Jump to content

disco custom kernel [v.b01]


Guest disc0

Recommended Posts

Guest D.Rosendahl
Then you are with luch, because you want need to cross compile a thing , you will be working with java, not c (: ..

It seems to me it will both be java and c. It's using java for the gui and stuff for android but are using (ndk?) som c-libaries to hook into kernel.

Link to comment
Share on other sites

It seems to me it will both be java and c. It's using java for the gui and stuff for android but are using (ndk?) som c-libaries to hook into kernel.

Oah, ok.. I download the source but I'm having a hard time to debug the f**king app !

I have the ndk on my system, and eclipse does all the job!..

EDIT:

I think that ndk (jni) want compile in windows without cygwin, do you have it installed??

Edited by disc0
Link to comment
Share on other sites

Guest D.Rosendahl
Oah, ok.. I download the source but I'm having a hard time to debug the f**king app !

I have the ndk on my system, and eclipse does all the job!..

EDIT:

I think that ndk (jni) want compile in windows withou cygwion, do you have it installed??

Havn't installed cygwin yet. New computer last week, (my old one gave up) so I need to install most things again...

I realized that it's also hard to know which branch of the code to use...

Link to comment
Share on other sites

Havn't installed cygwin yet. New computer last week, (my old one gave up) so I need to install most things again...

I realized that it's also hard to know which branch of the code to use...

Well I download from svn normally.

I get this error on logcat

I/TETHER -> NativeTask( 2874): Trying to load libnativetask.so

E/TETHER -> NativeTask( 2874): Could not load libnativetask.so

Hum..

Link to comment
Share on other sites

Guest D.Rosendahl
Well I download from svn normally.

I get this error on logcat

I/TETHER -> NativeTask( 2874): Trying to load libnativetask.so

E/TETHER -> NativeTask( 2874): Could not load libnativetask.so

Hum..

Have you built your own libnativetask.so or are you using the prebuilt one that comes with the source? (I think the source for the lib is also included)

Link to comment
Share on other sites

Guest chingy1788

Thought I'd post some benchmarks

Phone - MHz				 A1 - 998	  A1 - 768	  A1 - 384		G1 - 384	G1 - 528	Units

Graphics	  

Total graphics score		23.537497	 17.895359	 8.50352		 95.58539	109.82844

Draw opacity bitmap		 8.028224	  6.1152587	 2.948035		46.12613	51.2		MPixels per sec

Draw transparent bitmap	 6.3681593	 4.8301888	 2.2530253	   12.337349   15.975039   MPixels per sec

CPU Whetstone

Total CPU score			 516.10754	 392.9966	  183.75995	   108.21737   140.81575

MWIPS DP					35.701538	 27.21829	  12.618297	   7.6493535   9.867771	MWIPS(DP)

MWIPS SP					43.956043	 33.38898	  15.683813	   8.169935	10.513036   MWIPS(SP)

MFLOPS DP				   4.577629	  3.4007597	 1.5848045	   1.4890103   1.9575995   MFLOPS(DP)

MFLOPS SP				   6.6731443	 5.1465626	 2.4179213	   1.8678708   2.5706856   MFLOPS(SP)

VAX MIPS DP				 25.544647	 19.572577	 9.161985		5.3828325   7.0545645   VAX MIPS(DP)

VAX MIPS SP				 26.948242	 20.467522	 9.591156		5.5093546   7.1622076   VAX MIPS(SP)

Memory

Total memory score		  358.1191	  325.97748	 151.89786	   107.4707	108.71283

Copy memory				 325.41492	 296.20853	 138.02623	   97.65625	98.78494	Mb/sec

Filesystem

Total file system score	 126.13296	 117.21367	 51.327778	   29.957783   30.872993

Creating 1000 empty files   21.43		 7.192		 14.748		  21.401	  15.451	  sec

Deleting 1000 empty files   4.686		 3.3		   8.149		   7.315	   13.248	  sec

Write 1M into file		  3.7608123	 3.2583904	 2.2537751	   1.0990219   2.2316446   M/sec

Read 1M from file		   250		   232.55814	 101.0101		59.171597   59.88024	sM/sec

SD card perfomance

Creating 250 empty files	19.44		 19.296		22.3			69.596	  17.704	  sec

Deleting 250 empty files	19.668		17.772		19.144		  23.956	  22.6		sec

Write 1M into file		  14.265335	 10.869565	 11.560694	   4.8053823   7.621951	M/sec

Read 1M from file		   227.27272	 243.90244	 100			 58.82353	60.240963   sM/sec

Used Advanced Task Killer to kill all tasks before running the benchmarks

Used SetCPU to force the CPU to run at one speed

Used "Benchmark" from softweg to run the benchmarks

Edited by chingy1788
Link to comment
Share on other sites

Have you built your own libnativetask.so or are you using the prebuilt one that comes with the source? (I think the source for the lib is also included)

Native, I'll try to build from the source.. Whats bugging me is that the app does not request root access, so, the problem might be there !!

