[Intel-gfx] FW: (EXT-15760) [I+A] DRI_PRIME=1 glxgears tearing [8086:5917][1002:6900]

Wang, Quanxian quanxian.wang at intel.com
Wed Aug 30 04:39:34 UTC 2017 

Hi, All

We got a question from Dell Customer about Intel Graphics.
Anyone could help answer that? It is related with shared 3D buffer cache mechanism between Intel and AMD.

Thanks you in advance.

https://bugs.freedesktop.org/show_bug.cgi?id=101691#c37
Comment #37 from AMD engineer
I wonder if it could be related to the caching attributes of the system memory pages being shared between the GPUs. AFAICT the amdgpu driver should end up treating them as non-cacheable, and correspondingly program the AMD GPU not to participate in cache coherency protocol for them. How does the i915 driver treat the pages of imported buffer objects?

Background:
DRM_PRIME=1, glxgears corruption with compiz in KBL-U platform which has integrated Intel Graphics and discrete AMD graphics card.
AC mode corruption(Power). DC mode works (Battery)

Thanks

Regards

Quanxian Wang
From: Alex.Zhang1 at Dell.com [mailto:Alex.Zhang1 at Dell.com]
Sent: Wednesday, August 30, 2017 10:52 AM
To: Wang, Quanxian <quanxian.wang at intel.com>; David.D.Wang at dell.com
Cc: emily.chien at canonical.com; chih.ho at canonical.com; Zorro.Zhang at dell.com; Darren.Powell at amd.com; Tim.Writer at amd.com; Steven.Yen at amd.com
Subject: RE: (EXT-15760) [I+A] DRI_PRIME=1 glxgears tearing [8086:5917][1002:6900]

Dell - Internal Use - Confidential
Hi Quanxian,
  I also want to ask one question, how many framebuffer you create after the driver finish loading, about 3d window content, do you get the source just from one buffer?

From: Zhang1, Alex
Sent: Wednesday, August 30, 2017 10:37 AM
To: 'Wang, Quanxian' <quanxian.wang at intel.com<mailto:quanxian.wang at intel.com>>; Wang, David D <David_D_Wang at dell.com<mailto:David_D_Wang at dell.com>>
Cc: 'emily.chien at canonical.com' <emily.chien at canonical.com<mailto:emily.chien at canonical.com>>; 'chih.ho at canonical.com' <chih.ho at canonical.com<mailto:chih.ho at canonical.com>>; Zhang, Zorro <Zorro_Zhang at Dell.com<mailto:Zorro_Zhang at Dell.com>>; 'Powell, Darren' <Darren.Powell at amd.com<mailto:Darren.Powell at amd.com>>; 'Writer, Tim' <Tim.Writer at amd.com<mailto:Tim.Writer at amd.com>>; 'Yen, Steven' <Steven.Yen at amd.com<mailto:Steven.Yen at amd.com>>
Subject: RE: (EXT-15760) [I+A] DRI_PRIME=1 glxgears tearing [8086:5917][1002:6900]

Dell - Internal Use - Confidential
Remove AMD,
Hi Quanxian,
  Please reply https://bugs.freedesktop.org/show_bug.cgi?id=101691#c37
comment 37.
Thanks

From: Wang, Quanxian [mailto:quanxian.wang at intel.com]
Sent: Friday, August 25, 2017 3:21 AM
To: Olsak, Marek <Marek.Olsak at amd.com<mailto:Marek.Olsak at amd.com>>; Wang, Napoleon <Napoleon.Wang at amd.com<mailto:Napoleon.Wang at amd.com>>; Zhang1, Alex <Alex_Zhang1 at Dell.com<mailto:Alex_Zhang1 at Dell.com>>; Daenzer, Michel <Michel.Daenzer at amd.com<mailto:Michel.Daenzer at amd.com>>
Cc: emily.chien at canonical.com<mailto:emily.chien at canonical.com>; chih.ho at canonical.com<mailto:chih.ho at canonical.com>; Zhang, Zorro <Zorro_Zhang at Dell.com<mailto:Zorro_Zhang at Dell.com>>; Powell, Darren <Darren.Powell at amd.com<mailto:Darren.Powell at amd.com>>; Writer, Tim <Tim.Writer at amd.com<mailto:Tim.Writer at amd.com>>; Yen, Steven <Steven.Yen at amd.com<mailto:Steven.Yen at amd.com>>
Subject: RE: (EXT-15760) [I+A] DRI_PRIME=1 glxgears tearing [8086:5917][1002:6900]

No, amdgpu will also use Xserver/Compiz/Kernel. AMD is a discrete card, by default it should be used instead of integrated card. We don’t find the same issue on KBL platform without AMD.

You can refer to 101691’s comment #18 from Timo, and #20 from Ethan. If amdgpu loaded failed, no corruption which make Timo think the bug is fixed. If amdgpu loaded, corruption is back! Ethan’s test proves that. Amdgpu.pm=0. Amdgpu load failed, means in xserver, Integrated card (Intel) takes the role.

