Mailing List Archive: MythTV: Dev

Support for upcoming VDPAU on AMD and Myth Player

Index | Next | Previous | View Threaded

dargllun at googlemail

Jun 6, 2011, 11:02 AM

Post #1 of 14 (2166 views)
Permalink

Support for upcoming VDPAU on AMD and Myth Player

Hello,

I asked the following on mythtv-users the other day, but perhaps the
question really is one for the devs.

The pipe-video branch [1] in Mesa brings VDPAU Support to r600/Gallium3D
hardware. I was actually able to try it with some help from the main dev
for this, and could successfully run "mplayer -v vdpau <MPEG>". No
further modifications to the mplayer where needed.

Now, how would I go about trying this for MythTV's internal player? I
did select VDPAU but it's not working that way. Apparently some nvidia
specific library is attempted to be loaded, which obviously doesn't
work. Is it possible to have the internal player omit any nvidia
proprietary extensions/additions and just use plain VDPAU like mplayer does?

Thanks for any hint or insight.

Greg

[1] http://cgit.freedesktop.org/mesa/mesa/log/?h=pipe-video
_______________________________________________
mythtv-dev mailing list
mythtv-dev [at] mythtv
http://www.mythtv.org/mailman/listinfo/mythtv-dev

mark.kendall at gmail

Jun 7, 2011, 10:00 PM

Post #2 of 14 (2085 views)
Permalink

Re: Support for upcoming VDPAU on AMD and Myth Player [In reply to]

On 7 June 2011 02:02, dargllun <dargllun [at] googlemail> wrote:
> Hello,
>
> I asked the following on mythtv-users the other day, but perhaps the
> question really is one for the devs.
>
> The pipe-video branch [1] in Mesa brings VDPAU Support to r600/Gallium3D
> hardware. I was actually able to try it with some help from the main dev
> for this, and could successfully run "mplayer -v vdpau <MPEG>". No
> further modifications to the mplayer where needed.
>
> Now, how would I go about trying this for MythTV's internal player? I
> did select VDPAU but it's not working that way. Apparently some nvidia
> specific library is attempted to be loaded, which obviously doesn't
> work. Is it possible to have the internal player omit any nvidia
> proprietary extensions/additions and just use plain VDPAU like mplayer does?

Greg

Can you post the log output with the errors reported. Did you 'make
distclean' before re-running configure?

regards

Mark
_______________________________________________
mythtv-dev mailing list
mythtv-dev [at] mythtv
http://www.mythtv.org/mailman/listinfo/mythtv-dev

dargllun at googlemail

Jun 8, 2011, 12:43 AM

Post #3 of 14 (2079 views)
Permalink

Re: Support for upcoming VDPAU on AMD and Myth Player [In reply to]

2011/6/8 Mark Kendall <mark.kendall [at] gmail>

> On 7 June 2011 02:02, dargllun <dargllun [at] googlemail> wrote:
> > Now, how would I go about trying this for MythTV's internal player? I
> > did select VDPAU but it's not working that way. Apparently some nvidia
> > specific library is attempted to be loaded, which obviously doesn't
> > work. Is it possible to have the internal player omit any nvidia
> > proprietary extensions/additions and just use plain VDPAU like mplayer
> does?
>
> Greg
>
> Can you post the log output with the errors reported. Did you 'make
> distclean' before re-running configure?
>

Er, pretty embarrassing to write on a -dev list, but actually I didn't
recompile MythTV for this test. I simply turned on VDPAU in the settings and
tried to get some results. I would not think that recompiling MythTV changed
the dependency on some nvidia specific library, except that I probably
wouldn't even be able to even compile it without errors.

What I suspect is that the player, while using VDPAU, also tinkers with some
nvidia specific things. I meant to analyze the sources for more insight but
haven't got round to it yet. If my suspicion is correct, then the generic
VDPAU would be needed to be factored out somehow into s.th. which is
generically usable.

Has anyone perhaps ever tried the VDPAU to XvBA bridge from Splitted Desktop
with the Catalyst driver? That should show pretty much the same issues, and
maybe they're solved already?

Regards
Greg

mark.kendall at gmail

Jun 8, 2011, 12:59 AM

Post #4 of 14 (2075 views)
Permalink

Re: Support for upcoming VDPAU on AMD and Myth Player [In reply to]

2011/6/8 Marius Gr�ger <dargllun [at] googlemail>:
> Er, pretty embarrassing to write on a -dev list, but actually I didn't
> recompile MythTV for this test. I simply turned on VDPAU in the settings and
> tried to get some results. I would not think that recompiling MythTV changed
> the dependency on some nvidia specific library, except that I probably
> wouldn't even be able to even compile it without errors.

