Android 4.4 RC1 + Radeon 7660D (A10 5800K)

[pstglia]

Hi friends,

I'm running this version with modesetting disabled on my config (AMD A10 5800K => Radeon 7660D IGP). 

Works fine, but I'd like to use radeon + dri to have acceleration support (videos and 3d stuff). 

 (  Asking too much, isn't it? :D  )

When trying to, I got some dmesg warnings/errors:

  * radeon bo/va conflicts with va/bo

  * segfault at xx ip xxxxxx sp yyyyy error 4 in libGLES_mesa.so

Before entering graphic mode (used DEBUG=2), the pseudo file vma in /sys/kernel/debug/dri/0/vma shows some addresses. After typing exit to enter graphic mode, this same file shows zero addresses. Could segfaults on libGLES_mesa be related to this?

Apparently, my hardware is not supported by this compiling of mesa for Android (GLES/egl maybe). Someone could just confirm this? If so, Is there any chance of support in the future?

* * *

Note (not important - read this just if you have time and patience ): I downloaded sources and compiled it replacing Mesa to a newer version (downloaded 10.0.4 and changed some Android.mk files, headers, etc). Also, compiled a newer kernel version ( 3.13.6 - used make oldconfig to get configs from the original kernel in the RC 4.4).

It boots, but the same behavior (just get rid of bo/va conflicts messages). 

These are some debug info on my "frankenstein" compiling :)

# dmesg output

6>[  487.963852] droid.launcher3[7749]: segfault at 0 ip 778750cc sp bfcef0f0 error 4 in libGLES_mesa.so[77821000+650000]

<6>[  488.178988] ndroid.systemui[7821]: segfault at 18 ip 77c42a93 sp bfcef6d0 error 4 in libGLES_mesa.so[77821000+650000]

# logcat output

I/DEBUG   ( 1516): *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** ***

I/DEBUG   ( 1516): Build fingerprint: 'Android-x86/android_x86/x86:4.4.2/KOT49H/eng.paulo.20140309.172714:userdebug/test-keys'

I/DEBUG   ( 1516): Revision: '0'

I/DEBUG   ( 1516): pid: 3189, tid: 3189, name: ndroid.systemui  >>> com.android.systemui <<<

I/DEBUG   ( 1516): signal 11 (SIGSEGV), code 1 (SEGV_MAPERR), fault addr 00000018

E/GRALLOC-KMS( 1518): failed to set crtc (Permission denied) (crtc_id 12, fb_id 59, conn 20, mode 1920x1080)

E/SurfaceFlinger( 1518): error posting framebuffer: -13

I/DEBUG   ( 1516):

I/DEBUG   ( 1516): backtrace:

I/DEBUG   ( 1516): AM write failure (32 / Broken pipe)

I/DEBUG   ( 1516):     #00  pc 0045ba63  /android/system/lib/egl/libGLES_mesa.so

I/DEBUG   ( 1516):     #01  pc 00052186  /android/system/lib/egl/libGLES_mesa.so

I/DEBUG   ( 1516):     #02  pc 0003dcf1  /android/system/lib/egl/libGLES_mesa.so

I/DEBUG   ( 1516):     #03  pc 0003a0b9  /android/system/lib/egl/libGLES_mesa.so

I/DEBUG   ( 1516):     #04  pc 00036330  /android/system/lib/egl/libGLES_mesa.so

I/DEBUG   ( 1516):     #05  pc 00036444  /android/system/lib/egl/libGLES_mesa.so

I/DEBUG   ( 1516):     #06  pc 00039f03  /android/system/lib/egl/libGLES_mesa.so

I/DEBUG   ( 1516):     #07  pc 0003fb92  /android/system/lib/egl/libGLES_mesa.so

I/DEBUG   ( 1516):     #08  pc 000355ed  /android/system/lib/egl/libGLES_mesa.so

I/DEBUG   ( 1516):     #09  pc 004bd530  /android/system/lib/egl/libGLES_mesa.so

I/DEBUG   ( 1516):     #10  pc 00411b54  /android/system/lib/egl/libGLES_mesa.so

I/DEBUG   ( 1516):     #11  pc 00411c4c  /android/system/lib/egl/libGLES_mesa.so

I/DEBUG   ( 1516):     #12  pc 0041fabc  /android/system/lib/egl/libGLES_mesa.so

I/DEBUG   ( 1516):     #13  pc 0040f6e1  /android/system/lib/egl/libGLES_mesa.so

I/DEBUG   ( 1516):     #14  pc 00034bc4  /android/system/lib/egl/libGLES_mesa.so

I/DEBUG   ( 1516):     #15  pc 003f6977  /android/system/lib/egl/libGLES_mesa.so

I/DEBUG   ( 1516):     #16  pc 0023ed19  /android/system/lib/egl/libGLES_mesa.so

I/DEBUG   ( 1516):     #17  pc 0031c4d9  /android/system/lib/egl/libGLES_mesa.so

I/DEBUG   ( 1516):     #18  pc 0031c9d0  /android/system/lib/egl/libGLES_mesa.so

I/DEBUG   ( 1516):     #19  pc 0007bc6b  /system/lib/libandroid_runtime.so

I/DEBUG   ( 1516):     #20  pc 0002a2fb  /system/lib/libdvm.so (dvmPlatformInvoke+79)

I/DEBUG   ( 1516):     #21  pc 00002e3f  <unknown>

I/DEBUG   ( 1516):     #22  pc 00086bf2  /system/lib/libdvm.so (dvmCallJNIMethod(unsigned int const*, JValue*, Method const*, Thread*)+434)

I/DEBUG   ( 1516):     #23  pc 00177618  /system/lib/libdvm.so

I/DEBUG   ( 1516):     #24  pc 00008bfb  <unknown>

I/DEBUG   ( 1516):     #25  pc 0003b7b2  /system/lib/libdvm.so (dvmMterpStd(Thread*)+66)

I/DEBUG   ( 1516):     #26  pc 00036e79  /system/lib/libdvm.so (dvmInterpret(Thread*, Method const*, JValue*)+217)

I/DEBUG   ( 1516):     #27  pc 000bc086  /system/lib/libdvm.so (dvmInvokeMethod(Object*, Method const*, ArrayObject*, ArrayObject*, ClassObject*, bool)+1750)

I/DEBUG   ( 1516):     #28  pc 000d19e0  /system/lib/libdvm.so

I/DEBUG   ( 1516):     #29  pc 00177618  /system/lib/libdvm.so

I/DEBUG   ( 1516):     #30  pc 00008eff  <unknown>

I/DEBUG   ( 1516):     #31  pc 0003b7b2  /system/lib/libdvm.so (dvmMterpStd(Thread*)+66)

I/DEBUG   ( 1516):

I/DEBUG   ( 1516): stack:

I/DEBUG   ( 1516):          bfe62b30  00000000

I/DEBUG   ( 1516):          bfe62b34  00010010

I/DEBUG   ( 1516):          bfe62b38  00000010

I/DEBUG   ( 1516):          bfe62b3c  77e21d70  /android/system/lib/egl/libGLES_mesa.so

I/DEBUG   ( 1516):          bfe62b40  7a8b9020

I/DEBUG   ( 1516):          bfe62b44  7c236380

I/DEBUG   ( 1516):          bfe62b48  40059da9  /system/lib/libc.so

I/DEBUG   ( 1516):          bfe62b4c  400e3fcc  /system/lib/libc.so

I/DEBUG   ( 1516):          bfe62b50  00010010

I/DEBUG   ( 1516):          bfe62b54  7cc81010

I/DEBUG   ( 1516):          bfe62b58  000002c0

I/DEBUG   ( 1516):          bfe62b5c  4005ad3f  /system/lib/libc.so (dlmalloc+351)

