[Graphic issues] AMD Chipset RS690M (with radeon x1250) Fujitsu Futro S550 ThinClient Computer

[Patrick Wirth]

Hello everyone!

I have a problem with the android-x86 project and one of my computers:
The Fujitsu Futro S550 is a silent, small and efficient thinclient workstation designed to use it only as a terminal for a bigger network-pc via remote desktop.
So it has nothing powerful:
1Ghz CPU and a Radeon X1250 graphic.
But after an upgrade to a 1,2ghz dualcore and 2 gb ram, i really want to use it as a mediastation!

Windows 7 runs pretty nice, but for youtube is the cpu still too slow.
I could choose a faster one, but the cooler is not designed for such a powerful cpu!
There are no fans!

_____Now to the main problem in android_____
Android runs like every other os like windows 7/10, ubuntu, lubuntu... but only android has graphic issues every time.

The device has only a dvi connector, so I need a adapter for my vga monitor.
In linux it detects now an internal "laptop" monitor and a external one (my monitor).
This is a general issue in linux, not in windows.

I have to enable 

video=LVDS-1:d

in the grub menu.

Is this a driver problem and can I fix this or will it be fixed in the future?
Android would be the perfect OS for my mediastation!

In the attachments are some pictures or better watch this video:
https://youtu.be/0tDdhr8UJaE

Thanks

[Patrick Wirth]

I´ve added a logcat.

11-26 08:02:55.086 E/GLConsumer( 1167): Failed to create image. size=1280x976 st=1280 usage=0x933 fmt=2

Could this be the problem?
Ignore the wifi errors, I didn´t connect a wifi device (wifi usb stick)

[Patrick Wirth]

Is there no hope to get the graphics work correct ?

[Mauro Rossi]

Il giorno domenica 27 novembre 2016 23:35:45 UTC+1, Patrick Wirth ha scritto:

Is there no hope to get the graphics work correct ?

 Hi Patrick,

a similar issue was reported by Abu Al,
the full diagonal artifacts and mouse cursor shadow are unrelated to the error 0x3003 (which may be due to missing support of RGBX format for external images, affecting also other drivers, with the exception of Intel's)

Tearing issue seams to be exactly the one that was affecting xorg, which (when solved) required to set a specific vsync parameter in xorg config,
but on EGL was apparently never solved.

Even if I could observe these similarities, implementing vsync/swap buffer correctly is another story, it requires to solve a mesa driver problem.
My suggestion is to file a bug to Mesa Gallium/Radeon/R300 radeon with images/video link and dmesg, logcat.

In this way the Bug will be at least triaged by R300 driver maintainers.
Mauro

[Patrick Wirth]

Thank you for your answer.
As you said, I will collect the dmesg, logcat is already here. 
And where should I post the files? Inside this post?

And why does the gallium driver work in linux, but not in android?
Btw. I tried to disable HWA by adding HWACC=0 into the command line.
Nothing has changed.
 

[Mauro Rossi]

Il giorno lunedì 28 novembre 2016 02:05:54 UTC+1, Patrick Wirth ha scritto:

Am Montag, 28. November 2016 01:42:47 UTC+1 schrieb Mauro Rossi:

Il giorno domenica 27 novembre 2016 23:35:45 UTC+1, Patrick Wirth ha scritto:

Is there no hope to get the graphics work correct ?

 Hi Patrick,

a similar issue was reported by Abu Al,
the full diagonal artifacts and mouse cursor shadow are unrelated to the error 0x3003 (which may be due to missing support of RGBX format for external images, affecting also other drivers, with the exception of Intel's)

Tearing issue seams to be exactly the one that was affecting xorg, which (when solved) required to set a specific vsync parameter in xorg config,
but on EGL was apparently never solved.

Even if I could observe these similarities, implementing vsync/swap buffer correctly is another story, it requires to solve a mesa driver problem.
My suggestion is to file a bug to Mesa Gallium/Radeon/R300 radeon with images/video link and dmesg, logcat.

In this way the Bug will be at least triaged by R300 driver maintainers.
Mauro

Thank you for your answer.
As you said, I will collect the dmesg, logcat is already here. 

Better to use full logcat.txt and also output of following command is  is useful.

dumpsys SurfaceFlinger > dumpsys.txt

 

And where should I post the files? Inside this post?

https://bugs.freedesktop.org/enter_bug.cgi

 

And why does the gallium driver work in linux, but not in android?