Well I'm guessing that however your version of MythTV was compiled, it
is linked against a vdpau library. In which case you may need to make
distclean, re-run configure and take it from there.

> What I suspect is that the player, while using VDPAU, also tinkers with some
> nvidia specific things. I meant to analyze the sources for more insight but
> haven't got round to it yet. If my suspicion is correct, then the generic
> VDPAU would be needed to be factored out somehow into s.th. which is
> generically usable.

VDPAU is an API. Yes - NVidia wrote it but you can't do anything
nvidia specific. You're either working with the API or you're not.

> Has anyone perhaps ever tried the VDPAU to XvBA bridge from Splitted Desktop
> with the Catalyst driver? That should show pretty much the same issues, and
> maybe they're solved already?

Do you mean VAAPI to XvBA?

Mark
_______________________________________________
mythtv-dev mailing list
mythtv-dev [at] mythtv
http://www.mythtv.org/mailman/listinfo/mythtv-dev

dargllun at googlemail

Jun 8, 2011, 5:37 AM

Post #5 of 14 (2070 views)
Permalink

Re: Support for upcoming VDPAU on AMD and Myth Player [In reply to]

2011/6/8 Mark Kendall <mark.kendall [at] gmail>

> 2011/6/8 Marius Gr�ger <dargllun [at] googlemail>:
>
> > What I suspect is that the player, while using VDPAU, also tinkers with
> some
> > nvidia specific things. I meant to analyze the sources for more insight
> but
> > haven't got round to it yet. If my suspicion is correct, then the generic
> > VDPAU would be needed to be factored out somehow into s.th. which is
> > generically usable.
>
> VDPAU is an API. Yes - NVidia wrote it but you can't do anything
> nvidia specific. You're either working with the API or you're not.
>

That was my understanding as well. So I will retry and check the logs again.
I'm away from the box right now, but from what I remember it tried to open
some kind of lib*nvidia*.so, which I obviously don't have.

>
> > Has anyone perhaps ever tried the VDPAU to XvBA bridge from Splitted
> Desktop
> > with the Catalyst driver? That should show pretty much the same issues,
> and
> > maybe they're solved already?
>
> Do you mean VAAPI to XvBA?
>

Well, yes and no. I never tried it myself, but AFAIK there are actually two
bridges, VDPAU to VAAPI, and VAAPI to XvBA. Bottom line is there is some
kind of VDPAU support, but since I was never really happy with the Catalyst
driver (bulky, ugly tearing) I have long since exclusively used the open
source driver which gives excellent Xv performance (but unfortunately poor
OpenGL rates - see also http://code.mythtv.org/trac/ticket/9216).

Greg

raymond at wagnerrp

Jun 8, 2011, 6:40 AM

Post #6 of 14 (2060 views)
Permalink

Re: Support for upcoming VDPAU on AMD and Myth Player [In reply to]

On 06/08/11 08:37, Marius Gr�ger wrote:
> Well, yes and no. I never tried it myself, but AFAIK there are
> actually two bridges, VDPAU to VAAPI, and VAAPI to XvBA. Bottom line
> is there is some kind of VDPAU support

Those drivers are not bi-directional. You can't use them to go both
directions. They are intended to allow ATI cards (XvBA) and nVidia
cards (VDPAU) to be used through VAAPI. You can't chain them to
translate commands from VDPAU, into VAAPI, into XvBA.
_______________________________________________
mythtv-dev mailing list
mythtv-dev [at] mythtv
http://www.mythtv.org/mailman/listinfo/mythtv-dev

dargllun at googlemail

Jun 8, 2011, 2:42 PM

Post #7 of 14 (2056 views)
Permalink

Re: Support for upcoming VDPAU on AMD and Myth Player [In reply to]

> VDPAU is an API. Yes - NVidia wrote it but you can't do anything
> nvidia specific. You're either working with the API or you're not.

Ok, I don't know what happened but when I tried to reproduce the issue
it seems to be gone. Meaning MythTV doesn't really complain about my
Gallium VDPAU anymore! However, output is rather jerky. Since it plays
fine in mplayer -vo vdpau, I thought I show you guys the -v
none,playback log. Perhaps you can spot anything suspicious.

I'm trying to play an mpeg2 dvb-t recording (HD won't be supported by
this vdpau implementation for now anyways). I didn't configure a
deinterlacer.