Regards
Quanxian Wang

From: Olsak, Marek [mailto:Marek.Olsak at amd.com]
Sent: Friday, August 25, 2017 1:11 AM
To: Wang, Quanxian <quanxian.wang at intel.com<mailto:quanxian.wang at intel.com>>; Wang, Napoleon <Napoleon.Wang at amd.com<mailto:Napoleon.Wang at amd.com>>; Alex.Zhang1 at Dell.com<mailto:Alex.Zhang1 at Dell.com>; Daenzer, Michel <Michel.Daenzer at amd.com<mailto:Michel.Daenzer at amd.com>>
Cc: emily.chien at canonical.com<mailto:emily.chien at canonical.com>; chih.ho at canonical.com<mailto:chih.ho at canonical.com>; Zorro.Zhang at dell.com<mailto:Zorro.Zhang at dell.com>; Powell, Darren <Darren.Powell at amd.com<mailto:Darren.Powell at amd.com>>; Writer, Tim <Tim.Writer at amd.com<mailto:Tim.Writer at amd.com>>; Yen, Steven <Steven.Yen at amd.com<mailto:Steven.Yen at amd.com>>
Subject: Re: (EXT-15760) [I+A] DRI_PRIME=1 glxgears tearing [8086:5917][1002:6900]


Hi,



My understanding is that only the application that has DRI_PRIME=1 runs on amdgpu (which is glxgears). Everything else runs on Intel. Compiz and X run on Intel. amdgpu only renders the contents of the glxgears framebuffer to an off-screen buffer. The amdgpu work ends there. The off-screen buffer is used as a texture by compiz, which runs on Intel and renders the whole desktop. From Intel's point of view, it's a simple texturing task.



Marek

________________________________
From: Wang, Quanxian <quanxian.wang at intel.com<mailto:quanxian.wang at intel.com>>
Sent: Thursday, August 24, 2017 2:14:01 PM
To: Wang, Napoleon; Alex.Zhang1 at Dell.com<mailto:Alex.Zhang1 at Dell.com>; Olsak, Marek; Daenzer, Michel
Cc: emily.chien at canonical.com<mailto:emily.chien at canonical.com>; chih.ho at canonical.com<mailto:chih.ho at canonical.com>; Zorro.Zhang at dell.com<mailto:Zorro.Zhang at dell.com>; Powell, Darren; Writer, Tim; Yen, Steven; Wang, Quanxian
Subject: RE: (EXT-15760) [I+A] DRI_PRIME=1 glxgears tearing [8086:5917][1002:6900]

Hi, All
I draw a design for whole graphic stack. The role of every card is tagged in the picture.

Currently the doubt point is 5th step. 1 and 2 are same output from glxgears. From the testing, they are fine.
5th should have some issues for 3D composition. (the issue should be here).

if possible, could dump the compiz 3D snapshot buffer to take a look.

[cid:image002.jpg at 01D320EE.BFE3F7D0]

From: Wang, Quanxian
Sent: Thursday, August 24, 2017 6:36 PM
To: 'Wang, Napoleon' <Napoleon.Wang at amd.com<mailto:Napoleon.Wang at amd.com>>; Alex.Zhang1 at Dell.com<mailto:Alex.Zhang1 at Dell.com>; Olsak, Marek <Marek.Olsak at amd.com<mailto:Marek.Olsak at amd.com>>; Daenzer, Michel <Michel.Daenzer at amd.com<mailto:Michel.Daenzer at amd.com>>
Cc: emily.chien at canonical.com<mailto:emily.chien at canonical.com>; chih.ho at canonical.com<mailto:chih.ho at canonical.com>; Zorro.Zhang at dell.com<mailto:Zorro.Zhang at dell.com>; Powell, Darren <Darren.Powell at amd.com<mailto:Darren.Powell at amd.com>>; Writer, Tim <Tim.Writer at amd.com<mailto:Tim.Writer at amd.com>>; Yen, Steven <Steven.Yen at amd.com<mailto:Steven.Yen at amd.com>>
Subject: RE: (EXT-15760) [I+A] DRI_PRIME=1 glxgears tearing [8086:5917][1002:6900]

Hi All

Back to this original email.

DRI_PRIME=1 means use AMDGPU for rendering. From the corruption of picture, we found only glxgears corruption, about other parts in the screen, it works normally.

As AMD engineer described, without compiz, glxgears works. With compiz, glxgears corrupted. The problem should happens between glxgear and compiz. The buffer of glxgears is composited wrongly by AMD GPU rendering. This is part of framebuffer of compiz (compositor manager).

Compiz should talk with mesa (amdgpu 3d driver) and then connected AMD kernel driver for rendering if I am not wrong. And then rendering the whole screen which include glxgears parts.

From the beginning to the kernel driver end, does Intel graphics card (Integrated card) take any role?

Thanks

Regards

Quanxian Wang


From: Wang, Napoleon [mailto:Napoleon.Wang at amd.com]
Sent: Wednesday, August 23, 2017 2:51 PM
To: Alex.Zhang1 at Dell.com<mailto:Alex.Zhang1 at Dell.com>; Wang, Quanxian <quanxian.wang at intel.com<mailto:quanxian.wang at intel.com>>; Olsak, Marek <Marek.Olsak at amd.com<mailto:Marek.Olsak at amd.com>>; Daenzer, Michel <Michel.Daenzer at amd.com<mailto:Michel.Daenzer at amd.com>>
Cc: emily.chien at canonical.com<mailto:emily.chien at canonical.com>; chih.ho at canonical.com<mailto:chih.ho at canonical.com>; Zorro.Zhang at dell.com<mailto:Zorro.Zhang at dell.com>; Powell, Darren <Darren.Powell at amd.com<mailto:Darren.Powell at amd.com>>; Writer, Tim <Tim.Writer at amd.com<mailto:Tim.Writer at amd.com>>; Yen, Steven <Steven.Yen at amd.com<mailto:Steven.Yen at amd.com>>
Subject: RE: (EXT-15760) [I+A] DRI_PRIME=1 glxgears tearing [8086:5917][1002:6900]

Loop in Michel and Marek, Tim from our Linux team.