I/DEBUG   ( 1516):          bfe62b60  400e8800

I/DEBUG   ( 1516):          bfe62b64  7c237068

I/DEBUG   ( 1516):          bfe62b68  40055d59  /system/lib/libc.so (pthread_mutex_lock+9)

I/DEBUG   ( 1516):          bfe62b6c  77c7ba5d  /android/system/lib/egl/libGLES_mesa.so

I/DEBUG   ( 1516):     #00  bfe62b70  7c236820

I/DEBUG   ( 1516):          bfe62b74  bfe62bac  [stack]

I/DEBUG   ( 1516):          bfe62b78  000000c8

I/DEBUG   ( 1516):          bfe62b7c  77ce753d  /android/system/lib/egl/libGLES_mesa.so

I/DEBUG   ( 1516):          bfe62b80  7c236b05

I/DEBUG   ( 1516):          bfe62b84  0000104c

I/DEBUG   ( 1516):          bfe62b88  000000c2

I/DEBUG   ( 1516):          bfe62b8c  40055f29  /system/lib/libc.so (pthread_mutex_unlock+25)

I/DEBUG   ( 1516):          bfe62b90  00000006

I/DEBUG   ( 1516):          bfe62b94  00002850

I/DEBUG   ( 1516):          bfe62b98  000000c0

I/DEBUG   ( 1516):          bfe62b9c  400e3fcc  /system/lib/libc.so

I/DEBUG   ( 1516):          bfe62ba0  00000001

I/DEBUG   ( 1516):          bfe62ba4  78e88940

I/DEBUG   ( 1516):          bfe62ba8  77cdc149  /android/system/lib/egl/libGLES_mesa.so

I/DEBUG   ( 1516):          bfe62bac  77c68d41  /android/system/lib/egl/libGLES_mesa.so

I/DEBUG   ( 1516):          ........  ........

I/DEBUG   ( 1516):     #01  bfe62bd0  00000000

I/DEBUG   ( 1516):          bfe62bd4  00000000

I/DEBUG   ( 1516):          bfe62bd8  00000000

I/DEBUG   ( 1516):          bfe62bdc  00000000

I/DEBUG   ( 1516):          bfe62be0  00000040

I/DEBUG   ( 1516):          bfe62be4  00000040

I/DEBUG   ( 1516):          bfe62be8  00000086

I/DEBUG   ( 1516):          bfe62bec  7cc81010

I/DEBUG   ( 1516):          bfe62bf0  00000000

I/DEBUG   ( 1516):          bfe62bf4  00000000

I/DEBUG   ( 1516):          bfe62bf8  00000003

I/DEBUG   ( 1516):          bfe62bfc  00000000

I/DEBUG   ( 1516):          bfe62c00  7cc81010

I/DEBUG   ( 1516):          bfe62c04  00000001

I/DEBUG   ( 1516):          bfe62c08  00000068

I/DEBUG   ( 1516):          bfe62c0c  00000034

I/DEBUG   ( 1516):          ........  ........

[Chih-Wei Huang]

2014-03-25 9:47 GMT+08:00 pstglia <pst...@gmail.com>:
> Hi friends,
>
> I'm running this version with modesetting disabled on my config (AMD A10
> 5800K => Radeon 7660D IGP).
>
> Works fine, but I'd like to use radeon + dri to have acceleration support
> (videos and 3d stuff).
> ( Asking too much, isn't it? :D )

Not at all!
I hope to support more GPUs as possible.

> When trying to, I got some dmesg warnings/errors:
> * radeon bo/va conflicts with va/bo
> * segfault at xx ip xxxxxx sp yyyyy error 4 in libGLES_mesa.so
>
> Before entering graphic mode (used DEBUG=2), the pseudo file vma in
> /sys/kernel/debug/dri/0/vma shows some addresses. After typing exit to enter
> graphic mode, this same file shows zero addresses. Could segfaults on
> libGLES_mesa be related to this?
>
> Apparently, my hardware is not supported by this compiling of mesa for
> Android (GLES/egl maybe). Someone could just confirm this? If so, Is there
> any chance of support in the future?

I'm not sure.
Probably you could ask AMD mesa developers to see
if the GPU is supported by the mesa version we used.

> * * *
>
> Note (not important - read this just if you have time and patience ): I
> downloaded sources and compiled it replacing Mesa to a newer version
> (downloaded 10.0.4 and changed some Android.mk files, headers, etc). Also,
> compiled a newer kernel version ( 3.13.6 - used make oldconfig to get
> configs from the original kernel in the RC 4.4).
> It boots, but the same behavior (just get rid of bo/va conflicts messages).
>
> These are some debug info on my "frankenstein" compiling :)
>
> # dmesg output
> 6>[ 487.963852] droid.launcher3[7749]: segfault at 0 ip 778750cc sp
> bfcef0f0 error 4 in libGLES_mesa.so[77821000+650000]
> <6>[ 488.178988] ndroid.systemui[7821]: segfault at 18 ip 77c42a93 sp
> bfcef6d0 error 4 in libGLES_mesa.so[77821000+650000]

Thank you for the valuable input.
Unfortunately I have no such a device to test.
I suggest you debug the segfault and co-work with mesa developers
for it.

--
Chih-Wei
Android-x86 project
http://www.android-x86.org

[Andrews Magnoni]

can you post what you did to replace mesa and the kernel?

i want to check if with the new version some problems of gl textures are solved!

im instaling on an notebook amd c60!

[pstglia]

Hi,

Sorry for the delay, I have just checked the group messages today.

I'm afraid did I lot of mess to compile it. It was based on some forum info and "try and error". It compiles, but cannot be sure if this is right or not.

To replace mesa (version 10.0.4), I moved the original mesa directory in "external" to another directory (~/Downloads/mesa_ori), downloaded version 10.0.4 and copied it to external/mesa. Then I did the changes bellow:

1) Changes in ANDROID_X86_SRC_PATH/external/mesa/src/gallium/winsys/radeon/drm/radeon_winsys.h:

From: #include "libdrm/radeon_surface.h"

To: #include "ANDROID_X86_SRC_PATH/external/drm/radeon/radeon_surface.h"

Reason: I was having the error messages during compiling:

  external/mesa/src/gallium/winsys/radeon/drm/radeon_drm_winsys.c:493: error: undefined reference to 'radeon_surface_best'

  external/mesa/src/gallium/winsys/radeon/drm/radeon_drm_winsys.c:485: error: undefined reference to 'radeon_surface_init'

  external/mesa/src/gallium/winsys/radeon/drm/radeon_drm_winsys.c:444: error: undefined reference to 'radeon_surface_manager_free'

  external/mesa/src/gallium/winsys/radeon/drm/radeon_drm_winsys.c:630: error: undefined reference to 'radeon_surface_manager_new'

  external/mesa/src/gallium/winsys/radeon/drm/radeon_drm_winsys.c:667: error: undefined reference to 'radeon_surface_manager_free'

2) Changes in ANDROID_X86_SRC_PATH/external/mesa/src/gallium/auxiliary/util/u_math.h

removed the following code:

#ifdef PIPE_OS_ANDROID

static INLINE

double log2(double d)

