Kodi Community Forum
Initial native support for DXVA2 in SVN - Time to say goodbye to your firstborns - Printable Version

+- Kodi Community Forum (https://forum.kodi.tv)
+-- Forum: Support (https://forum.kodi.tv/forumdisplay.php?fid=33)
+--- Forum: General Support (https://forum.kodi.tv/forumdisplay.php?fid=111)
+---- Forum: Windows (https://forum.kodi.tv/forumdisplay.php?fid=59)
+---- Thread: Initial native support for DXVA2 in SVN - Time to say goodbye to your firstborns (/showthread.php?tid=69306)



Awsome~! - gmags - 2010-05-13

Thanks for getting this in ya'll! My little ASUS Eee PC just got a confidence boost!


- nathanblgos - 2010-05-13

nathanblgos Wrote:it seems to work if I set non full screen in the system settings.

edit: seems like I have a new problem, after 5-10 mins of watching a show it looks like to tries to change the refresh rate to a setting that is far too high and therefore the screen goes completely blank Sad

It seems this issue of the video crashing is not limited to xbmc, it also accord when I was using mpc-hc, very weird Sad.


- pinn - 2010-05-13

Can I ask which build most of you are on as I'm getting quite a few lock ups with black screens when I try to play video?

It does seem to be refresh rate related and I know there's a patch but its not in the svn yet.

Another thing, when in (true) full screen, it appears to be over scanning as I can't even see the menus at top/bottom to change settings etc - any fix/ work around for this?

Thanks


- steelman1991 - 2010-05-13

CrystalP Wrote:I just submitted a patch in Trac 9172 (http://trac.xbmc.org/ticket/9172) to improve refresh rate switching for Vista and higher. It avoids the resource destructions/creations that throw DXVA in for a loop and delay the start of playback at best. Works great for me with DXVA, non-DXVA, true fullscreen, fake fullscreen.

elupus, since the DXVA code refactoring of a couple days ago, I hit a deadlock a few times with DXVA and refresh rate change. That's because of the OnDeviceLost()/OnDeviceReset() activity, but I'm not sure how I could present you the information. Call stacks of the two deadlocked threads?
(of course, this doesn't happen anymore in D3D9Ex with the patch I just submitted, since OnDeviceLost/Reset are not called anymore)

CrystalP - this is the first opportunity I have had to test your patch in svn (Craig over at sshcs was having problems compiling svn builds over the past few weeks). The build I have is r30092. Some of the issues I had have been resolved - such as video stutter - however I am still having audio sync issues and excessive frame drop in all combinations (both fullscreen modes, vsync enabled/disabled). Would I be right that some alteration took place to your original patch, which may have accounted for this.

I have attached a log http://pastebin.com/Pf40v1Us


- CrystalP - 2010-05-13

Seems the video reference clock doesn't detect the refresh rate change - that can cause all kinds of weird stuff.
You're OK when 23.98 is set before starting XBMC, right?

No the patch wasn't affected by anything else.


- steelman1991 - 2010-05-13

CrystalP Wrote:Seems the video reference clock doesn't detect the refresh rate change - that can cause all kinds of weird stuff.
You're OK when 23.98 is set before starting XBMC, right?

No the patch wasn't affected by anything else.

Hi Mate - from a native 23.98Hz display resolution the issues remain, though with slightly less audio to video sync issues and certainly fewer dropped frames than running from 50Hz to 23.98Hz with the adjust refresh rate enabled.

Thanks for having a look.


- nathanblgos - 2010-05-14

nathanblgos Wrote:It seems this issue of the video crashing is not limited to xbmc, it also accord when I was using mpc-hc, very weird Sad.

ok, I think this is my own fault Shocked. I'm not 100% sure yet but I believe this has been fixed by upgrading 10.4a drivers to 10.4, I will try the r30092 build tonight to see if the fullscreen issues is still happening.


- cyradix - 2010-05-14

Running build 30130 and auto refresh rate works fine... if I disable vsync in xbmc.
If vsync is enabled the refresh rate detection seems to fail and video plays at around 11fps

Still a good step forward. Keep up the good work!! Smile


- thegrinch - 2010-05-15

vanvonno Wrote:I did try VLC 1.1 with Intel graphics G45.
It is not working, no difference between HA on or off.

R30092 still doesn't work with G45, is it supposed to work at all? Are there any specific criteria for this chipset? MPC-HC works fine using DXVA.


- gulp - 2010-05-15

With Rev30179 i don't have green screen anymore on various ATI gpu.
Previously with some (not all) files, all i had was a green screen, now with the same files, there are some artifacts, like a striped macroblock over the image (visibles especially on dark scenes), but with Changeset [30179] there was a big improvement!!


- CrystalP - 2010-05-15

Same observations as gulp, 30179 gets rid of the green, this is a big step forward for ATI!
No change for a few artifacts on nVidia Purevideo wmv9 sample, like a wrong purple color instead of black, sort of striped sky until the next I-frame (or the equivalent name for wmv).
Thanks elupus for the update!


- chunk1982 - 2010-05-16

hd4670 1gb dxva2 enabled rev30199, SWEET PLAYBACK AND OFFLOADING Nod Rofl Nod
many many thanks


- kimp93 - 2010-05-16

On hd4850, tested 30ish h.264 files, no artifact !! great work elupus ! WMV/VC-1 are not working with dxva2 yet though.


- athloni - 2010-05-16

kimp93 Wrote:On hd4850, tested 30ish h.264 files, no artifact !! great work elupus ! WMV/VC-1 are not working with dxva2 yet though.

Thanks elupus!!!

Even the "Gladiator Sample" now works without green screen on my ATI 4350:
Image


- introvert - 2010-05-16

I'm having HD playback issues in XBMC on a Revo 1600 in Windows 7. Thinking of grabbing the source and compiling in order to see if this fixes the issue.

The 9.11 win32 build available on the download page from the main site... what rev is that? If the DXVA stuff is already integrated into the release build, then there's no sense in going through the effort of getting source and compiling it.