Hi Quanxian


The issue is still there with all below settings(Corruption and Tearing)

If you have further questions you can ask directly with our SW developers



R600_DEBUG=nodcc

R600_DEBUG=nodma

R600_DEBUG=nodma,nodcc

R600_DEBUG=notiling

R600_DEBUG=notiling,nodma



I suppose it's Intel's issue, as the gears with DRI3 works well without compiz.

corruption (as well as the tearing) only occurs with a compositing

manager (compiz), not without. But that only makes a difference on the

Intel GPU side, the AMD GPU side works exactly the same way in both

cases.


Napoleon Wang [Description: Description: Description: green]
Program Management | AMD CE Taiwan
6F, No.3 YuanCyu Street, NanKang Software Park,
Taipei 115, Taiwan
O +(886) 2-3789-3376    M +(886) 939-893-320
[Description: Description: Description: image004]   facebook<https://www.facebook.com/AMD>  |  amd.com<http://www.amd.com/>

From: Alex.Zhang1 at Dell.com<mailto:Alex.Zhang1 at Dell.com> [mailto:Alex.Zhang1 at Dell.com]
Sent: Wednesday, August 23, 2017 2:39 PM
To: Wang, Napoleon <Napoleon.Wang at amd.com<mailto:Napoleon.Wang at amd.com>>; quanxian.wang at intel.com<mailto:quanxian.wang at intel.com>
Cc: emily.chien at canonical.com<mailto:emily.chien at canonical.com>; chih.ho at canonical.com<mailto:chih.ho at canonical.com>; Zorro.Zhang at dell.com<mailto:Zorro.Zhang at dell.com>
Subject: RE: (EXT-15760) [I+A] DRI_PRIME=1 glxgears tearing [8086:5917][1002:6900]

Dell - Internal Use - Confidential
+Intel Quanxian. And Remove other people.
Loop AMD and Intel together to work on this issue.
Hi Quanxian,
  About DRI_PRIME=1 glxgears tearing issue, AMD think this issue is due to Intel driver problem.
Please work together with AMD to dig the root cause out.

Hi Napoleon & Quanxian,
 Loki 13 Armani 13 14 platform should prepare A-rev driver on 9/20, please make sure provide solution before 9/6.
Please co-work with each other, provide Dell daily detail update about this issue. Loop engineers necessary to this mail loop if needed.
Thanks


From: Wang, Napoleon [mailto:Napoleon.Wang at amd.com]
Sent: Wednesday, August 23, 2017 2:06 PM
To: Zhang1, Alex <Alex_Zhang1 at Dell.com<mailto:Alex_Zhang1 at Dell.com>>; emily.chien at canonical.com<mailto:emily.chien at canonical.com>
Cc: Lin, Will <Will.Lin at amd.com<mailto:Will.Lin at amd.com>>; chih.ho at canonical.com<mailto:chih.ho at canonical.com>; Wu, Frasier <Frasier.Wu at amd.com<mailto:Frasier.Wu at amd.com>>; jeffp_wang at compal.com<mailto:jeffp_wang at compal.com>; rabbit_tsao at compal.com<mailto:rabbit_tsao at compal.com>; kevin.lu at canonical.com<mailto:kevin.lu at canonical.com>; Powell, Darren <Darren.Powell at amd.com<mailto:Darren.Powell at amd.com>>; Zhang, Zorro <Zorro_Zhang at Dell.com<mailto:Zorro_Zhang at Dell.com>>; yc.cheng at canonical.com<mailto:yc.cheng at canonical.com>; ethan.hsieh at canonical.com<mailto:ethan.hsieh at canonical.com>; cyrus.lien at canonical.com<mailto:cyrus.lien at canonical.com>
Subject: RE: (EXT-15760) [I+A] DRI_PRIME=1 glxgears tearing [8086:5917][1002:6900]

Hi Alex

UMA is different from Hybrid Graphic or PowerXpress platform.
All display output rendered by Intel not AMD, what is Intel’s investigation that point to AMD issue

[cid:image005.jpg at 01D320EE.BFE3F7D0]


If you look at the updates from Canonical



It turned out the amdgpu.dpm=0 did not fix the issue.



Platform: AMN4-DVT1-C2 (201707-25595)
Graphics Card: Intel [8086:5917] + AMD [1002:6900]
Image: ubuntu-17.04-desktop-amd64.iso
BIOS: 0.2.7

1) tearing issue (run glxgears with full screen)
Cannot reproduce the issue on 17.04.