So what do you think?

Regards
Greg

Attachments:

vdpau.log (16.3 KB)

mark.kendall at gmail

Jun 9, 2011, 5:30 PM

Post #8 of 14 (2038 views)
Permalink

Re: Support for upcoming VDPAU on AMD and Myth Player [In reply to]

On 9 June 2011 05:42, Greg <dargllun [at] googlemail> wrote:
>
>> VDPAU is an API. Yes - NVidia wrote it but you can't do anything
>> nvidia specific. You're either working with the API or you're not.
>
> Ok, I don't know what happened but when I tried to reproduce the issue
> it seems to be gone. Meaning MythTV doesn't really complain about my
> Gallium VDPAU anymore! However, output is rather jerky. Since it plays
> fine in mplayer -vo vdpau, I thought I show you guys the -v
> none,playback log. Perhaps you can spot anything suspicious.
>
> I'm trying to play an mpeg2 dvb-t recording (HD won't be supported by
> this vdpau implementation for now anyways). I didn't configure a
> deinterlacer.

The most obvious 'issue' from that log is that it is using software
decode. I couldn't find any detail on the video-pipe branch so I don't
know whether it should be using the gpu for decoding or not.

Other than that, there's little of note in the logs (using straight
'-v playback' would give a more useful log) and certainly no hint as
to why the peformance isn't great.

regards

Mark
_______________________________________________
mythtv-dev mailing list
mythtv-dev [at] mythtv
http://www.mythtv.org/mailman/listinfo/mythtv-dev

jyavenard at gmail

Jun 9, 2011, 7:16 PM

Post #9 of 14 (2034 views)
Permalink

Re: Support for upcoming VDPAU on AMD and Myth Player [In reply to]

Hi

On 10 June 2011 10:30, Mark Kendall <mark.kendall [at] gmail> wrote:
> The most obvious 'issue' from that log is that it is using software
> decode. I couldn't find any detail on the video-pipe branch so I don't
> know whether it should be using the gpu for decoding or not.
>
> Other than that, there's little of note in the logs (using straight
> '-v playback' would give a more useful log) and certainly no hint as
> to why the peformance isn't great.

I have a new machine at home with an ATI graphic card ; I will try
testing this VDPAU things cool and see what I come up with.

JY
_______________________________________________
mythtv-dev mailing list
mythtv-dev [at] mythtv
http://www.mythtv.org/mailman/listinfo/mythtv-dev

elpinkus at yahoo

Jun 9, 2011, 9:00 PM

Post #10 of 14 (2047 views)
Permalink

Re: Support for upcoming VDPAU on AMD and Myth Player [In reply to]

> I have a new machine at home with an ATI graphic card ; I will try

> testing this VDPAU things cool and see what I come up with.
>
> JY

I believe MythTV sends video to the VDPAU API which in turn sends it to the nvidia proprietary driver which sends it to the hardware video decoder on a recent nvidia card.

Playing video using an AMD/ATI card's hardware decoder requires you to use AMD's XvBA API and AMD's proprietary driver, either directly or through the VA API --> XvBA backend.

I just bought a Zotac ZBOX AD02 with an AMD E-350 processor, so I'll be playing with the AMD side of things myself shortly. I don't believe there's any way to use the hardware video decoder on an AMD card through VDPAU. I would be happy to be wrong about this.

There's a patch ( http://code.mythtv.org/trac/ticket/8593 ) adding VA API support to MythTV which would give AMD users hardware video decoding but I think wasn't added due to immature drivers and general bugginess.

anaerin at gmail

Jun 9, 2011, 11:04 PM

Post #11 of 14 (2117 views)
Permalink

Re: Support for upcoming VDPAU on AMD and Myth Player [In reply to]

On 09/06/2011 10:00 PM, Zuriel wrote:
> > I have a new machine at home with an ATI graphic card ; I will try
> > testing this VDPAU things cool and see what I come up with.
> >
> > JY
>
> I believe MythTV sends video to the VDPAU API which in turn sends it to
> the nvidia proprietary driver which sends it to the hardware video
> decoder on a recent nvidia card.
>
> Playing video using an AMD/ATI card's hardware decoder requires you to
> use AMD's XvBA API and AMD's proprietary driver, either directly or
> through the VA API --> XvBA backend.
>
> I just bought a Zotac ZBOX AD02 with an AMD E-350 processor, so I'll be
> playing with the AMD side of things myself shortly. I don't believe
> there's any way to use the hardware video decoder on an AMD card through
> VDPAU. I would be happy to be wrong about this.
>
> There's a patch ( http://code.mythtv.org/trac/ticket/8593 ) adding VA
> API support to MythTV which would give AMD users hardware video decoding
> but I think wasn't added due to immature drivers and general bugginess.