{

   return log(d) * (1.0 / M_LN2);

}

/* workaround a conflict with main/imports.h */

#ifdef log2f

#undef log2f

#endif

static INLINE

float log2f(float f)

{

   return logf(f) * (float) (1.0 / M_LN2);

}

#endif

Reason: I was having these errors. I found other declarations of this same functions in other files, so I assumed it as duplicated (But I'm probably wrong)

  external/mesa/src/gallium/auxiliary/util/u_math.h:153:7: error: static declaration of 'log2f' follows non-static declaration

3) Changes in ANDROID_X86_SRC_PATH/external/mesa/Android.mk:

From: classic_drivers := i915 i965

To: classic_drivers := 

Reason: having problems compiling classic drivers - as my goal is radeon/gallium, removed them

  external/mesa/src/mesa/drivers/dri/common/dri_util.c:288: error: undefined reference to 'driDriverAPI'

  external/mesa/src/mesa/drivers/dri/common/dri_util.c:288: error: undefined reference to 'driDriverAPI'

4) Changes in ANDROID_X86_SRC_PATH/external/mesa/src/gallium/state_trackers/egl/android/native_android.cpp (Removed these lines):

   case HAL_PIXEL_FORMAT_RGBA_5551:

      /* fmt = PIPE_FORMAT_A1B5G5R5_UNORM; */

   case HAL_PIXEL_FORMAT_RGBA_4444:

      /* fmt = PIPE_FORMAT_A4B4G4R4_UNORM; */

Reason:  read some forum these constants are used in noveau. So, assumed no need for my compiling

  external/mesa/src/gallium/state_trackers/egl/android/native_android.cpp:150:9: error: 'HAL_PIXEL_FORMAT_RGBA_5551' was not declared in this scope

  external/mesa/src/gallium/state_trackers/egl/android/native_android.cpp:152:9: error: 'HAL_PIXEL_FORMAT_RGBA_4444' was not declared in this scope

5) Changes in ANDROID_X86_SRC_PATH/external/mesa/src/egl/main/Android.mk 

Added the following lines (marked with + in the beginning => Remove the plus symbol when pasting)

# r300g/r600g/radeonsi

ifneq ($(filter r300g r600g radeonsi, $(MESA_GPU_DRIVERS)),)

gallium_DRIVERS += libmesa_winsys_radeon

+ LOCAL_SHARED_LIBRARIES += libdrm_radeon

ifneq ($(filter r600g, $(MESA_GPU_DRIVERS)),)

gallium_DRIVERS += libmesa_pipe_r600

+ LOCAL_SHARED_LIBRARIES += libstlport

endif

Changed the following variable:

From: gallium_DRIVERS += libmesa_pipe_r600

To: gallium_DRIVERS += libmesa_pipe_r600 libmesa_pipe_radeon

Reason: Some unresolved references

  external/mesa/src/gallium/winsys/radeon/drm/radeon_drm_winsys.c:492: error: undefined reference to 'radeon_surface_best'

  external/mesa/src/gallium/winsys/radeon/drm/radeon_drm_winsys.c:484: error: undefined reference to 'radeon_surface_init'

  external/mesa/src/gallium/winsys/radeon/drm/radeon_drm_winsys.c:443: error: undefined reference to 'radeon_surface_manager_free'

  external/mesa/src/gallium/winsys/radeon/drm/radeon_drm_winsys.c:629: error: undefined reference to 'radeon_surface_manager_new'

  external/mesa/src/gallium/winsys/radeon/drm/radeon_drm_winsys.c:666: error: undefined reference to 'radeon_surface_manager_free'

6) Changes in ANDROID_X86_SRC_PATH/external/mesa/src/glsl/Android.mk

Included the following line (marked with + in the beginning - remove it when pasting)

      LOCAL_MODULE := libmesa_glsl

      + include external/stlport/libstlport.mk

      include $(LOCAL_PATH)/Android.gen.mk

Reason: This include (limits) is declared on stlport as I could check

  external/mesa/src/glsl/./s_expression.cpp:26:18: fatal error: limits: No such file or directory

7) Changes in ANDROID_X86_SRC_PATH/external/mesa/src/gallium/Android.mk

Changed the following lines:

From:  SUBDIRS += drivers/r600

To: SUBDIRS += drivers/r600 drivers/radeon

Reason: Got these errors of undefined references:

  external/mesa/src/gallium/drivers/r600/r600_pipe.c:1314: error: undefined reference to 'ruvd_is_format_supported'

  external/mesa/src/gallium/drivers/r600/r600_uvd.c:161: error: undefined reference to 'ruvd_set_dt_surfaces'

  external/mesa/src/gallium/drivers/r600/r600_uvd.c:114: error: undefined reference to 'ruvd_join_surfaces'

  external/mesa/src/gallium/drivers/r600/r600_uvd.c:176: error: undefined reference to 'ruvd_create_decoder'

  external/mesa/src/gallium/drivers/r600/r600_uvd.c:203: error: undefined reference to 'ruvd_get_video_param'

8) Copied original Android.mk from ~/Downloads/mesa_ori/src/gallium/drivers/radeon/Android.mk to ANDROID_X86_SRC_PATH/external/mesa/src/gallium/drivers/radeon/

Reason: After including the reference described on step 7, got this error:

  make: *** No rule to make target `external/mesa/src/gallium/drivers/radeon/Android.mk'.  Stop.

9) Changes in ANDROID_X86_SRC_PATH/external/mesa/src/egl/main/Android.mk

From: gallium_DRIVERS += libmesa_pipe_r600 

To:  gallium_DRIVERS += libmesa_pipe_r600 libmesa_pipe_radeon

Reason:

10) Changes in ANDROID_X86_SRC_PATH/external/mesa/src/egl/main/Android.mk

From: LOCAL_SHARED_LIBRARIES += libstlport

To: LOCAL_SHARED_LIBRARIES += libstlport libdrm_radeon

Reason: I was having these link errors (undefined references):

  external/mesa/src/gallium/drivers/r600/r600_pipe.c:1314: error: undefined reference to 'ruvd_is_format_supported'

  external/mesa/src/gallium/drivers/r600/r600_uvd.c:161: error: undefined reference to 'ruvd_set_dt_surfaces'

  external/mesa/src/gallium/drivers/r600/r600_uvd.c:114: error: undefined reference to 'ruvd_join_surfaces'

  external/mesa/src/gallium/drivers/r600/r600_uvd.c:176: error: undefined reference to 'ruvd_create_decoder'

  external/mesa/src/gallium/drivers/r600/r600_uvd.c:203: error: undefined reference to 'ruvd_get_video_param'

After doing this, I could compile it.