2) corruption issue (run glxgears with default window size)
----------------- glxinfo ------------------
libGL: pci id for fd 4: 8086:5917, driver i965
libGL: OpenDriver: trying /usr/lib/x86_64-linux-gnu/dri/tls/i965_dri.so
libGL: OpenDriver: trying /usr/lib/x86_64-linux-gnu/dri/i965_dri.so
libGL: Using DRI3 for screen 0
OpenGL renderer string: Mesa DRI Intel(R) Kabylake GT1.5
----------------- DRI_PRIME=1 glxinfo ------------------
libGL: pci id for fd 5: 1002:6900, driver radeonsi
libGL: OpenDriver: trying /usr/lib/x86_64-linux-gnu/dri/tls/radeonsi_dri.so
libGL: OpenDriver: trying /usr/lib/x86_64-linux-gnu/dri/radeonsi_dri.so
libGL: Using DRI3 for screen 0
OpenGL renderer string: Gallium 0.4 on AMD ICELAND (DRM 3.9.0 / 4.10.0-32-generic, LLVM 4.0.0)
---------------------------------------------------------

a. modeset
AC (Power adaptor): Failed
DC (Battery, Unplug Power adaptor): Failed
b. 20-intel.conf + 30-amdgpu.conf
AC (Power adaptor): Failed
DC (Battery, Unplug Power adaptor): Pass


Napoleon Wang [Description: Description: Description: green]
Program Management | AMD CE Taiwan
6F, No.3 YuanCyu Street, NanKang Software Park,
Taipei 115, Taiwan
O +(886) 2-3789-3376    M +(886) 939-893-320
[Description: Description: Description: image004]   facebook<https://www.facebook.com/AMD>  |  amd.com<http://www.amd.com/>

From: Alex.Zhang1 at Dell.com<mailto:Alex.Zhang1 at Dell.com> [mailto:Alex.Zhang1 at Dell.com]
Sent: Wednesday, August 23, 2017 1:35 PM
To: Wang, Napoleon <Napoleon.Wang at amd.com<mailto:Napoleon.Wang at amd.com>>; emily.chien at canonical.com<mailto:emily.chien at canonical.com>
Cc: Lin, Will <Will.Lin at amd.com<mailto:Will.Lin at amd.com>>; chih.ho at canonical.com<mailto:chih.ho at canonical.com>; Wu, Frasier <Frasier.Wu at amd.com<mailto:Frasier.Wu at amd.com>>; jeffp_wang at compal.com<mailto:jeffp_wang at compal.com>; rabbit_tsao at compal.com<mailto:rabbit_tsao at compal.com>; kevin.lu at canonical.com<mailto:kevin.lu at canonical.com>; Powell, Darren <Darren.Powell at amd.com<mailto:Darren.Powell at amd.com>>; Zorro.Zhang at dell.com<mailto:Zorro.Zhang at dell.com>; yc.cheng at canonical.com<mailto:yc.cheng at canonical.com>; ethan.hsieh at canonical.com<mailto:ethan.hsieh at canonical.com>; cyrus.lien at canonical.com<mailto:cyrus.lien at canonical.com>
Subject: RE: (EXT-15760) [I+A] DRI_PRIME=1 glxgears tearing [8086:5917][1002:6900]

Dell - Internal Use - Confidential
Hi Napoleon,
  We use UMA driver, no this issue occur. Intel will refute your conclusion with this argument.
Please explain us the following lifecycle of render output. If necessary I will bring Intel and AMD together.

From: Wang, Napoleon [mailto:Napoleon.Wang at amd.com]
Sent: Tuesday, August 22, 2017 10:26 PM
To: Emily Chien <emily.chien at canonical.com<mailto:emily.chien at canonical.com>>
Cc: Zhang1, Alex <Alex_Zhang1 at Dell.com<mailto:Alex_Zhang1 at Dell.com>>; Lin, Will <Will.Lin at amd.com<mailto:Will.Lin at amd.com>>; chih.ho at canonical.com<mailto:chih.ho at canonical.com>; Wu, Frasier <Frasier.Wu at amd.com<mailto:Frasier.Wu at amd.com>>; jeffp_wang at compal.com<mailto:jeffp_wang at compal.com>; rabbit_tsao at compal.com<mailto:rabbit_tsao at compal.com>; kevin.lu at canonical.com<mailto:kevin.lu at canonical.com>; Powell, Darren <Darren.Powell at amd.com<mailto:Darren.Powell at amd.com>>; Zhang, Zorro <Zorro_Zhang at Dell.com<mailto:Zorro_Zhang at Dell.com>>; YC Cheng <yc.cheng at canonical.com<mailto:yc.cheng at canonical.com>>; Ethan Hsieh <ethan.hsieh at canonical.com<mailto:ethan.hsieh at canonical.com>>; cyrus <cyrus.lien at canonical.com<mailto:cyrus.lien at canonical.com>>
Subject: RE: (EXT-15760) [I+A] DRI_PRIME=1 glxgears tearing [8086:5917][1002:6900]

Hi Emily


The upstream 4.4 kernel is definitely too old to avoid the tearing.

Unless Canonical is backporting the needed PRIME synchronization related changes from the >= 4.9 i915 & amdgpu drivers, that would explain the tearing.





The issue is still there with all below settings.

R600_DEBUG=nodcc

R600_DEBUG=nodma

R600_DEBUG=nodma,nodcc

R600_DEBUG=notiling

R600_DEBUG=notiling,nodma



I suppose it's Intel's issue, as the gears with DRI3 works well without compiz.

corruption (as well as the tearing) only occurs with a compositing

manager (compiz), not without. But that only makes a difference on the

Intel GPU side, the AMD GPU side works exactly the same way in both

cases.






