Any hope of vblank-synchronized rendering (from vivante) on i.MX6 soon?

Any hope of vblank-synchronized rendering (from vivante) on i.MX6 soon?

[Gerard Bucas]

On X.11 and Chromium there is still "tearing" of videos played back in
Chromium (on X11). From what I have read and been told by others this is
caused by the lack of DRI2 support in the Vivante graphics libraries (even
in the latest V5.x libraries included in the latest
fsl-yocto-3.10.31-1.1.0_beta BSP).

Is there any hope of Vivante implementing DRI2 soon or some other way to
implement vblank-synchronized rendering ("vsync") on the i.MX6 so that
Chrome/Chromium can playback videos without this serious "tearing" problem?

How do we convince Vivante to give this a high priority!?

Surely this must be a major problem for ALL i.MX6 users?

Curious....

Gerard Bucas

Re: Any hope of vblank-synchronized rendering (from vivante) on i.MX6 soon?

[Daiane Angolini]

On Tue, Sep 23, 2014 at 1:31 AM, Gerard Bucas <gerard.b@tekmagic.net> wrote:
> On X.11 and Chromium there is still "tearing" of videos played back in
> Chromium (on X11). From what I have read and been told by others this is
> caused by the lack of DRI2 support in the Vivante graphics libraries (even
> in the latest V5.x libraries included in the latest
> fsl-yocto-3.10.31-1.1.0_beta BSP).
>
> Is there any hope of Vivante implementing DRI2 soon or some other way to
> implement vblank-synchronized rendering ("vsync") on the i.MX6 so that
> Chrome/Chromium can playback videos without this serious "tearing" problem?
>
> How do we convince Vivante to give this a high priority!?

This is a interesting topic, and i think there is room for a healthy
discussion in our next community meeting. Please, keep a list of
issues and suggest it when the next agenda is published.


for now, let´s wait if someone else has some word on this thread.

Daiane

Re: Any hope of vblank-synchronized rendering (from vivante) on i.MX6 soon?

[Gerard Bucas]

Thanks Daiane - will do. 

When is the next such meeting and who is actually the interface with Vivante that could put some pressure on them!?

DRI2 is used all over in the X11 code according to this so I think this is a critical area to solve the serious shortcomings and performance in the current X11 port on the I.MX6. See: http://en.m.wikipedia.org/wiki/DRI2 for more details. 

Best Regards

Gerard Bucas
TekMagic



> On 23 Sep 2014, at 07:45, Daiane Angolini <daiane.list@gmail.com> wrote:
> 
>> On Tue, Sep 23, 2014 at 1:31 AM, Gerard Bucas <gerard.b@tekmagic.net> wrote:
>> On X.11 and Chromium there is still "tearing" of videos played back in
>> Chromium (on X11). From what I have read and been told by others this is
>> caused by the lack of DRI2 support in the Vivante graphics libraries (even
>> in the latest V5.x libraries included in the latest
>> fsl-yocto-3.10.31-1.1.0_beta BSP).
>> 
>> Is there any hope of Vivante implementing DRI2 soon or some other way to
>> implement vblank-synchronized rendering ("vsync") on the i.MX6 so that
>> Chrome/Chromium can playback videos without this serious "tearing" problem?
>> 
>> How do we convince Vivante to give this a high priority!?
> 
> This is a interesting topic, and i think there is room for a healthy
> discussion in our next community meeting. Please, keep a list of
> issues and suggest it when the next agenda is published.
> 
> 
> for now, let´s wait if someone else has some word on this thread.
> 
> Daiane

Re: Any hope of vblank-synchronized rendering (from vivante) on i.MX6 soon?

[Daiane Angolini]

On Tue, Sep 23, 2014 at 10:04 AM, Gerard Bucas <gerard.b@tekmagic.net> wrote:
> Thanks Daiane - will do.
>
> When is the next such meeting and who is actually the interface with Vivante that could put some pressure on them!?

Next meeting is going to be something near 10th, next moth, more or less....

Re: Any hope of vblank-synchronized rendering (from vivante) on i.MX6 soon?

[Carlos Rafael Giani]

Perhaps there is still time left to notify Freescale how critical this 
feature is for several people. Imagine tearing on huge displays for 
example ... this can easily become a showstopper. Freescale should 
consider this one of the most critical features to introduce in 3.10.31 GA.