======================================================================================================================================================

======================================================================================================================================================

To compile a newer kernel version, I did the following:

1) Moved the original kernel directory to ~/Downloads/kernel_ori

2) Downloaded 3.13.6 from kernel.org, unpacked it and copied to ANDROID_X86_SRC_PATH/kernel

3) Copied the original android directory to my new kernel directory:

cp -R ~/Downloads/kernel_ori/android ANDROID_X86_SRC_PATH/kernel

4) Copied Android.mk from original kernel directory to my new one:

cp ~/Downloads/kernel_ori/Android.mk ANDROID_X86_SRC_PATH/kernel

5) Copied the original kernel config file and copied it to my new kernel directory (note: $OUT directory configured with lunch, as described in http://www.android-x86.org/documents/customizekernel):

cp ~/Downloads/kernel_original/arch/x86/configs/android-x86_defconfig ANDROID_X86_SRC_PATH/$OUT/obj/kernel/.config

6) Entered in the new kernel directory and executed a "make oldconfig" to use previous kernel settings. Selected default for new/unmatched options:

cd ANDROID_X86_SRC_PATH

make -C kernel O=$OUT/obj/kernel ARCH=x86 oldconfig

7) Copied .config to default kernel configs:

cd ANDROID_X86_SRC_PATH/$OUT/obj/kernel/

cp .config ANDROID_X86_SRC_PATH/kernel/arch/x86/configs/android-x86_defconfig

8) Compiled it:

make kernel TARGET_PRODUCT=generic_x86

====================================================================================================

====================================================================================================

Hope this has some utility

Cheers,

pstglia

[pstglia]

Here's the link for the iso I generated. If some of you want to try it (AT YOUR OWN RISK)

Note: I had some problems with hard disks turning of and on. AGAIN, BE AWARE OF THE RISK OF DAMAGING YOUR HARDWARE!!!

### Download and use it at your own risk!! ###

https://drive.google.com/file/d/0BxO6THtB865fa1FkS3h0NG9sY0U/edit?usp=sharing

### Download and use it at your own risk!! ###

Cheers,

pstglia

[Andrews Magnoni]

im getting a black screen with this image!

thanks for the detailed instruction. i will try with different settings and/or versions!

[Mauro Rossi]

Hi, what kind of behavior is the one you describe with hdd turning on and off? From my understanding you used the default config with kernel 3.13.6

Did you use latest mesa 10.2?

[pstglia]

Hi Mauro,

I meant sometimes system become unresponsible with intense HDD activity. I can hear it "clicking" and spining up constantly. I interpreted it as "turning off and on constantly". Probably some wrong config with power save modules

I got the kernel config from Valentine's day release (3.10) and used a "make oldconfig" to create a kernel config to version 3.13.6

No, I used mesa 10.0.4. Tried the newer releases, but had problems during compiling I couldn't solve

Yesterday I compiled and packaged a image that worked on my hardware (some bugs, but usable). The only modification I made this time was replacing mesa version. 

Changing kernel requires more time and research (My video card wasn't being recognized with my compiled 3.13.6.). Not to mention skills I don't have :)

I posted a video showing it:

http://youtu.be/xOoFLtbhlDM

I also uploaded it to my google drive (replaced the previous image I posted on April, 6th):

https://drive.google.com/file/d/0BxO6THtB865fVnJoS0xKa1FmRlk/edit?usp=sharing

Cheers,

pstglia

[Mauro Rossi]

Hi! Testing last iso on Core2 with Radeon HD 4830 and works, I'm actually posting from your LiveCD iso.

The only strange thing is the looking into Settings -> About tablet the kernel version reported is 3.10.30
compiled Monday April 14th 00:56:14 BRT

Stopping message here because I'm going to run OpenGL benchmark and reporting back

M.

[Mauro Rossi]

Hi,

OpenGL version 3.0 reported by 'OpenGL Check' app,
but in this last iso OpenGl driver version is not reported correcly in Settings -> About tablet (all three seen as 'null', while in your video they were reported correctly, what has changed?)

Some OpenGL demo working like 'Octopus' and 'OpenGL demo' with iridescent cube, which is reporting about 150 fps
Very smoth Android GUI and Youtube app working like a charm

Considering the many apps crashing on 4.4 RC1, I would say radeon driver is doing his job on my Core2 + HD 4830 system

M.

[pstglia]

Kernel version is correct. The newer kernel ( 3.13.6 ) wasn't correctly setup ( my IGP couldn't be detected my drm_gralloc ), so I rolled back to the version 3.10.30 (android-x86 Kitkat official release).

Just changed mesa as described on the first Apr, 6th post.

By the way, I also uploadd my mesa source already modified (10.0.4).

https://drive.google.com/file/d/0BxO6THtB865fanJ3MVdXWUplams/edit?usp=sharing

Important note: classic drivers are disabled

[pstglia]

Good to know it's working!

In my case, OpenGL fields are being displayed all the time. Don't know the reason it's not being displayed for your GPU (missing required firmware? - Your card is a RV770 family as I could check). You could check dmesg and logcat output. Maybe there's some clue.

However, if you are capable of process 3D stuff, I guess it's not a big problem.

I runned a 3d game on my system (Lawless) and worked correctly. Maybe you could test it also

I would be even happier if I could run netflix on it :)

[Mauro Rossi]

Hi,

I've tried again, this time with first Grub option "Live" that came at 1280x1024 resolution (before I was usign "Live VESA")

Good news, OpenGL (X.Org, Gallium 0.4 on AMD RV770, Mesa 10.0.4) reported in Settings -> About tablet

Animated wallpapers, Google Chrome, Angry Birds, Pudding Monsters, mame4droid now all working

Great!

Now if we could have a build supporting Intel, AMD, Nvidia drivers Kernel 3.14 and latest Mesa 10.X (currently best performance is with Mesa 10.1, based on Phoronix reports/benchmarks) would be awesome!!!

Thanks for the image and instructions, I'd like to be able to 'homebrew' a build myself (I'll give it a try during Easter holidays)

Mauro

[Mauro Rossi]

Hi pstglia,

in case I fail with kernel 3.14 compiling (l'll be probably doomed)  can you help me, by describing the procedure to build only mesa,
I've understood from your instructions that Android.mk editing does the trick,

but what should I do to selectively compile mesa libs only and to replace them in an existing iso image/USB bootable?

Thanks for any info.

Mauro

[Paulo Sergio]

Hi Mauro.

I'm not an expert (in fact my knowledge is very basic/limited)
but I can try to help you, of course

Android.mk is among the files that I had to change. As I described, some other files had some small change (header files, c files, etc). To compile 10.1, you'll face other problems and check on web to find a solution for them.

I'm not sure just replacing mesa libs on an existing iso will work. When you create a iso, some libs are linked with mesa. So, replacing it could generate some unresolved symbols errors.
By this reason, maybe it is easier compiling the entire source.

> --
> You received this message because you are subscribed to a topic in the Google Groups "Android-x86" group.
> To unsubscribe from this topic, visit https://groups.google.com/d/topic/android-x86/lKPmDXtNzjM/unsubscribe.
> To unsubscribe from this group and all its topics, send an email to android-x86...@googlegroups.com.
> To post to this group, send email to andro...@googlegroups.com.
> Visit this group at http://groups.google.com/group/android-x86.
> For more options, visit https://groups.google.com/d/optout.

[Mauro Rossi]

Hi, I've check around,

in order to build a single project the following commands come handy:

cd [android repo]
source build/envsetup.sh
cd external/mesa
mm

Well it works, but since I already built mesa, nothing is done, how can I force rebuilding external/mesa?
Thnaks for any info from people in the Group.

Mauro

[pstglia]

I use to clean obj files. Like this:

cd $OUT/obj/STATIC_LIBRARIES/

rm -rf libmesa*

 

[dscm]

your mesa upload needs clean up.....


external/mesa/src/gallium/winsys/radeon/drm/radeon_winsys.h:44:82: fatal error: /home/paulo/Downloads/android-x86/external/drm/radeon/radeon_surface.h: No such file or directory


#include "pipebuffer/pb_buffer.h"
#include "/home/paulo/Downloads/android-x86/external/drm/radeon/radeon_surface.h"

[Paulo Sergio]

Yeah, it's ugly :)

It's using absolute path. Can be changed using relative path or changing it to the path on your environment.