Napoleon Wang [Description: Description: Description: green]
Program Management | AMD CE Taiwan
6F, No.3 YuanCyu Street, NanKang Software Park,
Taipei 115, Taiwan
O +(886) 2-3789-3376    M +(886) 939-893-320
[Description: Description: Description: image004]   facebook<https://www.facebook.com/AMD>  |  amd.com<http://www.amd.com/>

From: Emily Chien [mailto:emily.chien at canonical.com]
Sent: Tuesday, August 22, 2017 6:25 PM
To: Wang, Napoleon <Napoleon.Wang at amd.com<mailto:Napoleon.Wang at amd.com>>
Cc: Alex.Zhang1 at Dell.com<mailto:Alex.Zhang1 at Dell.com>; Lin, Will <Will.Lin at amd.com<mailto:Will.Lin at amd.com>>; chih.ho at canonical.com<mailto:chih.ho at canonical.com>; Wu, Frasier <Frasier.Wu at amd.com<mailto:Frasier.Wu at amd.com>>; jeffp_wang at compal.com<mailto:jeffp_wang at compal.com>; rabbit_tsao at compal.com<mailto:rabbit_tsao at compal.com>; kevin.lu at canonical.com<mailto:kevin.lu at canonical.com>; Powell, Darren <Darren.Powell at amd.com<mailto:Darren.Powell at amd.com>>; Zorro.Zhang at dell.com<mailto:Zorro.Zhang at dell.com>; YC Cheng <yc.cheng at canonical.com<mailto:yc.cheng at canonical.com>>; Ethan Hsieh <ethan.hsieh at canonical.com<mailto:ethan.hsieh at canonical.com>>; cyrus <cyrus.lien at canonical.com<mailto:cyrus.lien at canonical.com>>
Subject: Re: (EXT-15760) [I+A] DRI_PRIME=1 glxgears tearing [8086:5917][1002:6900]

Hi Napoleon,

To clarify a bit further, when we mentioned tearing issue is gone with 4.10, 4.10 is using amdgpu open source driver. 4.4 is using amd dkms driver.

4.4 kernel + amdgpu-pro, 17.10-414273 (dkms) => reproduced tearing
4.10 kernel (4.10 built-in amdgpu driver) => tearing issue not reproduced

It will need AMD to investigate further what is the difference between the dkms driver and the 4.10 built-in driver.

Thank you.
Emily


On Tue, Aug 22, 2017 at 5:23 PM, Emily Chien <emily.chien at canonical.com<mailto:emily.chien at canonical.com>> wrote:
Hi Napoleon,

As Timo put, tearing issue is gone with 4.10; while the corruption issue still remained.

If you already have a Launchpad account, I could subscribe you to these two issues for transparency.
Can you share what is your launchpad account id?

Corruption issue (original reported as tearing) [EXT-15760
​]​
 - https://bugs.launchpad.net/somerville/+bug/1698268

Tearing issue - https://bugs.launchpad.net/somerville/+bug/1711328

FYI. Timo is our HWE engineer and his comments is based on our input on launchpad bugs.

Thank you.
Emily



On Tue, Aug 22, 2017 at 5:07 PM, Wang, Napoleon <Napoleon.Wang at amd.com<mailto:Napoleon.Wang at amd.com>> wrote:
Hi Emily

Can you confirm if tearing is gone at Kernel 4.10 ?

Napoleon Wang [Description: Description: Description: green]
Program Management | AMD CE Taiwan
6F, No.3 YuanCyu Street, NanKang Software Park,
Taipei 115, Taiwan
O +(886) 2-3789-3376<tel:+886%202%203789%203376>    M +(886) 939-893-320<tel:+886%20939%20893%20320>
[Description: Description: Description: image004]   facebook<https://www.facebook.com/AMD>  |  amd.com<http://www.amd.com/>

From: Emily Chien [mailto:emily.chien at canonical.com<mailto:emily.chien at canonical.com>]
Sent: Tuesday, August 22, 2017 5:03 PM
To: Wang, Napoleon <Napoleon.Wang at amd.com<mailto:Napoleon.Wang at amd.com>>
Cc: Alex.Zhang1 at Dell.com<mailto:Alex.Zhang1 at Dell.com>; Lin, Will <Will.Lin at amd.com<mailto:Will.Lin at amd.com>>; chih.ho at canonical.com<mailto:chih.ho at canonical.com>; Wu, Frasier <Frasier.Wu at amd.com<mailto:Frasier.Wu at amd.com>>; jeffp_wang at compal.com<mailto:jeffp_wang at compal.com>; rabbit_tsao at compal.com<mailto:rabbit_tsao at compal.com>; kevin.lu at canonical.com<mailto:kevin.lu at canonical.com>; Powell, Darren <Darren.Powell at amd.com<mailto:Darren.Powell at amd.com>>; Zorro.Zhang at dell.com<mailto:Zorro.Zhang at dell.com>; YC Cheng <yc.cheng at canonical.com<mailto:yc.cheng at canonical.com>>

Subject: Re: (EXT-15760) [I+A] DRI_PRIME=1 glxgears tearing [8086:5917][1002:6900]

Hi Napoleon,

We are shipping with 4.4 kernel for Loki/Armani platforms.