On 09/23/2014 03:04 PM, Gerard Bucas wrote:
> Thanks Daiane - will do.
>
> When is the next such meeting and who is actually the interface with Vivante that could put some pressure on them!?
>
> DRI2 is used all over in the X11 code according to this so I think this is a critical area to solve the serious shortcomings and performance in the current X11 port on the I.MX6. See: http://en.m.wikipedia.org/wiki/DRI2 for more details.
>
> Best Regards
>
> Gerard Bucas
> TekMagic
>
>
>
>> On 23 Sep 2014, at 07:45, Daiane Angolini <daiane.list@gmail.com> wrote:
>>
>>> On Tue, Sep 23, 2014 at 1:31 AM, Gerard Bucas <gerard.b@tekmagic.net> wrote:
>>> On X.11 and Chromium there is still "tearing" of videos played back in
>>> Chromium (on X11). From what I have read and been told by others this is
>>> caused by the lack of DRI2 support in the Vivante graphics libraries (even
>>> in the latest V5.x libraries included in the latest
>>> fsl-yocto-3.10.31-1.1.0_beta BSP).
>>>
>>> Is there any hope of Vivante implementing DRI2 soon or some other way to
>>> implement vblank-synchronized rendering ("vsync") on the i.MX6 so that
>>> Chrome/Chromium can playback videos without this serious "tearing" problem?
>>>
>>> How do we convince Vivante to give this a high priority!?
>> This is a interesting topic, and i think there is room for a healthy
>> discussion in our next community meeting. Please, keep a list of
>> issues and suggest it when the next agenda is published.
>>
>>
>> for now, let´s wait if someone else has some word on this thread.
>>
>> Daiane

Re: Any hope of vblank-synchronized rendering (from vivante) on i.MX6 soon?

[Eric Nelson]

On 09/23/2014 07:00 AM, Carlos Rafael Giani wrote:
> Perhaps there is still time left to notify Freescale how critical this
> feature is for several people. Imagine tearing on huge displays for
> example ... this can easily become a showstopper. Freescale should
> consider this one of the most critical features to introduce in 3.10.31 GA.
> 

+1

This also affects animation under X, whether driven by Qt, HTML5,
or Java.

It's not a kernel thing though, since Wayland works very well.

Does anyone know the schedule for Xserver 1.6 in Yocto and what
the plans are for Vivante acceleration support for 1.6?

Regards,


Eric

Re: Any hope of vblank-synchronized rendering (from vivante) on i.MX6 soon?

[Gerard Bucas]

I agree Carlos!

This is a show-stopper for applications in markets like the Digital Signage market - which could be huge for the i.MX6. The i.MX6 could be the best "media player" out there but the problems related to video playback in a browser (like Chrome/Chromium) needs to be solved wih the highest priority and at the highest level.

There seem to be TWO main issues:

1. Lack of vsync synchronized support for video playback in vivante graphics libraries. This is causing serious tearing on video playback in Chromium (which is dependent on this support in X11 or ozone-wayland). In X11 this requires support for DRI2 in the vivante GPU/graphics libraries. See http://en.wikipedia.org/wiki/Direct_Rendering_Infrastructure for more details.

2. More hardware acceleration support in the GPU/VPU libraries (also mostly in the domain of vivante I guess). We need a clear summary (LIST) of:
- What exactly is currently being "hardware accelerated" (latest V5.x graphics libraries were a great step forward!)
- What additional support is required in the GPU/VPU libraries to further improve performance in X11 and ozone-wayland.

In my opinion, we need to appoint a "point person" at freescale and vivante to clearly focus on raising the priority of these issues, and keep focus on them with a clearly defined roadmap (timeline) and in parallel bring this to the attention of the most senior i.MX6 product manager (and marketing people) at freescale so that they realize the significant market opportunities they are missing (eg: Digital Signage) by ignoring these issues. The i.MX6 could be the best HRML5 "playback device" (which is dependent on a browser like Chrome/Chromium and hence X11 and/or ozone-wayland) for a number of industries. Having been the CEO of Scala (www.scala.com ) for 10 years, I know the Digital Signage industry very well and it is hungrily waiting for a good ARM-based "HTML5 playback device" to replace the current options available (which require more power, generate more heat, need fans which have high maintenance problems in industrial/commercial environments, are more expensive). I am REALLY surprised that freescale (and vivante!) is not putting more focus on  this!

Let’s get everyone on board with this and start applying some serious pressure on ALL parties involved!!!

