[fritsch]

I could not parse the question, sorry.

[gurabli]

@fritsch
In the first post you write: "While we are trying to get this code tested and merged into kodi v17...", I understand that VAAPI with EGL will not be included in Jarvis final. What will happen once Jarvis final is released, the ppa here will be based on v17 development builds?

Few post before you wrote: "I don't have my Braswell anymore - gave it away after everything we could do was implemented."
This means that fine tuning and bugfixes are still to do without implementing anything (since everything possible already works). What is on the to do list for this project?

Thanks!

[Sunflux]

You said you need an actual receiver to test the TrueHD issue. But if the issue is that the video won't play in the first place, do you need an actual TrueHD supporting receiver to test that part? So long as the system thinks it supports TrueHD and tries to bitstream, does it matter if the device on the other end can decode it?

[fritsch]

You said you need an actual receiver to test the TrueHD issue. But if the issue is that the video won't play in the first place, do you need an actual TrueHD supporting receiver to test that part? So long as the system thinks it supports TrueHD and tries to bitstream, does it matter if the device on the other end can decode it?

The logfile I read did not tell anything that would be obviously wrong. Only thing wrong: "I cannot hear anything" :-)

[Milhouse]

The problem with Dolby TrueHD seems to be passthrough related (I can reproduce this problem in my latest Nvidia_Legacy test build).

However if you disable passthrough then the TrueHD audio plays without a problem (I'm guessing this is now being decoded to 8 channel PCM etc.)

But with passthrough enabled for TrueHD, the stream will pause forever at start of playback.

[fritsch]

@fritsch
In the first post you write: "While we are trying to get this code tested and merged into kodi v17...", I understand that VAAPI with EGL will not be included in Jarvis final. What will happen once Jarvis final is released, the ppa here will be based on v17 development builds?

Few post before you wrote: "I don't have my Braswell anymore - gave it away after everything we could do was implemented."
This means that fine tuning and bugfixes are still to do without implementing anything (since everything possible already works). What is on the to do list for this project?

Thanks!

I think @FernetMenta will be so kind and create a branch directly after v16 stable was released - which can then be used for endusers until v17 is out. Currently we have reworked handling of IEC encapsulated audio to support Android's passthrough in the future. Basically I don't know what will be done next. There is still a lot stuff to cleanup. Getting EGL / Videoplayer rework to v17 and test it on all platforms, including Android, AMLogic, IMX and so on is not trivial. VDPAU / VAAPI as provided in this thread is one of the most tested versions currently.

I currently think of giving VAAPI scalers a new chance - but I am not yet sure - we need to improve our scaler performance by a lot - cause if we are honest - we currently cannot upscale 1080p24 to 3840x2160p60 without pushing the GPU to a limit. Future TVs and HDMI output will be 4k@60 and users want to run the GUI in 4k then. I need to evaluate those again (adding a VPP step for deinterlacing is easy) adding another for the progressive path is not that easy. We will see. I had this implemented some years (!) ago on a baytrail and it was able to do lanczos3 optimized scaling, which is why i am quite curious. On the other hand if someone could fix up our lanczos shader to use intermediate float to save some steps would also be nice.

Currently work is ongoing to implement mipmapping for textures to make the GUI looking better. So a lot stuff that needs love and time - while just love is existent most of the time :-)

In short: Get yourself a source copy and start hacking today. When you are ready I am sure there is still a lot to do for you.

[fritsch]

The problem with Dolby TrueHD seems to be passthrough related (I can reproduce this problem in my latest Nvidia_Legacy test build).

However if you disable passthrough then the TrueHD audio plays without a problem (I'm guessing this is now being decoded to 8 channel PCM etc.)

But with passthrough enabled for TrueHD, the stream will pause forever at start of playback.

Yes - that's what I figured. I need to debug with my AVR - but only having one in the living room which is occupied most of the time when I am at home :-)

[gurabli]

@fritsch
In the first post you write: "While we are trying to get this code tested and merged into kodi v17...", I understand that VAAPI with EGL will not be included in Jarvis final. What will happen once Jarvis final is released, the ppa here will be based on v17 development builds?

Few post before you wrote: "I don't have my Braswell anymore - gave it away after everything we could do was implemented."
This means that fine tuning and bugfixes are still to do without implementing anything (since everything possible already works). What is on the to do list for this project?

Thanks!