--

[dscm]

file doesn't exist in 10.04???

[Paulo Sergio]

It exists. It's under external/drm/radeon directory (outside mesa). On 9.2.1 it is solved by include paths in Android.mk files. As I couldn't find which one to change, I prefered doind this quick but ugly workaround

[dscm]

yes located correct implementation...

mesa9$ find . -type f -print0 | xargs -0 grep -i radeon_surface.h
./src/gallium/winsys/radeon/drm/radeon_winsys.h:#include "radeon/radeon_surface.h"

[pstglia]

Thanks!

I created a iso using mesa 10.1.1. Uploaded the modified source for 10.1.1 on my google drive:

https://drive.google.com/file/d/0BxO6THtB865fMTRhV0NIQnZiaEE/edit?usp=sharing

Any of you had problems with refresh/redraw in some OpenGL apps? For example, "Learn open GL ES" app, keeps a lead/track of previous frame. I filled a bug describing this:

https://bugs.freedesktop.org/show_bug.cgi?id=77966

Strange that many other OpenGL ES 2 apps works correctly ("Live Ocean Benchmark", "Lawless game", etc)

I want to know if this is specific for my hardware (Radeon 7660D IGP)

Thanks in advance!

pstglia

[pstglia]

Here's the iso I generated, in case someone wants to test it in a ATI/AMD/Nvidia hardware ( mesa 10.1.1 - No classic drivers, just gallium):

https://drive.google.com/file/d/0BxO6THtB865fMHpKb204ZUVRTGM/edit?usp=sharing

At your own risk, as usual :)

[dscm]

 not built yet...we're working kernel 3.14 from expermential issues...

[Vik Tanwar]

tried on AMD Phenom Gigabyte Mobo with 5670 pcie 4g RAM - Installation on USB Pendrive Ext3

1. Yes OpenGL ES leaves track of previous frame.
2. "Live Ocean Benchmark" crashes on launch
3. "Learn open GL ES" upto tutorial upto no. 6 works fine but crashes on last two
     Log shows following:
     "D dalvikvm: Trying to load lib /data/app-lib/com.learnopengles.android-1/libandroidgl20.so 0x43027690"
     "F libc    : Fatal signal 4 (SIGILL) at 0x7aa0a2a4 (code=2), thread 6949 (pengles.android)"

Though 2 and 3rd is happening in every build giving same error, any idea how to get around with it.

[pstglia]

Hi Vik Tanwar,

Thanks for checking this out.

About 1, a patch was created (77966) to fix the track/trail problem. Here's the new iso with it applied, in case you want it:

https://drive.google.com/file/d/0BxO6THtB865fS3dIbzBCNTFjcGs/edit?usp=sharing

About 2 and 3, You could check if your kernel is requiring some firmware for your radeon. Open terminal app and check dmesg, typing these commands:

su (will be prompted to grant root privileges)

busybox dmesg | grep -i firmware

According to "https://wiki.gentoo.org/wiki/Radeon", 4 firmware files are required for your radeon (radeon/REDWOOD_me.bin radeon/REDWOOD_pfp.bin radeon/REDWOOD_rlc.bin radeon/REDWOOD_smc.bin radeon/CYPRESS_uvd.bin).

The first 3 files are already on /system/lib/firmware/radeon directory. If dmesg output shows the other 2 files (specially REDWOOD_smc.bin), download it from "http://packages.gentoo.org/package/sys-kernel/linux-firmware" and copy to /system/lib/firmware/radeon, and try to boot android-x86 after this.

Also, check again logcat output. You can find other clues on it. 

Another question: Does the release from android-x86.org (Kitkat RC1) works for you? Maybe mesa 9.2.1 already supports your config.

Cheers,

pstglia

[Vik Tanwar]

I got following:
# dmesg | grep -i firmware
<6>[ 1.915164] pnp 00:0a: [Firmware Bug]: [mem 0x00000000-0xffffffff disabled] covers only part of AMD MMCONFIG area [mem 0xe0000000-0xe00fffff]; adding more reservations

firmware files are available:
root@x86:/system/lib/firmware/radeon # ls REDWOOD*
REDWOOD_me.bin
REDWOOD_pfp.bin
REDWOOD_rlc.bin
REDWOOD_smc.bin
root@x86:/system/lib/firmware/radeon # ls CYPRES*
CYPRESS_me.bin
CYPRESS_pfp.bin
CYPRESS_rlc.bin
CYPRESS_smc.bin
CYPRESS_uvd.bin

Kitkat RC1 is working fine, screen width/height 1980/1008 on HDMI, sound ok, network (wired) no problem, even iball webcam works in camera (should be attached to usb before booting).

"OpenGL Check" reports support upto 3.0 and "OpenGL Demo" no frame is below 115 (lowest value i could see)

syncing iso you provided, will post results.

[Vik Tanwar]

Good Work! Patched ISO is working fine, no trails shown. 

However, "Learn open GL ES" is still having issues with tutorial no.7 & 8

I also tested following apps:

Antutu - Runs and Give Highest Rating 
Aquarium - Crashes
Fish Tank HD  - Crashes
GPGUBenchmark - Runs
Magic Ripple Heart - Crashes
Live Ocean Benchmark - Crashes
Quadran Standard - Runs
Vellamo - Runs
GPU Bench - Crashes
Madagaskar 3d Benchmark - Crashes

These crashes are mainly due to unity player library libmono.so, any idea or resolution. 

One more thing is noted from log that few apps tries to find volume/mkdirat/ensure directory on /storage/sdcard1/Android/ and result in error as no SDCARD / USB Storgare is available.

[pstglia]

Hi, lessons 7 and 8 work ok on my machine. You could enable egl trace to logcat, so you can have more info on what`s going on. You can do this on
dev tools -> Developer Options -> Monitoring -> Enable OpenGL Traces -> logcat

Then, launch the app and run the lesson 7 or 8. After the crash, query logcat output (logcat -d | grep -v "cache full").

Maybe it will not produce usefull info, but we can try :)

* * *

Other apps from your list also crash on my hardware. One of them is Madagascar 3d Benchmark. Using egl trace and
lauching the app I checked it always crash after calling the same funcion (glUniform1i):
logcat -d | grep -v "cache full"

...
D/libEGL ( 5161): glGetActiveUniform(382, 1, 1024, (GLsizei*) 0x79d3a218, (GLint*) 0x79d3a21c, (GLenum*) 0x79d3a220, (GLchar*) 0x79d3a22c);
D/libEGL ( 5161): glGetUniformLocation(382, (const GLchar*) 0x79d3a22c);
D/libEGL ( 5161): glGetActiveUniform(382, 2, 1024, (GLsizei*) 0x79d3a218, (GLint*) 0x79d3a21c, (GLenum*) 0x79d3a220, (GLchar*) 0x79d3a22c);
D/libEGL ( 5161): glGetUniformLocation(382, (const GLchar*) 0x79d3a22c);
D/libEGL ( 5161): glUniform1i(2, 0);
D/Zygote ( 1418): Process 5161 terminated by signal (6)
I/ActivityManager( 1741): Process com.roofgame.madagascar3d (pid 5161) has died.

This function is related to shaders variables (I think), but I can`t say what is wrong. Maybe something missing on OpenGL implementation, radeon driver. I`m not thinking in a app bug because it runs on a "real" android device (Mean Smartphones / Tablets).