Regards

Gerard Bucas   


-----Original Message-----
From: meta-freescale-bounces@yoctoproject.org [mailto:meta-freescale-bounces@yoctoproject.org] On Behalf Of Carlos Rafael Giani
Sent: Tuesday, September 23, 2014 10:00 AM
To: meta-freescale@yoctoproject.org
Subject: Re: [meta-freescale] Any hope of vblank-synchronized rendering (from vivante) on i.MX6 soon?

Perhaps there is still time left to notify Freescale how critical this feature is for several people. Imagine tearing on huge displays for example ... this can easily become a showstopper. Freescale should consider this one of the most critical features to introduce in 3.10.31 GA.

On 09/23/2014 03:04 PM, Gerard Bucas wrote:
> Thanks Daiane - will do.
>
> When is the next such meeting and who is actually the interface with Vivante that could put some pressure on them!?
>
> DRI2 is used all over in the X11 code according to this so I think this is a critical area to solve the serious shortcomings and performance in the current X11 port on the I.MX6. See: http://en.m.wikipedia.org/wiki/DRI2 for more details.
>
> Best Regards
>
> Gerard Bucas
> TekMagic
>
>
>
>> On 23 Sep 2014, at 07:45, Daiane Angolini <daiane.list@gmail.com> wrote:
>>
>>> On Tue, Sep 23, 2014 at 1:31 AM, Gerard Bucas <gerard.b@tekmagic.net> wrote:
>>> On X.11 and Chromium there is still "tearing" of videos played back 
>>> in Chromium (on X11). From what I have read and been told by others 
>>> this is caused by the lack of DRI2 support in the Vivante graphics 
>>> libraries (even in the latest V5.x libraries included in the latest 
>>> fsl-yocto-3.10.31-1.1.0_beta BSP).
>>>
>>> Is there any hope of Vivante implementing DRI2 soon or some other 
>>> way to implement vblank-synchronized rendering ("vsync") on the 
>>> i.MX6 so that Chrome/Chromium can playback videos without this serious "tearing" problem?
>>>
>>> How do we convince Vivante to give this a high priority!?
>> This is a interesting topic, and i think there is room for a healthy 
>> discussion in our next community meeting. Please, keep a list of 
>> issues and suggest it when the next agenda is published.
>>
>>
>> for now, let´s wait if someone else has some word on this thread.
>>
>> Daiane

--
_______________________________________________
meta-freescale mailing list
meta-freescale@yoctoproject.org
https://lists.yoctoproject.org/listinfo/meta-freescale

Re: Any hope of vblank-synchronized rendering (from vivante) on i.MX6 soon?

[Daiane Angolini]

On Tue, Sep 23, 2014 at 11:31 AM, Gerard Bucas <gerard.b@tekmagic.net> wrote:
> I agree Carlos!
>
> This is a show-stopper for applications in markets like the Digital Signage market - which could be huge for the i.MX6. The i.MX6 could be the best "media player" out there but the problems related to video playback in a browser (like Chrome/Chromium) needs to be solved wih the highest priority and at the highest level.
>
> There seem to be TWO main issues:

Do you have only 2 issues? Have you double checked if they are/are not
reported in bugzilla?

Do you have the detailed step-by-step? I want to reproduce them.

Daiane

Re: Any hope of vblank-synchronized rendering (from vivante) on i.MX6 soon?

[Carlos Rafael Giani]

[-- Attachment #1: Type: text/plain, Size: 927 bytes --]

On 09/23/2014 04:50 PM, Daiane Angolini wrote:
> On Tue, Sep 23, 2014 at 11:31 AM, Gerard Bucas <gerard.b@tekmagic.net> wrote:
>> I agree Carlos!
>>
>> This is a show-stopper for applications in markets like the Digital Signage market - which could be huge for the i.MX6. The i.MX6 could be the best "media player" out there but the problems related to video playback in a browser (like Chrome/Chromium) needs to be solved wih the highest priority and at the highest level.
>>
>> There seem to be TWO main issues:
> Do you have only 2 issues? Have you double checked if they are/are not
> reported in bugzilla?
>
> Do you have the detailed step-by-step? I want to reproduce them.
>
> Daiane

Daiane,

try out the attached HTML example. It draws a white vertical bar on a 
black background, and moves that bar horizontally. Run this on an imx6 
in Chromium in X11, and you will see how severe the tearing is.

