Gingerbread to my Ideos S7 development

Balcora

Member
Feb 6, 2011
59
15
Success on at least one part, i now have a bootable Android 2.3.3 kernel 2.6.37 i believe it is; old userland still.
Buuuuuut, and here's the funny thing....
Cyanogen isn't using any of the arm optimizations... infact the makefiles enforce armv6
(cortex is armv7).
I added the optimizations as follows, and i havn't found any nasty issues yet; but i'm willing to bet my kernel is faster and better then the current CM built one.
-march=armv7-a -mtune=cortex-a8 -mfpu=neon -ftree-vectorize -mfloat-abi=softfp

I'm going to hop on the CM forums and see if anyone has a clue as to why there is no usage of optimizations or even correct ARM target....

Also; i am at an impass,
I'm not sure if i should continue along the line and get the 2.3.3 userland up and running; or if i should take a good crack at getting an LLVM'd linux kernel for stupidly fast androidyness first.
 

goodoane

Senior Member
Nov 25, 2010
269
90
Wow.....Congratulation!
It sound very prommising for a noob like me.
When do you think to release some alfa or beta kernels?..
Did you ported huawei s7 kernel over 2.3 kernel?
Keep us updated with your work. I wish to help you with something but i am just a noob.
Thank you.
 

Balcora

Member
Feb 6, 2011
59
15
Its the CM7.0rc2 kernel built with some alterations in the code for the MSM side stuff, and some optimization thrown in.
I may have to remove all Optimization now, as apparently i found why at least some of the optimizations aren't used by default...
There is broken assembly in the qualcomm stuff, second to this; there was an annoyance with the DRM module which took a little while to beat with a stick.

I could release a kernel, but its rather useless, as its missing several pieces of hardware,
and its just a new kernel; alot of the sexyness of 2.3 is in the userland, rather then in kernel.
Thats not to dispute there is some sexy in the new kernel... of course there is :p

But for the moment; its just a test, I'll be going to work in a few hours and i havn't slept; so i'll probably stop for the night,
and will get back into it later on in the week as i have a few papers i have to write by tomorrow.

Secondly to this; I trialed a LLVM build of the kernel, it is broken for the moment; but i'd really like to have a hard go at that; as it should be "lets rape all the competition with a kernel that self optimizes"
 

goodoane

Senior Member
Nov 25, 2010
269
90
You are really look for the best and I just hope that you will move forward our little toy.
Thank you.
 

meowtablet

Member
Feb 15, 2011
11
0
I could release a kernel, but its rather useless, as its missing several pieces of hardware,
and its just a new kernel; alot of the sexyness of 2.3 is in the userland, rather then in kernel.

Anyway I strongly suggest you to "release" it somewhere on github. At least I'd love to play with it.
It would be also really nice if you properly used git to keep tracking of changes, I mean the first commit in your git repo should be original CM7.0rc2 kernel, so that people like me could see what you have changed.
 

Balcora

Member
Feb 6, 2011
59
15
Tis a possibility when I get some effort; and its already in git on my system.
My code is a different branch.
I'm not sure if this is how i'd release it though;
As the CM sources are broken up into 400,000,000,000,000 different git repo's (I would strongly like to bludgeon them for this btw).
I'd probably like to consolidate their sources and segregate the code within a single repo.
Secondly; most of my alterations so far have been assembly, so most people probably won't like to "play with it" but look at the pretty numbers sure :p.
I would personally like to go back through and find who made certain alterations and school them on ARM asm vs Thumb/Thumb2/ThumbEE.
school them with a hammer, nails, redstone/wires and a naquada generator. oh yes... 2 pop culture references in a single malicious sentence....
I'll likely branch the code again for my LLVM work.
To be honest; I will probably go deep into the LLVM kernel build as I really think that would be an epic move compared to just "getting it working".
It would make it work; and then give something to the android community that would honestly make a huge amount of efficiency issues vanish (bad programming excluded of course).