[Vik Tanwar]

tried "Rajawali Example", no problem with any of example in the app.

Also tried few "Rajawali" based live wallpapers, of which "Nexus Neon Wave", "Nexus Neon Grid" & "Zen Flowers Live" works, but "Ocean Ruins" and "Galaxy Colors" crashes.

Though, I am no expert at reading logs, but considering repeated lines, observed that most of apps crash when tries to load lib from /data/app-lib/com.name.of.app/name.of.lib.so.

[pstglia]

You're probably right. If all of then crashes with signal 4 (SIGILL "invalid or malformed instruction", as you reported on OpenGL ES lesson 6 ), then there's something  unsupported for your hardware in mesa.

If these apps works correctly on 9.2.1 (the version bundled with Kitkat RC1 released on android-x86.org) then there's some regretion on newer mesa codes or a mistake in my compiling procedure.

Anyway, there's not much I can help at this point. The maximum I can do is compiling a new iso with last mesa version released (by the time of this answer is 10.1.2) and hope it has the fixes for the problems you're facing. 

Better option is wait for android-x86 developers release the next versions. They are very skilled people and will support as much hardware as possible.

Sorry not being able to give more help

Ps: Your vga is a Redwood family (R8xx). Mine is a Aruba/Trinity (APU)

[Somebody]

How do we build this? Do you have a GIT somewhere with the changes?

[Paulo Sergio]

No separated git. Basically applied these changes on android x86 git (mesa changes only).

--

[Somebody]

Need something more to go by. At least a PATCH.

You really should keep your changes hosted on a git server, like github -- its free. It makes it possible to access, understand, and integrate your work.

[Somebody]

It also makes it easier FOR YOU, because it keeps all of your changes organized in a sane manner.

On Thursday, May 22, 2014 11:59:52 AM UTC-4, pstglia wrote:

[Paulo Sergio]

Entire android src is more than 10gb. It would take a long long time to upload it with my connection.

I haven't changed much code. You can get mesa src files I put on my google drive ( https://drive.google.com/uc?id=0BxO6THtB865fanJ3MVdXWUplams&export=download) and replace external/mesa dir by this.

Cheers
Pstglia

[Somebody]

You really need to learn a bit more about git.

You don't need to upload the entire android-x86.

JUST THE REPOSITORIES YOU CHANGE!!!!

Even Android-x86 doesn't have a full copy of AOSP. Just the parts of it that are changed.

What this means, is that you clone MESA, and MANIFEST.

You need to make adjustments to the manifest to instruct it in where the different parts reside.

And for that matter, making your changes in an archive like that loses the entire change history of mesa.

Mesa is 122 MB. Manifest is negligible.

Here, look: https://github.com/lbdroid/android-x86_manifest/commit/633f9fbd8679679d89c0fdb62a4e41d0aac90f48

When you do your repo init, it will look similar to this;

repo init -u https://github.com/lbdroid/android-x86_manifest.git -b kitkat-x86

Don't do that with my repo right now though, since I'm still uploading a bone stock clone of mesa.

[Paulo Sergio]

I will consider this. Thanks for the explanation

[Somebody]

Just as an example of how difficult this becomes, I'm looking at your patched 10.1.1, and trying to find its diff against the official 10.1.1 release. The diff file created by the patch command is over 180 THOUSAND lines long. Obviously, you didn't use the official 10.1.1 release from mesa....

[Somebody]

https://github.com/lbdroid/android-x86_external_mesa/commit/bf2e77ca9e569ab2dde6690dd1e18dbad23ec60e

That should hopefully work.

[Francesco Diaz]

my apu a8 5600k, with the iso with mesa 10.1 run..., but certain apps like dolphin, chrome flashfox, suddenly crash

El martes, 25 de marzo de 2014 02:47:24 UTC+1, pstglia escribió:

Hi friends,

I'm running this version with modesetting disabled on my config (AMD A10 5800K => Radeon 7660D IGP). 

Works fine, but I'd like to use radeon + dri to have acceleration support (videos and 3d stuff). 

 (  Asking too much, isn't it? :D  )

When trying to, I got some dmesg warnings/errors:

  * radeon bo/va conflicts with va/bo

  * segfault at xx ip xxxxxx sp yyyyy error 4 in libGLES_mesa.so

Before entering graphic mode (used DEBUG=2), the pseudo file vma in /sys/kernel/debug/dri/0/vma shows some addresses. After typing exit to enter graphic mode, this same file shows zero addresses. Could segfaults on libGLES_mesa be related to this?

Apparently, my hardware is not supported by this compiling of mesa for Android (GLES/egl maybe). Someone could just confirm this? If so, Is there any chance of support in the future?

* * *

Note (not important - read this just if you have time and patience ): I downloaded sources and compiled it replacing Mesa to a newer version (downloaded 10.0.4 and changed some Android.mk files, headers, etc). Also, compiled a newer kernel version ( 3.13.6 - used make oldconfig to get configs from the original kernel in the RC 4.4).

It boots, but the same behavior (just get rid of bo/va conflicts messages). 

These are some debug info on my "frankenstein" compiling :)

# dmesg output

6>[  487.963852] droid.launcher3[7749]: segfault at 0 ip 778750cc sp bfcef0f0 error 4 in libGLES_mesa.so[77821000+650000]

<6>[  488.178988] ndroid.systemui[7821]: segfault at 18 ip 77c42a93 sp bfcef6d0 error 4 in libGLES_mesa.so[77821000+650000]

# logcat output

I/DEBUG   ( 1516): *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** ***

I/DEBUG   ( 1516): Build fingerprint: 'Android-x86/android_x86/x86:4.4.2/KOT49H/eng.paulo.20140309.172714:userdebug/test-keys'

I/DEBUG   ( 1516): Revision: '0'

I/DEBUG   ( 1516): pid: 3189, tid: 3189, name: ndroid.systemui  >>> com.android.systemui <<<

I/DEBUG   ( 1516): signal 11 (SIGSEGV), code 1 (SEGV_MAPERR), fault addr 00000018

E/GRALLOC-KMS( 1518): failed to set crtc (Permission denied) (crtc_id 12, fb_id 59, conn 20, mode 1920x1080)

E/SurfaceFlinger( 1518): error posting framebuffer: -13

I/DEBUG   ( 1516):

I/DEBUG   ( 1516): backtrace:

I/DEBUG   ( 1516): AM write failure (32 / Broken pipe)