[-- Attachment #2: vertical-bar.html --]
[-- Type: text/html, Size: 1152 bytes --]

Re: Any hope of vblank-synchronized rendering (from vivante) on i.MX6 soon? [Resending to avoid HTML problems]

[Carlos Rafael Giani]

[-- Attachment #1.1: Type: text/plain, Size: 1168 bytes --]

On 09/23/2014 04:55 PM, Carlos Rafael Giani wrote:
> On 09/23/2014 04:50 PM, Daiane Angolini wrote:
>> On Tue, Sep 23, 2014 at 11:31 AM, Gerard Bucas 
>> <gerard.b@tekmagic.net> wrote:
>>> I agree Carlos!
>>>
>>> This is a show-stopper for applications in markets like the Digital 
>>> Signage market - which could be huge for the i.MX6. The i.MX6 could 
>>> be the best "media player" out there but the problems related to 
>>> video playback in a browser (like Chrome/Chromium) needs to be 
>>> solved wih the highest priority and at the highest level.
>>>
>>> There seem to be TWO main issues:
>> Do you have only 2 issues? Have you double checked if they are/are not
>> reported in bugzilla?
>>
>> Do you have the detailed step-by-step? I want to reproduce them.
>>
>> Daiane
>
> Daiane,
>
> try out the attached HTML example. It draws a white vertical bar on a 
> black background, and moves that bar horizontally. Run this on an imx6 
> in Chromium in X11, and you will see how severe the tearing is.
>
>

(Resending this mail with a GZ-compressed html file as attachment to 
avoid mail clients using a black background in the mail

[-- Attachment #1.2: Type: text/html, Size: 2086 bytes --]

[-- Attachment #2: vertical-bar.html.gz --]
[-- Type: application/gzip, Size: 596 bytes --]

Re: Any hope of vblank-synchronized rendering (from vivante) on i.MX6 soon?

[Prabhu S]

[-- Attachment #1: Type: text/plain, Size: 1680 bytes --]

i.MX6 X11 EGL is single buffered and there is no support for VSYNC. The
tearing with X11 is a known issue. We did investigate to fix this, but the
solution becomes too hacky and cannot be used in production systems.

We are adding VSYNC support in Wayland. eglSwapInterval will be supported
for FBDEV and Wayland backends in the upcoming release(in january). So
Ozone-wayland can take advantage of this.

On Tue, Sep 23, 2014 at 9:55 AM, Carlos Rafael Giani <dv@pseudoterminal.org>
wrote:

> On 09/23/2014 04:50 PM, Daiane Angolini wrote:
>
>> On Tue, Sep 23, 2014 at 11:31 AM, Gerard Bucas <gerard.b@tekmagic.net>
>> wrote:
>>
>>> I agree Carlos!
>>>
>>> This is a show-stopper for applications in markets like the Digital
>>> Signage market - which could be huge for the i.MX6. The i.MX6 could be the
>>> best "media player" out there but the problems related to video playback in
>>> a browser (like Chrome/Chromium) needs to be solved wih the highest
>>> priority and at the highest level.
>>>
>>> There seem to be TWO main issues:
>>>
>> Do you have only 2 issues? Have you double checked if they are/are not
>> reported in bugzilla?
>>
>> Do you have the detailed step-by-step? I want to reproduce them.
>>
>> Daiane
>>
>
> Daiane,
>
> try out the attached HTML example. It draws a white vertical bar on a
> black background, and moves that bar horizontally. Run this on an imx6 in
> Chromium in X11, and you will see how severe the tearing is.
>
> --
> _______________________________________________
> meta-freescale mailing list
> meta-freescale@yoctoproject.org
> https://lists.yoctoproject.org/listinfo/meta-freescale
>
>

[-- Attachment #2: Type: text/html, Size: 2434 bytes --]

Re: Any hope of vblank-synchronized rendering (from vivante) on i.MX6 soon?

[Carlos Rafael Giani]

[-- Attachment #1: Type: text/plain, Size: 3080 bytes --]

This is probably because v4l somehow uses vsync directly. Keep in mind 
that v4l bypasses X11.
Also, you can use qt5 with pure framebuffer output. There, vsync works.

