Package Details: libva-intel-driver-g45-h264 1:2.3.0-1

Git Clone URL: https://aur.archlinux.org/libva-intel-driver-g45-h264.git (read-only)
Package Base: libva-intel-driver-g45-h264
Description: Video Acceleration (VA) API for Linux
Upstream URL: https://01.org/linuxmedia/vaapi
Licenses: MIT
Conflicts: intel-gpu-tools, libva-intel-driver
Provides: libva-intel-driver
Replaces: libva-intel-driver
Submitter: pedromj
Maintainer: pedromj (alium)
Last Packager: alium
Votes: 60
Popularity: 1.025443
First Submitted: 2014-08-04 02:05
Last Updated: 2018-12-28 10:50

Required by (22)

Sources (1)

Latest Comments

« First ‹ Previous 1 2 3 4 5 6 7 8 9 10 ... Next › Last »

Raziel23 commented on 2017-04-22 15:45

@tegularius: Yes your patch works without patching mpv or ffmpeg.

@alium: Please apply the patch from @tegularius.

tegularius commented on 2017-04-22 13:44

Parts of the driver had not been patched for support h264.
It seems that the driver support for ironlake is the same as it should be for g4x-devices, so I patched it to return the same surface-attribute-information.
Works as far as I can tell.

https://gist.github.com/anonymous/e5383bd3c78bf389d124fcecde5a1556

Raziel23 commented on 2017-04-21 17:27

@tegularius: I dug a little bit about this issue. When mpv is compiled against ffmpeg 3.2.4 it uses an old vaapi infrastructure, which means that vaapi is handled by mpv player. On the other hand when mpv is compiled against ffmpeg 3.3 it uses a new API, which means that vaapi is handled by ffmpeg. They differently handle vaapi. FFmpeg rely on codec/format configuration that is returned by libva-intel-driver-g45-h264. I don't know if this can be fixed in libva-intel-driver-g45-h264 (I looked into a source code but don't have enough knowledge), but there are two other solution that works for me:

1. Use old vaapi infrastructure in mpv when compiled against ffmpeg 3.3. Disclaimer: old vaapi infrastructure maybe removed from mpv in the future. Compile mpv with the following flags:
https://gist.github.com/anonymous/3e0daaedc6fcfa1e41365696c79883ba

2. Fix ffmpeg 3.3 with patch which do: if hardware configuration returned by libva-intel-driver-g45-h264 driver doesn't have any relevant format then returns a full set of image formats and check one by one if any format is relevant to decode h264:
https://gist.github.com/anonymous/8670b535237f0e650947092586e9446d

Use either one solution. I don't know if this can be fixed in libva-intel-driver-g45-h264 package.

Edit: Upload patches to gist.

tegularius commented on 2017-04-20 19:45

Sadly doesn't seem to work, mpv says:
[vd] Pixel formats supported by decoder: vdpau vaapi_vld yuv420p
[vd] Codec profile: Main (0x4d)
[ffmpeg/video] h264: Hardware does not offer any usable surface formats.
[vd] Pixel formats supported by decoder: vdpau yuv420p
[vd] Codec profile: Main (0x4d)
[ffmpeg/video] h264: Reinit context to 1280x720, pix_fmt: yuv420p
[vd] Falling back to software decoding.

vainfo reports:
libva info: VA-API version 0.40.0
libva info: va_getDriverName() returns 0
libva info: Trying to open /usr/lib/dri/i965_drv_video.so
libva info: Found init function __vaDriverInit_0_40
libva info: va_openDriver() returns 0
vainfo: VA-API version: 0.40 (libva )
vainfo: Driver version: Intel i965 driver for Intel(R) GM45 Express Chipset - 1.8.1
vainfo: Supported profile and entrypoints
VAProfileMPEG2Simple : VAEntrypointVLD
VAProfileMPEG2Main : VAEntrypointVLD
VAProfileH264ConstrainedBaseline: VAEntrypointVLD
VAProfileH264Main : VAEntrypointVLD
VAProfileH264High : VAEntrypointVLD

pedromj commented on 2016-09-18 04:50

As we stated in previous comments, neither me nor @alium (if nothing changed) have access to G45 chipset. If any of you, current users of G45 chipsets, can maintain this package, please contribute a fixed version and become the new maintainer. Thank you.

kwsk commented on 2016-09-18 03:00

libva 1.7.2 was released

resolving dependencies...
looking for conflicting packages...
error: failed to prepare transaction (could not satisfy dependencies)
:: libva-intel-driver-g45-h264: installing libva (1.7.2-1) breaks dependency 'libva=1.7.1'

pedromj commented on 2016-01-27 15:35

Actually @alium has been the main contributor to this package. I have added him to the list of co-maintainers so he can update it when a new release is out. Moreover, I no longer have access to a machine with a G45 chipset, so I cannot test if this work before updating it.

Raziel23 commented on 2016-01-27 15:15

@alium: I repeat here on AUR what I sent to you in an email, thus all interested users will know. I found a cause of my issue [1]. After applying a patch [2] from there on intel-driver-g45-h264-1.6.2.tar.gz source green artifacts disappeared. This patch is already in mainline [3]. Please apply this patch in PKGBUILD or intel-driver-g45-h264-1.6.2.tar.gz source. I want to thank you for maintaining this source.

BTW. You should be a co-maintainer of libva-intel-driver-g45-h264 package on AUR.

[1] https://bugs.freedesktop.org/show_bug.cgi?id=93499
[2] https://bugs.freedesktop.org/attachment.cgi?id=120803
[3] http://cgit.freedesktop.org/vaapi/intel-driver/commit/?id=10c4cdd103f730a9e07ffcf0f940e8f63ee5738d

Raziel23 commented on 2016-01-25 08:47

@alium: Thank you for your response. I will send you an email with details.

Raziel23 commented on 2016-01-24 14:14

@alium: With your new version I get green artifacts at the bottom of a video. With previous version this issue doesn't exist. I don't know if this can be fixed, I just want let you know about this.