I think @FernetMenta will be so kind and create a branch directly after v16 stable was released - which can then be used for endusers until v17 is out. Currently we have reworked handling of IEC encapsulated audio to support Android's passthrough in the future. Basically I don't know what will be done next. There is still a lot stuff to cleanup. Getting EGL / Videoplayer rework to v17 and test it on all platforms, including Android, AMLogic, IMX and so on is not trivial. VDPAU / VAAPI as provided in this thread is one of the most tested versions currently.

I currently think of giving VAAPI scalers a new chance - but I am not yet sure - we need to improve our scaler performance by a lot - cause if we are honest - we currently cannot upscale 1080p24 to 3840x2160p60 without pushing the GPU to a limit. Future TVs and HDMI output will be 4k@60 and users want to run the GUI in 4k then. I need to evaluate those again (adding a VPP step for deinterlacing is easy) adding another for the progressive path is not that easy. We will see. I had this implemented some years (!) ago on a baytrail and it was able to do lanczos3 optimized scaling, which is why i am quite curious. On the other hand if someone could fix up our lanczos shader to use intermediate float to save some steps would also be nice.

Currently work is ongoing to implement mipmapping for textures to make the GUI looking better. So a lot stuff that needs love and time - while just love is existent most of the time :-)

In short: Get yourself a source copy and start hacking today. When you are ready I am sure there is still a lot to do for you.

Thanks fritsch for summing up everything so nice and clear. Good to know and see the details. It is such a great work, the Braswell based hw is giving amazing results and htpc experience. To me, the only thing missing is hdmi 2.0 and CEC a little, but I can easily live without these, but guess Intel needs to hold back HDMI 2.0 ports to be able to sell some new hw later.

[furii]

Currently work is ongoing to implement mipmapping for textures to make the GUI looking better.

so happy to hear this! as you know from the other thread this has been bothering me for quite a while. there hadn't been any update there stating work had commenced so i was unaware. great news considering 4k tvs are becoming more common.

[Robotica]

Currently work is ongoing to implement mipmapping for textures to make the GUI looking better. So a lot stuff that needs love and time - while just love is existent most of the time :-)

whahaha! (sorry for this useless post)

[FernetMenta]

You said you need an actual receiver to test the TrueHD issue. But if the issue is that the video won't play in the first place, do you need an actual TrueHD supporting receiver to test that part? So long as the system thinks it supports TrueHD and tries to bitstream, does it matter if the device on the other end can decode it?

absolutely correct. with a simple tweak this should be able to get debugged on any system

[fritsch]

Yeah - it can obviously be debugged - but at the end - you need to test it on the AVR ... don't like to shoot builds at people that might kill their ears with it.

[Robotica]

I think @FernetMenta will be so kind and create a branch directly after v16 stable was released - which can then be used for endusers until v17 is out. Currently we have reworked handling of IEC encapsulated audio to support Android's passthrough in the future. Basically I don't know what will be done next. There is still a lot stuff to cleanup. Getting EGL / Videoplayer rework to v17 and test it on all platforms, including Android, AMLogic, IMX and so on is not trivial. VDPAU / VAAPI as provided in this thread is one of the most tested versions currently.

I currently think of giving VAAPI scalers a new chance - but I am not yet sure - we need to improve our scaler performance by a lot - cause if we are honest - we currently cannot upscale 1080p24 to 3840x2160p60 without pushing the GPU to a limit. Future TVs and HDMI output will be 4k@60 and users want to run the GUI in 4k then. I need to evaluate those again (adding a VPP step for deinterlacing is easy) adding another for the progressive path is not that easy. We will see. I had this implemented some years (!) ago on a baytrail and it was able to do lanczos3 optimized scaling, which is why i am quite curious. On the other hand if someone could fix up our lanczos shader to use intermediate float to save some steps would also be nice.

Currently work is ongoing to implement mipmapping for textures to make the GUI looking better. So a lot stuff that needs love and time - while just love is existent most of the time :-)
..

Are you interested in sharing your ideas about Vulkan and Kodi?

[fritsch]

Sharing my ideas: No ideas at all -> offtopic here. If you have some nice ideas, where it helps all our (linux) platforms, code - then you are the man.

[teki4u]

@fritsch please don't forget about the freeze/hang bug on Chromebox using ppa version. Post with all info is few pages back .
T.I.A