On 2014-09-23 19:19, Alfonso Tamés wrote:
>
>
> Yes, the attached test video is tearing in EGL FB without X11/wayland.
>
> This tears:
>
> gst-launch-1.0 -v filesrc location=vttest.mp4 typefind=true ! qtdemux 
> ! h264parse ! imxvpudec ! imxeglvivsink
>
> This does not tear:
>
> gst-launch -v filesrc location=vttest.mp4 typefind=true ! aiurdemux ! 
> queue ! vpudec ! mfw_v4lsink
>
> I was jut starting a project using qt-gst gl video sink. I guess it 
> will be a dead end
>
> :(
>
>
>
>
>
> On Sep 23, 2014, at 11:48 AM, Prabhu S <prabhusundar@gmail.com 
> <mailto:prabhusundar@gmail.com>> wrote:
>
>> i.MX6 X11 EGL is single buffered and there is no support for VSYNC. 
>> The tearing with X11 is a known issue. We did investigate to fix 
>> this, but the solution becomes too hacky and cannot be used in 
>> production systems.
>>
>> We are adding VSYNC support in Wayland. eglSwapInterval will be 
>> supported for FBDEV and Wayland backends in the upcoming release(in 
>> january). So Ozone-wayland can take advantage of this.
>>
>> On Tue, Sep 23, 2014 at 9:55 AM, Carlos Rafael Giani 
>> <dv@pseudoterminal.org <mailto:dv@pseudoterminal.org>> wrote:
>>
>>     On 09/23/2014 04:50 PM, Daiane Angolini wrote:
>>
>>         On Tue, Sep 23, 2014 at 11:31 AM, Gerard Bucas
>>         <gerard.b@tekmagic.net <mailto:gerard.b@tekmagic.net>> wrote:
>>
>>             I agree Carlos!
>>
>>             This is a show-stopper for applications in markets like
>>             the Digital Signage market - which could be huge for the
>>             i.MX6. The i.MX6 could be the best "media player" out
>>             there but the problems related to video playback in a
>>             browser (like Chrome/Chromium) needs to be solved wih the
>>             highest priority and at the highest level.
>>
>>             There seem to be TWO main issues:
>>
>>         Do you have only 2 issues? Have you double checked if they
>>         are/are not
>>         reported in bugzilla?
>>
>>         Do you have the detailed step-by-step? I want to reproduce them.
>>
>>         Daiane
>>
>>
>>     Daiane,
>>
>>     try out the attached HTML example. It draws a white vertical bar
>>     on a black background, and moves that bar horizontally. Run this
>>     on an imx6 in Chromium in X11, and you will see how severe the
>>     tearing is.
>>
>>     --
>>     _______________________________________________
>>     meta-freescale mailing list
>>     meta-freescale@yoctoproject.org
>>     <mailto:meta-freescale@yoctoproject.org>
>>     https://lists.yoctoproject.org/listinfo/meta-freescale
>>
>>
>> -- 
>> _______________________________________________
>> meta-freescale mailing list
>> meta-freescale@yoctoproject.org <mailto:meta-freescale@yoctoproject.org>
>> https://lists.yoctoproject.org/listinfo/meta-freescale
>


[-- Attachment #2: Type: text/html, Size: 6358 bytes --]

Re: Any hope of vblank-synchronized rendering (from vivante) on i.MX6 soon?

[Carlos Rafael Giani]

[-- Attachment #1: Type: text/plain, Size: 2249 bytes --]

I wonder why it is single buffered. is it because of the memory 
consumption for a second buffer? The buffers could be flipped, 
eliminating any need for blits, so the memory bandwidth should not be an 
issue.

On 2014-09-23 18:48, Prabhu S wrote:
> i.MX6 X11 EGL is single buffered and there is no support for VSYNC. 
> The tearing with X11 is a known issue. We did investigate to fix this, 
> but the solution becomes too hacky and cannot be used in production 
> systems.
>
> We are adding VSYNC support in Wayland. eglSwapInterval will be 
> supported for FBDEV and Wayland backends in the upcoming release(in 
> january). So Ozone-wayland can take advantage of this.
>
> On Tue, Sep 23, 2014 at 9:55 AM, Carlos Rafael Giani 
> <dv@pseudoterminal.org <mailto:dv@pseudoterminal.org>> wrote:
>
>     On 09/23/2014 04:50 PM, Daiane Angolini wrote:
>
>         On Tue, Sep 23, 2014 at 11:31 AM, Gerard Bucas
>         <gerard.b@tekmagic.net <mailto:gerard.b@tekmagic.net>> wrote:
>
>             I agree Carlos!
>
>             This is a show-stopper for applications in markets like
>             the Digital Signage market - which could be huge for the
>             i.MX6. The i.MX6 could be the best "media player" out
>             there but the problems related to video playback in a
>             browser (like Chrome/Chromium) needs to be solved wih the
>             highest priority and at the highest level.
>
>             There seem to be TWO main issues:
>
>         Do you have only 2 issues? Have you double checked if they
>         are/are not
>         reported in bugzilla?
>
>         Do you have the detailed step-by-step? I want to reproduce them.
>
>         Daiane
>
>
>     Daiane,
>
>     try out the attached HTML example. It draws a white vertical bar
>     on a black background, and moves that bar horizontally. Run this
>     on an imx6 in Chromium in X11, and you will see how severe the
>     tearing is.
>
>     --
>     _______________________________________________
>     meta-freescale mailing list
>     meta-freescale@yoctoproject.org
>     <mailto:meta-freescale@yoctoproject.org>
>     https://lists.yoctoproject.org/listinfo/meta-freescale
>
>