I'm honestly going to need to spend a large amount more work getting this to build properly; as currently you can't really do anything bar booting and entering nothingness :p.
If anyone else knows ARM ASM and Thumb2, help would be much appreciated; as frankly its really really irritating with respect to some of the broken MSM stuff.
 

rictec

Member
Jan 30, 2011
127
24
Its the CM7.0rc2 kernel built with some alterations in the code for the MSM side stuff, and some optimization thrown in.
I may have to remove all Optimization now, as apparently i found why at least some of the optimizations aren't used by default...
There is broken assembly in the qualcomm stuff, second to this; there was an annoyance with the DRM module which took a little while to beat with a stick.

I could release a kernel, but its rather useless, as its missing several pieces of hardware,
and its just a new kernel; alot of the sexyness of 2.3 is in the userland, rather then in kernel.
Thats not to dispute there is some sexy in the new kernel... of course there is :p

But for the moment; its just a test, I'll be going to work in a few hours and i havn't slept; so i'll probably stop for the night,
and will get back into it later on in the week as i have a few papers i have to write by tomorrow.

Secondly to this; I trialed a LLVM build of the kernel, it is broken for the moment; but i'd really like to have a hard go at that; as it should be "lets rape all the competition with a kernel that self optimizes"


are you getting MSM from codeaurora? can be faster donne (slower working) then writting it from scratch
 

rictec

Member
Jan 30, 2011
127
24
Success on at least one part, i now have a bootable Android 2.3.3 kernel 2.6.37 i believe it is; old userland still.
Buuuuuut, and here's the funny thing....
Cyanogen isn't using any of the arm optimizations... infact the makefiles enforce armv6
(cortex is armv7).
I added the optimizations as follows, and i havn't found any nasty issues yet; but i'm willing to bet my kernel is faster and better then the current CM built one.
-march=armv7-a -mtune=cortex-a8 -mfpu=neon -ftree-vectorize -mfloat-abi=softfp

I'm going to hop on the CM forums and see if anyone has a clue as to why there is no usage of optimizations or even correct ARM target....

Also; i am at an impass,
I'm not sure if i should continue along the line and get the 2.3.3 userland up and running; or if i should take a good crack at getting an LLVM'd linux kernel for stupidly fast androidyness first.

got some compiler errors trying 2.3.3 kernel 2.6.37.4 with that DRM not found

compile the beast with drm disabled on config
 
Last edited:

Balcora

Member
Feb 6, 2011
59
15
Problem with DRM disabled is, we kinda loose all usefulness in graphical land :p.
Its more that it was trying to use something that didn't exist in the sources, an annoyance, but still fixable.
I believe no one bothered to care about the assembly madness as no one seems to be using any of the real tunings for ARMv7 (still using some of the ARMv6 tuning though by the looks :p).
As for more like desire then passion.
To be honest, it really wouldn't matter which of them we looked at more closely, as both of them share the QSD8250; same as the S7.
I likely won't be doing a lot on it tonight; but perhaps tomorrow when I don't have papers i still havn't written coming up on their deadline :p.
 

rictec

Member
Jan 30, 2011
127
24
is it any way better (or can it be done without much work) to just tune kernel to ARMv7 and leave the rest of the userland ARMv6 tuned?
so where do we find those DRM files? can the older ones be used instead?

i didnt notice that both where QSD8250 just notice passion so i was trying to port that ;)
write faster s7 needs you and we need good .h files ;)
 

Balcora

Member
Feb 6, 2011
59
15
lulz; It would be preferable to get the kernel as optimized as possible without breaking things,
as for the userland, optimize for ARMv7, but depending on what works with the fine tuning (i.e. neon ThumbEE etc etc), is what we can use in userland.

I'd much prefer to get the kernel stable but optimized. and the userland also optimized, but possibly not as heavily as the kernel... depends on what we can get away with.
 

rictec

Member
Jan 30, 2011
127
24
For the rest of the drm etc wanna read get to IRC lol

Sent from my sdk using Android Tablet Forum App
 

Balcora

Member
Feb 6, 2011
59
15
Well, now with the DRM issue fixed, we can slowly walk through the rest of the kernel issues; and get a fully working kernel.
 

rictec

Member
Jan 30, 2011
127
24
so did you get any more info about that bcm_bt problem ..CM devs aren't helping much with it so we realy
 
Top