I/DEBUG   ( 1516):     #00  pc 0045ba63  /android/system/lib/egl/libGLES_mesa.so

I/DEBUG   ( 1516):     #01  pc 00052186  /android/system/lib/egl/libGLES_mesa.so

I/DEBUG   ( 1516):     #02  pc 0003dcf1  /android/system/lib/egl/libGLES_mesa.so

I/DEBUG   ( 1516):     #03  pc 0003a0b9  /android/system/lib/egl/libGLES_mesa.so

I/DEBUG   ( 1516):     #04  pc 00036330  /android/system/lib/egl/libGLES_mesa.so

I/DEBUG   ( 1516):     #05  pc 00036444  /android/system/lib/egl/libGLES_mesa.so

I/DEBUG   ( 1516):     #06  pc 00039f03  /android/system/lib/egl/libGLES_mesa.so

I/DEBUG   ( 1516):     #07  pc 0003fb92  /android/system/lib/egl/libGLES_mesa.so

I/DEBUG   ( 1516):     #08  pc 000355ed  /android/system/lib/egl/libGLES_mesa.so

I/DEBUG   ( 1516):     #09  pc 004bd530  /android/system/lib/egl/libGLES_mesa.so

I/DEBUG   ( 1516):     #10  pc 00411b54  /android/system/lib/egl/libGLES_mesa.so

I/DEBUG   ( 1516):     #11  pc 00411c4c  /android/system/lib/egl/libGLES_mesa.so

I/DEBUG   ( 1516):     #12  pc 0041fabc  /android/system/lib/egl/libGLES_mesa.so

I/DEBUG   ( 1516):     #13  pc 0040f6e1  /android/system/lib/egl/libGLES_mesa.so

I/DEBUG   ( 1516):     #14  pc 00034bc4  /android/system/lib/egl/libGLES_mesa.so

I/DEBUG   ( 1516):     #15  pc 003f6977  /android/system/lib/egl/libGLES_mesa.so

I/DEBUG   ( 1516):     #16  pc 0023ed19  /android/system/lib/egl/libGLES_mesa.so

I/DEBUG   ( 1516):     #17  pc 0031c4d9  /android/system/lib/egl/libGLES_mesa.so

I/DEBUG   ( 1516):     #18  pc 0031c9d0  /android/system/lib/egl/libGLES_mesa.so

I/DEBUG   ( 1516):     #19  pc 0007bc6b  /system/lib/libandroid_runtime.so

I/DEBUG   ( 1516):     #20  pc 0002a2fb  /system/lib/libdvm.so (dvmPlatformInvoke+79)

I/DEBUG   ( 1516):     #21  pc 00002e3f  <unknown>

I/DEBUG   ( 1516):     #22  pc 00086bf2  /system/lib/libdvm.so (dvmCallJNIMethod(unsigned int const*, JValue*, Method const*, Thread*)+434)

I/DEBUG   ( 1516):     #23  pc 00177618  /system/lib/libdvm.so

I/DEBUG   ( 1516):     #24  pc 00008bfb  <unknown>

I/DEBUG   ( 1516):     #25  pc 0003b7b2  /system/lib/libdvm.so (dvmMterpStd(Thread*)+66)

I/DEBUG   ( 1516):     #26  pc 00036e79  /system/lib/libdvm.so (dvmInterpret(Thread*, Method const*, JValue*)+217)

I/DEBUG   ( 1516):     #27  pc 000bc086  /system/lib/libdvm.so (dvmInvokeMethod(Object*, Method const*, ArrayObject*, ArrayObject*, ClassObject*, bool)+1750)

I/DEBUG   ( 1516):     #28  pc 000d19e0  /system/lib/libdvm.so

I/DEBUG   ( 1516):     #29  pc 00177618  /system/lib/libdvm.so

I/DEBUG   ( 1516):     #30  pc 00008eff  <unknown>

I/DEBUG   ( 1516):     #31  pc 0003b7b2  /system/lib/libdvm.so (dvmMterpStd(Thread*)+66)

I/DEBUG   ( 1516):

I/DEBUG   ( 1516): stack:

I/DEBUG   ( 1516):          bfe62b30  00000000

I/DEBUG   ( 1516):          bfe62b34  00010010

I/DEBUG   ( 1516):          bfe62b38  00000010

I/DEBUG   ( 1516):          bfe62b3c  77e21d70  /android/system/lib/egl/libGLES_mesa.so

I/DEBUG   ( 1516):          bfe62b40  7a8b9020

I/DEBUG   ( 1516):          bfe62b44  7c236380

I/DEBUG   ( 1516):          bfe62b48  40059da9  /system/lib/libc.so

I/DEBUG   ( 1516):          bfe62b4c  400e3fcc  /system/lib/libc.so

I/DEBUG   ( 1516):          bfe62b50  00010010

I/DEBUG   ( 1516):          bfe62b54  7cc81010

I/DEBUG   ( 1516):          bfe62b58  000002c0

I/DEBUG   ( 1516):          bfe62b5c  4005ad3f  /system/lib/libc.so (dlmalloc+351)

I/DEBUG   ( 1516):          bfe62b60  400e8800

I/DEBUG   ( 1516):          bfe62b64  7c237068

I/DEBUG   ( 1516):          bfe62b68  40055d59  /system/lib/libc.so (pthread_mutex_lock+9)

I/DEBUG   ( 1516):          bfe62b6c  77c7ba5d  /android/system/lib/egl/libGLES_mesa.so

I/DEBUG   ( 1516):     #00  bfe62b70  7c236820

I/DEBUG   ( 1516):          bfe62b74  bfe62bac  [stack]

I/DEBUG   ( 1516):          bfe62b78  000000c8

I/DEBUG   ( 1516):          bfe62b7c  77ce753d  /android/system/lib/egl/libGLES_mesa.so

I/DEBUG   ( 1516):          bfe62b80  7c236b05

I/DEBUG   ( 1516):          bfe62b84  0000104c

I/DEBUG   ( 1516):          bfe62b88  000000c2

I/DEBUG   ( 1516):          bfe62b8c  40055f29  /system/lib/libc.so (pthread_mutex_unlock+25)

I/DEBUG   ( 1516):          bfe62b90  00000006

I/DEBUG   ( 1516):          bfe62b94  00002850

I/DEBUG   ( 1516):          bfe62b98  000000c0

I/DEBUG   ( 1516):          bfe62b9c  400e3fcc  /system/lib/libc.so

I/DEBUG   ( 1516):          bfe62ba0  00000001

I/DEBUG   ( 1516):          bfe62ba4  78e88940

I/DEBUG   ( 1516):          bfe62ba8  77cdc149  /android/system/lib/egl/libGLES_mesa.so

I/DEBUG   ( 1516):          bfe62bac  77c68d41  /android/system/lib/egl/libGLES_mesa.so

I/DEBUG   ( 1516):          ........  ........

I/DEBUG   ( 1516):     #01  bfe62bd0  00000000

I/DEBUG   ( 1516):          bfe62bd4  00000000

I/DEBUG   ( 1516):          bfe62bd8  00000000

I/DEBUG   ( 1516):          bfe62bdc  00000000

I/DEBUG   ( 1516):          bfe62be0  00000040

I/DEBUG   ( 1516):          bfe62be4  00000040

I/DEBUG   ( 1516):          bfe62be8  00000086

I/DEBUG   ( 1516):          bfe62bec  7cc81010

I/DEBUG   ( 1516):          bfe62bf0  00000000

I/DEBUG   ( 1516):          bfe62bf4  00000000

I/DEBUG   ( 1516):          bfe62bf8  00000003

I/DEBUG   ( 1516):          bfe62bfc  00000000

I/DEBUG   ( 1516):          bfe62c00  7cc81010

I/DEBUG   ( 1516):          bfe62c04  00000001

I/DEBUG   ( 1516):          bfe62c08  00000068

I/DEBUG   ( 1516):          bfe62c0c  00000034

I/DEBUG   ( 1516):          ........  ........

[Paulo Sergio]

Yes I did :)
Downloaded from mesa3d.org

I just changed what I described in the previous posts (Android.mk files and some header files).
Strange your diff out show so many changes.

I'll do the same to confirm when have time

[pstglia]

Probably you already figured out that I send to you Mesa src 10.0.4 instead of 10.1.1. My fault 

Here's the correct link for 10.1.1:

https://drive.google.com/file/d/0BxO6THtB865fMTRhV0NIQnZiaEE/edit?usp=sharing

diff -r between release src and modified returns less than 60 lines

Thanks for creating this git.