[-- Attachment #2: Type: text/html, Size: 4122 bytes --]

Re: Any hope of vblank-synchronized rendering (from vivante) on i.MX6 soon?

[Eric Nelson]

Hi Carlos,

On 09/23/2014 10:52 AM, Carlos Rafael Giani wrote:
>
> I wonder why it is single buffered. is it because of the memory
> consumption for a second buffer? The buffers could be flipped,
> eliminating any need for blits, so the memory bandwidth should not be an
> issue.
>

There is a memory bandwidth implication.

Assuming frame buffers A and B in a double-buffered system,
rendering will occur into B while A is displayed, then the
two will be swapped, such that B is displayed and A is the
new rendering output.

In order for apps to only update parts of their windows, any
"dirty" blocks need to be copied from B to A in the process
of the flip.

> On 2014-09-23 18:48, Prabhu S wrote:
>> i.MX6 X11 EGL is single buffered and there is no support for VSYNC.
>> The tearing with X11 is a known issue. We did investigate to fix this,
>> but the solution becomes too hacky and cannot be used in production
>> systems.
>>
>> We are adding VSYNC support in Wayland. eglSwapInterval will be
>> supported for FBDEV and Wayland backends in the upcoming release(in
>> january). So Ozone-wayland can take advantage of this.
>>

?? This is already supported in Wayland, though it is controlled
by an environment variable.

You can see it in action by watching /sys/class/graphics/fb0/pan.

Regards,


Eric

Re: Any hope of vblank-synchronized rendering (from vivante) on i.MX6 soon?

[Prabhu S]

[-- Attachment #1: Type: text/plain, Size: 1636 bytes --]

On Tue, Sep 23, 2014 at 1:46 PM, Eric Nelson <
eric.nelson@boundarydevices.com> wrote:

> Hi Carlos,
>
> On 09/23/2014 10:52 AM, Carlos Rafael Giani wrote:
> >
> > I wonder why it is single buffered. is it because of the memory
> > consumption for a second buffer? The buffers could be flipped,
> > eliminating any need for blits, so the memory bandwidth should not be an
> > issue.
> >
>
> There is a memory bandwidth implication.
>
> Assuming frame buffers A and B in a double-buffered system,
> rendering will occur into B while A is displayed, then the
> two will be swapped, such that B is displayed and A is the
> new rendering output.
>
> In order for apps to only update parts of their windows, any
> "dirty" blocks need to be copied from B to A in the process
> of the flip.
>
> > On 2014-09-23 18:48, Prabhu S wrote:
> >> i.MX6 X11 EGL is single buffered and there is no support for VSYNC.
> >> The tearing with X11 is a known issue. We did investigate to fix this,
> >> but the solution becomes too hacky and cannot be used in production
> >> systems.
> >>
> >> We are adding VSYNC support in Wayland. eglSwapInterval will be
> >> supported for FBDEV and Wayland backends in the upcoming release(in
> >> january). So Ozone-wayland can take advantage of this.
> >>
>
> ?? This is already supported in Wayland, though it is controlled
> by an environment variable.
>
export FB_MULTI_BUFFER=2. By doing this server side is VYSNC controlled.
The client side support in next release.

>
> You can see it in action by watching /sys/class/graphics/fb0/pan.
>
> Regards,
>
>
> Eric
>

[-- Attachment #2: Type: text/html, Size: 2323 bytes --]

Re: Any hope of vblank-synchronized rendering (from vivante) on i.MX6 soon?