Thank you.
Emily

On Tue, Aug 22, 2017 at 4:59 PM, Wang, Napoleon <Napoleon.Wang at amd.com<mailto:Napoleon.Wang at amd.com>> wrote:
Hi Emily/Alex

https://bugs.freedesktop.org/show_bug.cgi?id=101691#c27

[cid:image008.png at 01D320EE.BFE3F7D0]

Tearing is gone with Kernel 4.10, which kernel are you plan to ship with Dell Loki/Armani projects?

Napoleon Wang [Description: Description: Description: green]
Program Management | AMD CE Taiwan
6F, No.3 YuanCyu Street, NanKang Software Park,
Taipei 115, Taiwan
O +(886) 2-3789-3376<tel:+886%202%203789%203376>    M +(886) 939-893-320<tel:+886%20939%20893%20320>
[Description: Description: Description: image004]   facebook<https://www.facebook.com/AMD>  |  amd.com<http://www.amd.com/>

From: Emily Chien [mailto:emily.chien at canonical.com<mailto:emily.chien at canonical.com>]
Sent: Wednesday, August 9, 2017 4:53 PM
To: Wang, Napoleon <Napoleon.Wang at amd.com<mailto:Napoleon.Wang at amd.com>>
Cc: Alex.Zhang1 at Dell.com<mailto:Alex.Zhang1 at Dell.com>; Lin, Will <Will.Lin at amd.com<mailto:Will.Lin at amd.com>>; chih.ho at canonical.com<mailto:chih.ho at canonical.com>; Wu, Frasier <Frasier.Wu at amd.com<mailto:Frasier.Wu at amd.com>>; jeffp_wang at compal.com<mailto:jeffp_wang at compal.com>; rabbit_tsao at compal.com<mailto:rabbit_tsao at compal.com>; kevin.lu at canonical.com<mailto:kevin.lu at canonical.com>; Powell, Darren <Darren.Powell at amd.com<mailto:Darren.Powell at amd.com>>; Zorro.Zhang at dell.com<mailto:Zorro.Zhang at dell.com>
Subject: Re: (EXT-15760) [I+A] DRI_PRIME=1 glxgears tearing [8086:5917][1002:6900]

Hi Napoleon,

Is there update on this issue from your SW team?

Thank you.
Emily

On Thu, Aug 3, 2017 at 1:59 PM, Wang, Napoleon <Napoleon.Wang at amd.com<mailto:Napoleon.Wang at amd.com>> wrote:
Hi Alex

I will follow up with SW team tonight

Napoleon Wang [Description: Description: Description: green]
Program Management | AMD CE Taiwan
6F, No.3 YuanCyu Street, NanKang Software Park,
Taipei 115, Taiwan
O +(886) 2-3789-3376<tel:+886%202%203789%203376>    M +(886) 939-893-320<tel:+886%20939%20893%20320>
[Description: Description: Description: image004]   facebook<https://www.facebook.com/AMD>  |  amd.com<http://www.amd.com/>

From: Alex.Zhang1 at Dell.com<mailto:Alex.Zhang1 at Dell.com> [mailto:Alex.Zhang1 at Dell.com<mailto:Alex.Zhang1 at Dell.com>]
Sent: Thursday, August 3, 2017 1:49 PM
To: Wang, Napoleon <Napoleon.Wang at amd.com<mailto:Napoleon.Wang at amd.com>>; NoReply-JIRA <NoReply-JIRA at amd.com<mailto:NoReply-JIRA at amd.com>>; Lin, Will <Will.Lin at amd.com<mailto:Will.Lin at amd.com>>; chih.ho at canonical.com<mailto:chih.ho at canonical.com>; kent.lin at canonical.com<mailto:kent.lin at canonical.com>; Wu, Frasier <Frasier.Wu at amd.com<mailto:Frasier.Wu at amd.com>>; jeffp_wang at compal.com<mailto:jeffp_wang at compal.com>; rabbit_tsao at compal.com<mailto:rabbit_tsao at compal.com>; emily.chien at canonical.com<mailto:emily.chien at canonical.com>; kevin.lu at canonical.com<mailto:kevin.lu at canonical.com>; Powell, Darren <Darren.Powell at amd.com<mailto:Darren.Powell at amd.com>>
Cc: Zorro.Zhang at dell.com<mailto:Zorro.Zhang at dell.com>
Subject: RE: (EXT-15760) [I+A] DRI_PRIME=1 glxgears tearing [8086:5917][1002:6900]

Dell - Internal Use - Confidential
Hi Napoleon,
  Is there any reply from your team?

From: Zhang1, Alex
Sent: Monday, July 31, 2017 5:54 PM
To: 'Wang, Napoleon' <Napoleon.Wang at amd.com<mailto:Napoleon.Wang at amd.com>>; NoReply-JIRA <NoReply-JIRA at amd.com<mailto:NoReply-JIRA at amd.com>>; Lin, Will <Will.Lin at amd.com<mailto:Will.Lin at amd.com>>; chih.ho at canonical.com<mailto:chih.ho at canonical.com>; kent.lin at canonical.com<mailto:kent.lin at canonical.com>; Wu, Frasier <Frasier.Wu at amd.com<mailto:Frasier.Wu at amd.com>>; jeffp_wang at compal.com<mailto:jeffp_wang at compal.com>; rabbit_tsao at compal.com<mailto:rabbit_tsao at compal.com>; emily.chien at canonical.com<mailto:emily.chien at canonical.com>; kevin.lu at canonical.com<mailto:kevin.lu at canonical.com>; Powell, Darren <Darren.Powell at amd.com<mailto:Darren.Powell at amd.com>>
Cc: Zhang, Zorro <Zorro_Zhang at Dell.com<mailto:Zorro_Zhang at Dell.com>>
Subject: RE: (EXT-15760) [I+A] DRI_PRIME=1 glxgears tearing [8086:5917][1002:6900]

Dell - Internal Use - Confidential
Thanks Napoleon.
As our schedule plan, we will release final driver version on 8/10, so it is appreciated if you can provide a workable solution before 8/8, so we can have an internal test.

