vo_drm: add KMS/DRM renderer support #1811
Conversation
For now, it displays noise. Obviously this is just a placeholder for real draw functions.
With this implementation I experienced all sorts of problems, such as lags and tearing. But it's a good starting point for future improvements.
Rather than assuming the first connector to be the one that's currently connected, we iterate and choose first connector that is truly available.
|
What I have tested:
What I haven't tested:
|
video/out/vo_drm.c
Outdated
There was a problem hiding this comment.
Also, for defensive programming, talloc_zero(vo->priv, struct modeset_dev) would be better.
|
Looks pretty good. I made some comments about mostly minor issues. I don't know the DRM API, so I can't say anything about that part. |
|
Thanks. I'll fix all of these tomorrow. |
video/out/vo_drm.c
Outdated
There was a problem hiding this comment.
Instead of this, you can just write struct drm_mode_create_dumb creq = {0};.
There was a problem hiding this comment.
PS: or actually struct drm_mode_create_dumb creq = { .width = buf->width, ....
|
I added some more comments. There are just some minor things, and I'd say it's ready for merge. The OSD redrawing thing is possibly a bit more complicated, but optional. |
|
@wm4 I'll fix these ASAP I get home. |
|
I gave this a test. The image appeared, but in the next moment the monitor turned itself off. Switching to a different VT and back, the monitor was on again, showing a segmentation fault. Then I figured, I used a 16/9 video, let's try a 4/3 one. This time it worked great. I have a 16/10 monitor. So there's an issue when the video has a wider aspect than the monitor. |
|
@Gusar321: Can you please provide me with a video that has aspect ratio 2:1 so I can fix things? (This was on the list of things I mentioned I haven't tested, mainly due lack of test material.) |
|
Just use |
|
Crash confirmed, I suspect the culrpit to be |
|
So is it basically done? |
|
Yep, I don't plan on adding anything unless someone discovers more bugs, which I'm unable to. |
|
Squashed and merged. |
ghost
closed this
2 participants
This pull request introduces new video output,
vo_drm. It uses Direct Rendering Manager (DRM) and Kernel Mode Setting (KMS) to enable watching videos on Linux systems, where neither X window system nor wayland server are available. Its only dependency islibdrm(it doesn't even need OpenGL) and thus I draw a bold conclusion that in terms of dependencies, it is the most lightweight video renderer for Linux.Because of this, it doesn't use hardware acceleration at all, and heavy videos / old hardware will experience performance problems. This was tested on a 5-year-old netbook equipped with AMD C-50 (1gHz) and Radeon 6250. Old movies were played acceptably, while 720p action videos with rich ass subtitles experienced massive frame drops. This behavior could be fixed to some extent with extreme
--vf=scale=-1:180 --sws-scaler=point.I have no idea how it performs on better hardware. I'm going to purchase a powerful computer within one month.
Additionally, I (eventually) plan to develop a renderer for DRM that utilizes hardware acceleration, but it will need
libegland almost surelylibgbmthat come with their own (remarkable) chain of dependencies. In my opinion, these dependencies are what makesvo_drmattractive.