[Somebody]

I already found that link from further up the thread.

Compared to official mesa 10.1.1 http://cgit.freedesktop.org/mesa/mesa/snapshot/mesa-10.1.1.tar.gz

$ diff -Naur mesa-10.1.1 mesa-10.1.1_pstglia | wc -l

180617

$ du -sh mesa*

54M mesa-10.1.1

60M mesa-10.1.1_pstglia

So... somehow, your source is much *larger*.

Anyhow, whatever the cause of the source differences, the point is that keeping everything neatly organized in a git solves it.

[Somebody]

Anybody who wants to is free to make a build off my git.

Its working code, mesa 10.1.4

repo init -u https://github.com/lbdroid/android-x86_manifest.git -b kitkat-x86

repo sync

. build/envsetup.sh

lunch 6

m -jX iso_img

(where X is the SMALLER of the number of physical CPU cores you have, and the number of cores you WANT TO USE).


You can see what has actually been done by looking in these two repos. The second one is the more important.
https://github.com/lbdroid/android-x86_manifest/commits/kitkat-x86

https://github.com/lbdroid/android-x86_external_mesa/commits/android_10.1.4

[dscm]

ah yes but....working code is a relative term no????

it's radeon specific I assume.....??

i915, i965 are yet to be debugged/fixed??

[rbg]

did a build using your mesa 10.1.4

uploaded for test here....https://drive.google.com/folderview?id=0B2evBKvIpdYkSkVBWkxMSWtZWk0&usp=sharing

On Friday, May 23, 2014 4:10:45 PM UTC-4, Somebody wrote:

[Somebody]

No, "working" is not a relative term. It is pretty specific and means that it is predominantly functional and can be used normally.

[Somebody]

As far as i915/i965, they should work as far as the i915g driver can take them.

If you need the classic drivers and want to patch them up in mesa, go for it and send me a pull request with the changes needed.

On Friday, May 23, 2014 5:28:15 PM UTC-4, dscm wrote:

[Chih-Wei Huang]

2014-05-26 20:30 GMT+08:00 Somebody <defines...@gmail.com>:
> As far as i915/i965, they should work as far as the i915g driver can take
> them.

i965 and i915 are different drivers.
i915g is the gallium replacement of i915.
Actually I have tried to use it in jb-x86 4.3 (the 0725 release)
However, it's too slow that most games are unplayable.
So I gave it up and switch back to i915.

There is also a gallium replacement of i965
which is call ilo (developed by olv).
But it's not mature enough as my last test.

Anyway, I'd like to hear test updates
based on the newer mesa branch.

[Somebody]

I don't have any intel hardware to test with.

[Meerio]

i've an atom n450 and gma3150, your iso is too slow compared to rc2, i wasn't even able to access tablet's info to check mesa version.

[rbg]

my test of the build on various intel processors was basically unusable...I'd assume because of
the i915/i965 drivers you referenced.....

using vesa mode was reasonable...so guess it had to do with acceleration...

On Monday, May 26, 2014 11:27:34 AM UTC-4, Chih-Wei Huang wrote:

[rbg]

if you read the thread on this you'll see it's geared to radeon NOT intel....so that's no suprise

the intel stuff needs to be debugged/updated before it's going to work correctly for their chipsets...IMO

[Meerio]

I've read the thread, I was trying to help Chih-Wei and Somebody who were asking if on intel hardware this build was funcitional, sorry if i haven't specified

[rbg]

ya see that now...always in too much of a hurry am I...

try as I did on my n450 with vesa mode....little more reasonable but still not good...

[Somebody]

We probably should be looking to get mesa 10.2 working. It is actually very exciting, SPECIFICALLY for android-x86.... due to the openmax state tracker.

Android uses openmax for video encode/decode acceleration, so this state tracker would be directly usable by android for radeon UVD.

[Daniel Labella de la Cruz]

Great news!, how can I help? I have an amd e-350 palm.

[Mauro Rossi]

Hi,

while trying to build ISO with patched libdrm 2.4.53 and mesa 10.2.3 with Nouveau gralloc_drm code by pstglia,
I discovered that also radeon code was patched in libdrm, thus enabling a build with mesa 10.2.3 also for radeon.

It seams to work on my Radeon 48xx as per former mesa 10.1.x

A question for Somebody: how can we check if openmax is working properly?

Here is the ISO, for everyone interested in giving it a try:
 
https://drive.google.com/file/d/0B_OFHiIqgpSFUXg1eWhCOUZPeWM/edit?usp=sharing

Mauro

[Somebody]

If you want to check if openmax is working properly, this is how;

IT'S NOT.

I said that its exciting, not that its an instant fix.

[Francesco Diaz]

For me , with  aradeon 7560D, google chrome crash every time I try to see a movie, firefox =, :(

[pstglia]

For me , with  aradeon 7560D, google chrome crash every time I try to see a movie, firefox =, :

Does the same happens with 10.1 mesa or with Android-x86 4.4 r1 (released today)?

# Android-x86 Build with Mesa 10.1.5

https://drive.google.com/file/d/0BxO6THtB865fSHFoVEVjdkZjYms/edit?usp=sharing

 

# Android-x86 r1 (released 2014-08-08)

http://sourceforge.net/projects/android-x86/files/Release%204.4/android-x86-4.4-r1.iso/download

[Andromalius dex]

  The latest amd gpus will horribly bug out if they are using  D3D9Ex  because the newest gpu's have support for mantle
being the mantle support is bugging out the D3D9Ex hardware support  i recomend shutting anything off on the iso that links in anyway to D3D9Ex
 

[pstglia]

Hi

  The latest amd gpus will horribly bug out if they are using  D3D9Ex  because the newest gpu's have support for mantle
being the mantle support is bugging out the D3D9Ex hardware support  i recomend shutting anything off on the iso that links in anyway to D3D9Ex
 

D3D9Ex is part of a Windows API for graphics (DirectX). It's not used by Android-x86.

Android-x86 (as Linux) uses Mesa, which is an OpenGL implementation.

The main reason latest AMD GPU's doesn't work today on Android-x86 (the CGN ones, Kanibi, Kavery, etc) is some coding changes on Mesa. For this reason, radeonsi (the driver which supports these GPUs) is disabled on Android-x86

Also, some minor kernel changes may be required.

Some guys are working to add this support. Until this is done, you have to wait. 

However, you can still use nomodeset /  vesa mode to boot Android (without hw accel)

Regards,

pstglia

[Cristian Magadan]

Hi pstglia,

For me is Crashing too, Google Chrome, Firefox, when trying to set animated wallpapers.. I try 10.1 mesa, 10.2.3 mesa and the new r1...

My Video Card is NVIDIA GTX650 Ti.

regards,

Cristian

[pstglia]

As Radeon cgn cards, Nvidia ones are not supported yet.
Some people ( Chih Wei, Mauro Rossi, Emil Velikov and me) are trying to include enable it.

Have you tried our custom builds from Nouveau thread?
https://drive.google.com/file/d/0BxO6THtB865fRmN4VE5HR1B5dDg/edit?usp=sharing

They still have bugs, long delays and crashes eventually, but some apps are already working.

[Francesco Diaz]

The problem is with the latest r1 and with the latest iso with mesa 10.2... published in the forum :(

[Francesco Diaz]

would be fine if we have an iso with mesa 10.3