From: Wang, Napoleon [mailto:Napoleon.Wang at amd.com]
Sent: Monday, July 31, 2017 5:37 PM
To: Zhang1, Alex <Alex_Zhang1 at Dell.com<mailto:Alex_Zhang1 at Dell.com>>; NoReply-JIRA <NoReply-JIRA at amd.com<mailto:NoReply-JIRA at amd.com>>; Lin, Will <Will.Lin at amd.com<mailto:Will.Lin at amd.com>>; chih.ho at canonical.com<mailto:chih.ho at canonical.com>; kent.lin at canonical.com<mailto:kent.lin at canonical.com>; Wu, Frasier <Frasier.Wu at amd.com<mailto:Frasier.Wu at amd.com>>; jeffp_wang at compal.com<mailto:jeffp_wang at compal.com>; rabbit_tsao at compal.com<mailto:rabbit_tsao at compal.com>; emily.chien at canonical.com<mailto:emily.chien at canonical.com>; kevin.lu at canonical.com<mailto:kevin.lu at canonical.com>; Powell, Darren <Darren.Powell at amd.com<mailto:Darren.Powell at amd.com>>
Cc: Zhang, Zorro <Zorro_Zhang at Dell.com<mailto:Zorro_Zhang at Dell.com>>
Subject: RE: (EXT-15760) [I+A] DRI_PRIME=1 glxgears tearing [8086:5917][1002:6900]

Hi Alex

I will check with our Internal SW team

Napoleon Wang [Description: Description: Description: green]
Program Management | AMD CE Taiwan
6F, No.3 YuanCyu Street, NanKang Software Park,
Taipei 115, Taiwan
O +(886) 2-3789-3376<tel:+886%202%203789%203376>    M +(886) 939-893-320<tel:+886%20939%20893%20320>
[Description: Description: Description: image004]   facebook<https://www.facebook.com/AMD>  |  amd.com<http://www.amd.com/>

From: Alex.Zhang1 at Dell.com<mailto:Alex.Zhang1 at Dell.com> [mailto:Alex.Zhang1 at Dell.com]
Sent: Monday, July 31, 2017 3:48 PM
To: NoReply-JIRA <NoReply-JIRA at amd.com<mailto:NoReply-JIRA at amd.com>>; Lin, Will <Will.Lin at amd.com<mailto:Will.Lin at amd.com>>; chih.ho at canonical.com<mailto:chih.ho at canonical.com>; Wang, Napoleon <Napoleon.Wang at amd.com<mailto:Napoleon.Wang at amd.com>>; kent.lin at canonical.com<mailto:kent.lin at canonical.com>; Wu, Frasier <Frasier.Wu at amd.com<mailto:Frasier.Wu at amd.com>>; jeffp_wang at compal.com<mailto:jeffp_wang at compal.com>; rabbit_tsao at compal.com<mailto:rabbit_tsao at compal.com>; emily.chien at canonical.com<mailto:emily.chien at canonical.com>; kevin.lu at canonical.com<mailto:kevin.lu at canonical.com>; Powell, Darren <Darren.Powell at amd.com<mailto:Darren.Powell at amd.com>>
Cc: Zorro.Zhang at dell.com<mailto:Zorro.Zhang at dell.com>
Subject: RE: (EXT-15760) [I+A] DRI_PRIME=1 glxgears tearing [8086:5917][1002:6900]

Dell - Internal Use - Confidential
Hi AMDs,
  Could you please help to check this issue? There might be some connection between this issue and AMD dpm function.
Thanks

From: Emily Chien (JIRA) [mailto:NoReply-JIRA at amd.com]
Sent: Monday, July 31, 2017 11:04 AM
To: Will.Lin at amd.com<mailto:Will.Lin at amd.com>; chih.ho at canonical.com<mailto:chih.ho at canonical.com>; Napoleon.Wang at amd.com<mailto:Napoleon.Wang at amd.com>; kent.lin at canonical.com<mailto:kent.lin at canonical.com>; Frasier.Wu at amd.com<mailto:Frasier.Wu at amd.com>; Zhang1, Alex <Alex_Zhang1 at Dell.com<mailto:Alex_Zhang1 at Dell.com>>; jeffp_wang at compal.com<mailto:jeffp_wang at compal.com>; rabbit_tsao at compal.com<mailto:rabbit_tsao at compal.com>; emily.chien at canonical.com<mailto:emily.chien at canonical.com>; kevin.lu at canonical.com<mailto:kevin.lu at canonical.com>; Darren.Powell at amd.com<mailto:Darren.Powell at amd.com>
Subject: (EXT-15760) [I+A] DRI_PRIME=1 glxgears tearing [8086:5917][1002:6900]


Emily Chien<https://ontrack.amd.com/secure/ViewProfile.jspa?name=emily.chien%40canonical.com> commented on [Ticket] EXT-15760<https://ontrack.amd.com/browse/EXT-15760>




Re: [I+A] DRI_PRIME=1 glxgears tearing [8086:5917][1002:6900]<https://ontrack.amd.com/browse/EXT-15760>



We still need AMD to investigate on this issue.