If you read the whole thread, you'd find that there is some preliminary
VDPAU support in the Mesa/Gallium drivers for AMD/ATI cards, which is
what's being discussed here. It is, essentially, VDPAU->VAAPI->XvBA
bridge(s), but it enables some VDPAU support on ATI.

At the moment, Greg <dargllun [at] googlemail> has it up and running, and
has posted a log with some debug output, but he is having issues with
playback not being smooth (Whereas it is smooth using "mplayer -vo vdpau")

This Recap was brought to you by the letters V, A, T and I. :)
_______________________________________________
mythtv-dev mailing list
mythtv-dev [at] mythtv
http://www.mythtv.org/mailman/listinfo/mythtv-dev

jyavenard at gmail

Jun 10, 2011, 2:50 AM

Post #12 of 14 (2026 views)
Permalink

Re: Support for upcoming VDPAU on AMD and Myth Player [In reply to]

On 10 June 2011 14:00, Zuriel <elpinkus [at] yahoo> wrote:

> I believe MythTV sends video to the VDPAU API which in turn sends it to the
> nvidia proprietary driver which sends it to the hardware video decoder on a
> recent nvidia card.

The VDPAU library is designed to be hardware agnostic.

The VDPAU library source code is open source. If I remember correctly,
while it's very easy to add support for another vendor than nvidia
with libvdpau, it still needs to be done.

You need to tell libvdpau where the VDPAU drivers are located

> There's a patch ( http://code.mythtv.org/trac/ticket/8593 ) adding VA API
> support to MythTV which would give AMD users hardware video decoding but I
> think wasn't added due to immature drivers and general bugginess.

I'm more keen on getting VDPAU working first, because I'm sure it's
going to be far less painless to integrate the ATI vdpau drivers, with
the generic libvdpau

JY
_______________________________________________
mythtv-dev mailing list
mythtv-dev [at] mythtv
http://www.mythtv.org/mailman/listinfo/mythtv-dev

dargllun at googlemail

Jun 10, 2011, 8:35 AM

Post #13 of 14 (2015 views)
Permalink

Re: Support for upcoming VDPAU on AMD and Myth Player [In reply to]

On 10.06.2011 08:04, Robert Johnston wrote:
> If you read the whole thread, you'd find that there is some preliminary
> VDPAU support in the Mesa/Gallium drivers for AMD/ATI cards, which is
> what's being discussed here. It is, essentially, VDPAU->VAAPI->XvBA
> bridge(s), but it enables some VDPAU support on ATI.

Actually this thread is about the VDPAU state tracker for Mesa on top of
the open source r600g driver.

The VDPAU->VAAPI->XvBA bridge were brought into the discussion as
another example for non-NVidia VDPAU.

These two use cases should not be confused.

Greg
_______________________________________________
mythtv-dev mailing list
mythtv-dev [at] mythtv
http://www.mythtv.org/mailman/listinfo/mythtv-dev

dargllun at googlemail

Jun 10, 2011, 8:38 AM

Post #14 of 14 (2014 views)
Permalink

Re: Support for upcoming VDPAU on AMD and Myth Player [In reply to]

On 10.06.2011 04:16, Jean-Yves Avenard wrote:
> Hi
>
> On 10 June 2011 10:30, Mark Kendall<mark.kendall [at] gmail> wrote:
>> The most obvious 'issue' from that log is that it is using software
>> decode. I couldn't find any detail on the video-pipe branch so I don't
>> know whether it should be using the gpu for decoding or not.
>>
>> Other than that, there's little of note in the logs (using straight
>> '-v playback' would give a more useful log) and certainly no hint as
>> to why the peformance isn't great.
>
> I have a new machine at home with an ATI graphic card ; I will try
> testing this VDPAU things cool and see what I come up with.

Let me know if you need assistance in setting up the pipe-video branch
of Mesa/r600g for this experiment.

Vice versa, if you succeed in getting smooth playback with the VDPAU on
Catalyst+Splitted Desktop bridges I'd be very interested in getting
assistance from you :-)

Greg

_______________________________________________
mythtv-dev mailing list
mythtv-dev [at] mythtv
http://www.mythtv.org/mailman/listinfo/mythtv-dev

Index | Next | Previous | View Threaded

Interested in having your list archived? Contact Gossamer Threads