My guess is some kind of fix/workaround was applied for Xorg (glx) and not in mesa specific android code.

In my partial understanding of the two visible problems: 

1) the shadowing mouse cursor make me thing that mouse is HW accelerated but the issue is in the scanout/vsync code

2) the diagonal or better full screen triangulare shape flashing is symptom that during screen refresh in some scanouts the correct buffer is used,

in others not

Btw. I tried to disable HWA by adding HWACC=0 into the command line.

Which command line? Or is it a grub line entry?

 

Nothing has changed.

If in non accelerated path (please try also LiveCD  VESA menu entry) 

the full screen triangular artifact is still there this info will be useful when opening the bug.

In other similar bugs submissions people refferred that the issue appeared with introduction of KMS,

so it is something which may be very ancient, but the Android bugs is not there and it is not a duplicate.

Please consider that any patches that may be proposed by developers may need to be tested on kitkat-x86 only,

this will require to build mesa 13.1.0devel for kitkat-x86 ( I can help with that when needed)

Mauro

[Mauro Rossi]

Better to use full logcat.txt and also output of following command is  is useful.

dumpsys SurfaceFlinger > dumpsys.txt

 

And where should I post the files? Inside this post?

https://bugs.freedesktop.org/enter_bug.cgi

 

And why does the gallium driver work in linux, but not in android?

My guess is some kind of fix/workaround was applied for Xorg (glx) and not in mesa specific android code.

 

Hi, I don't know if the issue is exactly the one described here:

https://bugs.freedesktop.org/show_bug.cgi?id=69076

but it would make sense to try mesa 13.0 branch, which includes this commit:

https://cgit.freedesktop.org/mesa/mesa/commit/?id=02675622b02742960678c438f1b239321c075f50

Futro S550 AMD CPUs should have SSE3 and with SSSE3 emulation should boot kitkat-x86,

while, I'm 100% not sure, but lollipop-x86 or later should not even boot with 32 bit image.

If that's the case I will try to build a kitkat-x86 iso with mesa 13.0.2, that will most probably lack nouveau and swrast drivers,

in a matter of some days, just for the sake of RS690 testing.

Mauro

[Mauro Rossi]

Futro S550 AMD CPUs should have SSE3 and with SSSE3 emulation should boot kitkat-x86,

while, I'm 100% not sure, but lollipop-x86 or later should not even boot with 32 bit image.

If that's the case I will try to build a kitkat-x86 iso with mesa 13.0.2, that will most probably lack nouveau and swrast drivers,

in a matter of some days, just for the sake of RS690 testing.

Mauro

Hi Patrick,

mesa 13.0.2 was possible to build (with several changes) but the resulting iso was not working.

After having discovered that the r300g patch had been backported to mesa 12.0.5,

I built kitkat-x86 with mesa 12.0.5 (without swrast)

It's here: https://drive.google.com/drive/folders/0B_OFHiIqgpSFelZpOUxtZlVMLWs?usp=sharing 

Would you like to give it a try to this experiment, to see if it solves the RS690 triangle flickering in kitkat-x86?

Mauro

[Patrick Wirth]

Hello,
Thank you for your work.
I will try it now...
Sorry for the late answer, didn´t receive a email notification...
And:
The AMD Athlon 64 X2 (upgraded by me) can boot up all x86 images without problems, also the new marshmallow images.
I only have problems with Android x64 and Android Tv on all my pc hardware (including my core2quad pc)
Works just fine on my laptop with newer BIOS and core i5

I will give feedback soon.

[Patrick Wirth]

I installed the build on the 1GB CF (300mb space left lol) and tried some things:

The issue itself is fixed now but the systemui is very laggy.
It´s also impossible to play HD video files. 480p works just fine.

 

[Mauro Rossi]

I also experienced slowliness and GUI instability when installing on CF

If you can boot 6.0r2 x86 image, now it has mesa 13.0.3
So triangle flickering should be gone
Mauro

[Patrick Wirth]

So the laggy system ui is caused by the cf card? 
It is mounted as IDE drive and with x333 speed so this is a little bit confusing.
Due to this, it´s nothing special to install windows on it, if i would have enough space on my 1 gb card...

 Thank you so much for your help with my problem :)