the test build since the default frequency was 1728 which is too high for testing. Please download again ... sorry.
 
Heat is my fault :(

I have updated the test build. Please download again ... sorry. The default frequency was 1728. You are finding problems so quickly. Thanks!
 
The 06132013 build of MiRaGe kernel is available in the OP with the following changes

- Synced with Linux 3.0.82
- backported all of the rwsem updates from Linux 3.10-rc5
- backported many tlb updates from Linux 3.10-rc5
- backported many ARM updates from Linux 3.10-rc5

Enjoy!
 
You must log in or register to reply here.
Back
Top Bottom