[Bug 789240] New: Add Multi-Frame Context support for Parallel Encode use cases

["GStreamer" (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=789240

Bug ID: 789240
Summary: Add Multi-Frame Context support for Parallel Encode
use cases
Classification: Platform
Product: GStreamer
Version: git master
OS: Linux
Status: NEW
Severity: enhancement
Priority: Normal
Component: gstreamer-vaapi
Assignee: gstreamer-***@lists.freedesktop.org
Reporter: ***@gmail.com
QA Contact: gstreamer-***@lists.freedesktop.org
CC: ***@gmail.com, ***@igalia.com
GNOME version: ---

VA-API is going to bring some new apis to support Multi-Frame processing.


Currently, Multi-frame processing is an optimization for making better hardware
resource utilization in multi-stream encoding/transcoding applications.
This allows combining several jobs from different parallel pipelines inside one
GPU task execution to better reuse engines that can't be loaded fully in single
frame case, as well as decrease CPU overhead for task submission.

Here is the Pull Request landed in
libva:https://github.com/01org/libva/pull/112

I am not sure how we could implement this in gstreamer.

These new APIs bringing the idea of a having a single "super-encoder" element
which can accept multiple input streams and encode them to different formats
like one h264 and the other to hevc or multiple h264 encoded streams.

From VA-API perspective there could be 3-4 new APIs

vaCreateMFContext(VADisplay dpy, VAMFContextID *mf_context) : Create a new
MultiFrameContext

vaMFAddContext (VADisplay dpy, VAMFContextID mf_context, VAContextID context):
To add an already created VAContext to the MFContext

vaMFReleaseContext (VADisplay dpy, VAMFContextID mf_context, VAContextID
context): To release the vacontext from global context

vaMFSubmit (VADisplay dpy, VAMFContextID mf_context, VAContextID * contexts,
int num_contexts): Make the end of rendering for a pictures in contexts passed
with submission

vaQueryXXX: needed for checking the feature support for specific entrypoint
combinations. This not not included in current PullRequest though.


Fron Intel-vaapi-driver perspective,
If a context is attached to MultiFrame Context, underlined driver won't begin
the actual batchbuffer submission in vaEndPicture(), instead
it waits for the vaMFSubmit from middleware to start actual processing.

["GStreamer" (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=789240

Hyunjun Ko <***@igalia.com> changed:

What |Removed |Added
----------------------------------------------------------------------------
CC| |***@igalia.com

["GStreamer" (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=789240

GStreamer system administrator <***@gstreamer.net> changed:

What |Removed |Added
----------------------------------------------------------------------------
Status|NEW |RESOLVED
Resolution|--- |OBSOLETE

--- Comment #1 from GStreamer system administrator <***@gstreamer.net> ---
-- GitLab Migration Automatic Message --

This bug has been migrated to freedesktop.org's GitLab instance and has been
closed from further activity.

You can subscribe and participate further through the new bug through this link
to our GitLab instance:
https://gitlab.freedesktop.org/gstreamer/gstreamer-vaapi/issues/71.