[Eric Nelson]

Hi Prabhu,

On 09/23/2014 11:50 AM, Prabhu S wrote:
> 
> 
> On Tue, Sep 23, 2014 at 1:46 PM, Eric Nelson
> <eric.nelson@boundarydevices.com
> <mailto:eric.nelson@boundarydevices.com>> wrote:
> 
> <snip>
>
>     >> We are adding VSYNC support in Wayland. eglSwapInterval will be
>     >> supported for FBDEV and Wayland backends in the upcoming release(in
>     >> january). So Ozone-wayland can take advantage of this.
>     >>
> 
>     ?? This is already supported in Wayland, though it is controlled
>     by an environment variable.
> 
> export FB_MULTI_BUFFER=2. By doing this server side is VYSNC controlled.
> The client side support in next release.
> 

Thanks for the clarification.

Re: Any hope of vblank-synchronized rendering (from vivante) on i.MX6 soon?

[Carlos Rafael Giani]

While this is correct for output that is drawn directly to the 
framebuffer (with damage regions etc.), it isn't for compositor-based 
output, such as what Weston does. There, it is not uncommon to repaint  
the entire screen. If it uses GLES as the backend, this is the only way 
to do it, unless some extension for swapping buffer subregions exists. 
Perhaps by using a compositor in X buffer flipping would be viable.

On 2014-09-23 20:46, Eric Nelson wrote:
> Hi Carlos,
>
> On 09/23/2014 10:52 AM, Carlos Rafael Giani wrote:
>> I wonder why it is single buffered. is it because of the memory
>> consumption for a second buffer? The buffers could be flipped,
>> eliminating any need for blits, so the memory bandwidth should not be an
>> issue.
>>
> There is a memory bandwidth implication.
>
> Assuming frame buffers A and B in a double-buffered system,
> rendering will occur into B while A is displayed, then the
> two will be swapped, such that B is displayed and A is the
> new rendering output.
>
> In order for apps to only update parts of their windows, any
> "dirty" blocks need to be copied from B to A in the process
> of the flip.
>
>> On 2014-09-23 18:48, Prabhu S wrote:
>>> i.MX6 X11 EGL is single buffered and there is no support for VSYNC.
>>> The tearing with X11 is a known issue. We did investigate to fix this,
>>> but the solution becomes too hacky and cannot be used in production
>>> systems.
>>>
>>> We are adding VSYNC support in Wayland. eglSwapInterval will be
>>> supported for FBDEV and Wayland backends in the upcoming release(in
>>> january). So Ozone-wayland can take advantage of this.
>>>
> ?? This is already supported in Wayland, though it is controlled
> by an environment variable.
>
> You can see it in action by watching /sys/class/graphics/fb0/pan.
>
> Regards,
>
>
> Eric

Re: Any hope of vblank-synchronized rendering (from vivante) on i.MX6 soon?

[Daiane Angolini]

On Tue, Sep 23, 2014 at 5:09 PM, Carlos Rafael Giani
<dv@pseudoterminal.org> wrote:
> While this is correct for output that is drawn directly to the framebuffer
> (with damage regions etc.), it isn't for compositor-based output, such as
> what Weston does. There, it is not uncommon to repaint  the entire screen.
> If it uses GLES as the backend, this is the only way to do it, unless some
> extension for swapping buffer subregions exists. Perhaps by using a
> compositor in X buffer flipping would be viable.

Carlos, would you mind to open a bug on bugzilla for this VSYNC issue?

It looks like you can describe it better than me. Please add the html
file you shared and the DISTRO_FEATURES needed to reproduce the issue.

Also, please, say which kernel version you used and which GPU version.

I can then, update this with the comments from Prabhu, and change its
status accordingly.

Thanks in advance,
Daiane

>
>
> On 2014-09-23 20:46, Eric Nelson wrote:
>>
>> Hi Carlos,
>>
>> On 09/23/2014 10:52 AM, Carlos Rafael Giani wrote:
>>>
>>> I wonder why it is single buffered. is it because of the memory
>>> consumption for a second buffer? The buffers could be flipped,
>>> eliminating any need for blits, so the memory bandwidth should not be an
>>> issue.
>>>
>> There is a memory bandwidth implication.
>>
>> Assuming frame buffers A and B in a double-buffered system,
>> rendering will occur into B while A is displayed, then the
>> two will be swapped, such that B is displayed and A is the
>> new rendering output.
>>
>> In order for apps to only update parts of their windows, any
>> "dirty" blocks need to be copied from B to A in the process
>> of the flip.
>>
>>> On 2014-09-23 18:48, Prabhu S wrote:
>>>>
>>>> i.MX6 X11 EGL is single buffered and there is no support for VSYNC.
>>>> The tearing with X11 is a known issue. We did investigate to fix this,
>>>> but the solution becomes too hacky and cannot be used in production
>>>> systems.
>>>>
>>>> We are adding VSYNC support in Wayland. eglSwapInterval will be
>>>> supported for FBDEV and Wayland backends in the upcoming release(in
>>>> january). So Ozone-wayland can take advantage of this.
>>>>
>> ?? This is already supported in Wayland, though it is controlled
>> by an environment variable.
>>
>> You can see it in action by watching /sys/class/graphics/fb0/pan.
>>
>> Regards,
>>
>>
>> Eric
>
>
> --
> _______________________________________________
> meta-freescale mailing list
> meta-freescale@yoctoproject.org
> https://lists.yoctoproject.org/listinfo/meta-freescale