Here are further findings:

1. Cannot reproduce the issue in DC mode.
2. Cyrus cannot reproduce the issue in AC and DC mode after he adds "amdpgu.dpm=0" into kernel command line.

Test Environment:
1. Image: dell-bto-xenial-loki-13-armani-13-14-X75-iso-20170713-0.iso
2. Platform: ANM4-DVT1-C2 (201707-25595) (5917), LOIT5-DVT1-C2 (201706-25560) (5917)

Test Result [8086:5917]:

1. AC Mode:
$ DRI_PRIME=1 glxgear -fullscreen: tearing
$ DRI_PRIME=1 glxgear: tearing

2. DC Mode:
$ DRI_PRIME=1 glxgear -fullscreen: tearing
$ DRI_PRIME=1 glxgear: PASS

3. AC Mode + amdgpu.dpm=0:
$ DRI_PRIME=1 glxgear -fullscreen: PASS
$ DRI_PRIME=1 glxgear: PASS

4. DC Mode + amdgpu.dpm=0:
$ DRI_PRIME=1 glxgear -fullscreen: PASS
$ DRI_PRIME=1 glxgear: PASS


[Add Comment]<https://ontrack.amd.com/browse/EXT-15760#add-comment>

Add Comment<https://ontrack.amd.com/browse/EXT-15760#add-comment>






This message was sent by Atlassian JIRA (v7.2.7#72009-sha1:68b7d86)

[Atlassian logo]







--


--

Emily Chien

Project Manager

Canonical Group Ltd.

e-mail:emily.chien at canonical.com<mailto:e-mail%3Aemily.chien at canonical.com>
Tel: +886-2-8729-6815<tel:+886%202%208729%206815>
Ubuntu | www.ubuntu.com<http://www.ubuntu.com>



--


--

Emily Chien

Project Manager

Canonical Group Ltd.

e-mail:emily.chien at canonical.com<mailto:e-mail%3Aemily.chien at canonical.com>
Tel: +886-2-8729-6815<tel:+886%202%208729%206815>
Ubuntu | www.ubuntu.com<http://www.ubuntu.com>



--


--

Emily Chien

Project Manager

Canonical Group Ltd.

e-mail:emily.chien at canonical.com<mailto:e-mail%3Aemily.chien at canonical.com>
Tel: +886-2-8729-6815<tel:+886%202%208729%206815>
Ubuntu | www.ubuntu.com<http://www.ubuntu.com>



--


--

Emily Chien

Project Manager

Canonical Group Ltd.

e-mail:emily.chien at canonical.com<mailto:e-mail%3Aemily.chien at canonical.com>
Tel: +886-2-8729-6815
Ubuntu | www.ubuntu.com<http://www.ubuntu.com>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/intel-gfx/attachments/20170830/89d5f9e4/attachment-0001.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image001.jpg
Type: image/jpeg
Size: 35837 bytes
Desc: image001.jpg
URL: <https://lists.freedesktop.org/archives/intel-gfx/attachments/20170830/89d5f9e4/attachment-0002.jpg>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image002.png
Type: image/png
Size: 473 bytes
Desc: image002.png
URL: <https://lists.freedesktop.org/archives/intel-gfx/attachments/20170830/89d5f9e4/attachment-0008.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image003.png
Type: image/png
Size: 3857 bytes
Desc: image003.png
URL: <https://lists.freedesktop.org/archives/intel-gfx/attachments/20170830/89d5f9e4/attachment-0009.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image004.jpg
Type: image/jpeg
Size: 46833 bytes
Desc: image004.jpg
URL: <https://lists.freedesktop.org/archives/intel-gfx/attachments/20170830/89d5f9e4/attachment-0003.jpg>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image005.png
Type: image/png
Size: 349 bytes
Desc: image005.png
URL: <https://lists.freedesktop.org/archives/intel-gfx/attachments/20170830/89d5f9e4/attachment-0010.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image006.png
Type: image/png
Size: 974 bytes
Desc: image006.png
URL: <https://lists.freedesktop.org/archives/intel-gfx/attachments/20170830/89d5f9e4/attachment-0011.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image007.png
Type: image/png
Size: 12535 bytes
Desc: image007.png
URL: <https://lists.freedesktop.org/archives/intel-gfx/attachments/20170830/89d5f9e4/attachment-0012.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image008.png
Type: image/png
Size: 666 bytes
Desc: image008.png
URL: <https://lists.freedesktop.org/archives/intel-gfx/attachments/20170830/89d5f9e4/attachment-0013.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image009.png
Type: image/png
Size: 1084 bytes
Desc: image009.png
URL: <https://lists.freedesktop.org/archives/intel-gfx/attachments/20170830/89d5f9e4/attachment-0014.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image010.png
Type: image/png
Size: 2983 bytes
Desc: image010.png
URL: <https://lists.freedesktop.org/archives/intel-gfx/attachments/20170830/89d5f9e4/attachment-0015.png>

More information about the Intel-gfx mailing list