Link to comment
Share on other sites

Guest xian1243

@disc0

Yes I did test it (the tun.ko you are referring to), it made my Liquid hang on the android boot logo. Had to flash a nandroid backup to get it working again.

EDIT: Sorry, I misread your post... I need to sleep... I only tested the tun.ko module from issue 48, not the patched VPN client GUI (posted under issue 38). That GUI is for cisco VPNs, not openvpn, but even so, I couldn't even get the kernel module working in the first place.

Edited by xian1243
Link to comment
Share on other sites

Guest D.Rosendahl
Native, I'll try to build from the source.. Whats bugging me is that the app does not request root access, so, the problem might be there !!

Maybe, it's should need root access according to all intructions. Installing cygwin now, but I have problem finding gnu make for it. (Need to find a repo that is including it)

You should probably put a notice in http://code.google.com/p/android-wifi-teth...etail?id=250#c1. Harald seems very active and could maybe help.

(I dont want to post anything before I have a kernal and a build environment that works)

Link to comment
Share on other sites

@disc0

Yes I did test it (the tun.ko you are referring to), it made my Liquid hang on the android boot logo. Had to flash a nandroid backup to get it working again.

EDIT: Sorry, I misread your post... I need to sleep... I only tested the tun.ko module from issue 48, not the patched VPN client GUI (posted under issue 38). That GUI is for cisco VPNs, not openvpn, but even so, I couldn't even get the kernel module working in the first place.

Ok, sorry to hear that :\ ..

Maybe, it's should need root access according to all intructions. Installing cygwin now, but I have problem finding gnu make for it. (Need to find a repo that is including it)

You should probably put a notice in http://code.google.com/p/android-wifi-teth...etail?id=250#c1. Harald seems very active and could maybe help.

(I dont want to post anything before I have a kernal and a build environment that works)

Thanks, I left a comment there (: ..

EDIT:

Well, the problem resides right there, the lib is there and the android system does not loads it :s ..

EDIT2:

Finnaly managed to compile the bloody lib! And app finnaly started (didn't tested yet) but got erros on log..

EDIT3:

This is getting me crazy :s .. stupid wifi interface !

EDIT4:

We are f**ked up, spent all night on this almost for nothing. Same comment I left on the issue tracker:

Ok, here comes the problems.. It seems that liquid as the same problem as cliq (uses

same module [dhd.ko] and as same chip), no one knows how to bring up the eth0

interface, and if it wasn't enough, there are two firmware images in the wifi folder

BCM4325.bin and BCM4325Fac.bin and two configuration files wifi.conf and wifiFac.conf..

I spotted something curious, in either conf files,

FIRMWARE_LOADER=wlan_loader

but there is no wlan_loader anywhere in the system, on any folder!

What can be done for us, liquid owners to run wifi tether ? :\

F**k, now I'm going to sleep completely pissed off!

Edited by disc0
Link to comment
Share on other sites

Guest chingy1788

disc0,

Is it possible to trade multi touch for single touch, but when two fingers are pressed, it wont take the mid point, but the first/one finger pressed point?

I had my fun with pinch zooming... problem is I touch the menu button when I do the one handed pinch zoom... kinda annoying

so, returning one finger's point instead of the mid point would be good... only if it affects the keyboard

Link to comment
Share on other sites

Guest D.Rosendahl
Ok, sorry to hear that :\ ..

Thanks, I left a comment there (: ..

EDIT:

Well, the problem resides right there, the lib is there and the android system does not loads it :s ..

EDIT2:

Finnaly managed to compile the bloody lib! And app finnaly started (didn't tested yet) but got erros on log..

EDIT3:

This is getting me crazy :s .. stupid wifi interface !

EDIT4:

We are f**ked up, spent all night on this almost for nothing. Same comment I left on the issue tracker:

F**k, now I'm going to sleep completely pissed off!