Re: Any hope of vblank-synchronized rendering (from vivante) on i.MX6 soon?

[Carlos Rafael Giani]

Done. See https://bugzilla.yoctoproject.org/show_bug.cgi?id=6760

On 2014-09-24 17:36, Daiane Angolini wrote:
> On Tue, Sep 23, 2014 at 5:09 PM, Carlos Rafael Giani
> <dv@pseudoterminal.org> wrote:
>> While this is correct for output that is drawn directly to the framebuffer
>> (with damage regions etc.), it isn't for compositor-based output, such as
>> what Weston does. There, it is not uncommon to repaint  the entire screen.
>> If it uses GLES as the backend, this is the only way to do it, unless some
>> extension for swapping buffer subregions exists. Perhaps by using a
>> compositor in X buffer flipping would be viable.
> Carlos, would you mind to open a bug on bugzilla for this VSYNC issue?
>
> It looks like you can describe it better than me. Please add the html
> file you shared and the DISTRO_FEATURES needed to reproduce the issue.
>
> Also, please, say which kernel version you used and which GPU version.
>
> I can then, update this with the comments from Prabhu, and change its
> status accordingly.
>
> Thanks in advance,
> Daiane
>
>>
>> On 2014-09-23 20:46, Eric Nelson wrote:
>>> Hi Carlos,
>>>
>>> On 09/23/2014 10:52 AM, Carlos Rafael Giani wrote:
>>>> I wonder why it is single buffered. is it because of the memory
>>>> consumption for a second buffer? The buffers could be flipped,
>>>> eliminating any need for blits, so the memory bandwidth should not be an
>>>> issue.
>>>>
>>> There is a memory bandwidth implication.
>>>
>>> Assuming frame buffers A and B in a double-buffered system,
>>> rendering will occur into B while A is displayed, then the
>>> two will be swapped, such that B is displayed and A is the
>>> new rendering output.
>>>
>>> In order for apps to only update parts of their windows, any
>>> "dirty" blocks need to be copied from B to A in the process
>>> of the flip.
>>>
>>>> On 2014-09-23 18:48, Prabhu S wrote:
>>>>> i.MX6 X11 EGL is single buffered and there is no support for VSYNC.
>>>>> The tearing with X11 is a known issue. We did investigate to fix this,
>>>>> but the solution becomes too hacky and cannot be used in production
>>>>> systems.
>>>>>
>>>>> We are adding VSYNC support in Wayland. eglSwapInterval will be
>>>>> supported for FBDEV and Wayland backends in the upcoming release(in
>>>>> january). So Ozone-wayland can take advantage of this.
>>>>>
>>> ?? This is already supported in Wayland, though it is controlled
>>> by an environment variable.
>>>
>>> You can see it in action by watching /sys/class/graphics/fb0/pan.
>>>
>>> Regards,
>>>
>>>
>>> Eric
>>
>> --
>> _______________________________________________
>> meta-freescale mailing list
>> meta-freescale@yoctoproject.org
>> https://lists.yoctoproject.org/listinfo/meta-freescale

Any hope of vblank-synchronized rendering (from vivante) on i.MX6 soon?

[Mike Folcik]

I have been up against this problem for a few weeks now, and it is
definitely a show stopper for me using the i.MX6 for my application.  I,
too, am using Chromium atop X11 and need smooth CSS3 accelerated animations.
I have seen the Bugzilla bug; and am wondering, is this definitely in the
pipeline for the 1.8 release?  Is there any chance to increase the severity
to get it sooner?

From Gerard's original post, this IS a critical defect.  My application is
high volume, and I'm sure others interested in this defect also have high
volume, which should motivate Freescale & Vivante to act on this.

Mike