Thanks for your effort! So we need to find a way to bring eth0 up. Hm, how is android doing it?

Link to comment
Share on other sites

disc0,

Is it possible to trade multi touch for single touch, but when two fingers are pressed, it wont take the mid point, but the first/one finger pressed point?

I had my fun with pinch zooming... problem is I touch the menu button when I do the one handed pinch zoom... kinda annoying

so, returning one finger's point instead of the mid point would be good... only if it affects the keyboard

I need you to describe better the problem please.

You facing that problem in dolphin [ witch version ] ??

You prefer to remove the MT ability and discard completly the second touch, is it that?

Thanks for your effort! So we need to find a way to bring eth0 up. Hm, how is android doing it?

yes, and like in motorola cliq, I don't find any reference of any succeeding it.. Thats the point, I don't find nothing that can give a clue hoe android is doing it :\ ..

Link to comment
Share on other sites

Guest chingy1788
I need you to describe better the problem please.

You facing that problem in dolphin [ witch version ] ??

You prefer to remove the MT ability and discard completly the second touch, is it that?

Yes, remove MT ability and replace with just one touch

so, if I put one finger down, it returns a point, then I put a second finger down, it doesnt take the mid point, but takes the first finger's point

Its not so much a problem with dolphin, its more a problem with doing a one hand pinch zoom. My palm touches the menu button by accident

Link to comment
Share on other sites

Guest maxisma
Yes, remove MT ability and replace with just one touch

so, if I put one finger down, it returns a point, then I put a second finger down, it doesnt take the mid point, but takes the first finger's point

Its not so much a problem with dolphin, its more a problem with doing a one hand pinch zoom. My palm touches the menu button by accident

Menu button is NOT the touchscreen!

Link to comment
Share on other sites

Guest maxisma
Menu button IS capacitive!

Of course. But touchscreen & the soft keys are two different things and your problem won't be solved if MT get's deactivated.

There would have to be some sort of

"if Touchscreen is press ignore touchpanel presses"

But I have no idea where to integrate this, mabýbe disc0 knows it.. But it would be bad for apps like Gameboid.

Edited by maxisma
Link to comment
Share on other sites

Guest chingy1788
Of course. But touchscreen & the soft keys are two different things and your problem won't be solved if MT get's deactivated.

There would have to be some sort of

"if Touchscreen is press ignore touchpanel presses"

But I have no idea where to integrate this, mabýbe disc0 knows it.. But it would be bad for apps like Gameboid.

OK...

The stock keyboard does not support multi touch

What I want is to have the system not to return the mid point when two fingers are pressed

it seems that this can be done in the kernel

Might not need to remove MT

from an old post reffering to drivers/input/touchscreen/auo_ts.c

http://android.modaco.com/content/acer-liq...rce-here-it-is/

we see

	if (!finger2_pressed) {

		/* No multitouch -- force width to zero */;

		width = 0;

		x = coord[0][0];

		y = coord[0][1];

	} else {

		/* (x,y) coords for multitouch are at midpt between fingers */

		x = (coord[0][0] + coord[1][0]) / 2;

		y = (coord[0][1] + coord[1][1]) / 2;


		/* Report the width according to the abs distance of x-axis */

		width = abs((coord[0][0] - coord[1][0]));

	}
instead of having the mid point being the x, y coords
		x = (coord[0][0] + coord[1][0]) / 2;

		y = (coord[0][1] + coord[1][1]) / 2;
can we put the first point i.e.
		x = coord[0][0];

		y = coord[0][1];

from what it looks like it might affect the multi touch someway

doing this might make the keyboard easier to use or more accurate when speed typing with two thumbs, since there will not be a case where the character at the mid point is typed in

Edited by chingy1788
Link to comment
Share on other sites

Guest behnaam
JIT....JIT....JIT lol lets get JIT working on Donut Liquid, :lol: from what I am reading it will increase x1.74 in speed and more battery also.... ill try to backport it :D

Edit:

here is the backport:

His repo is synced to the current kernel release, so its a matter of syncing and start working on implementing JIT.

Im a bit interested to see what BFS patching will for more speed increase :D

Link to comment
Share on other sites

His repo is synced to the current kernel release, so its a matter of syncing and start working on implementing JIT.

Im a bit interested to see what BFS patching will for more speed increase :D

will cfs help also? do you know how they both work, im a little confused on bfs and cfs :lol:

EDIT: ill be looking into this... :D

Edited by Angio
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.