From: Eric Nelson <eric.nelson@boundarydevices.com>
To: Carlos Rafael Giani <dv@pseudoterminal.org>,
Otavio Salvador <otavio@ossystems.com.br>
Cc: "meta-freescale@yoctoproject.org"
<meta-freescale@yoctoproject.org>,
Ashwin Kirpalani <akcooper8@gmail.com>
Subject: Re: gstreamer 1.x freescale plugings
Date: Tue, 15 Oct 2013 12:16:13 -0700 [thread overview]
Message-ID: <525D947D.4060703@boundarydevices.com> (raw)
In-Reply-To: <525D8E49.9090703@pseudoterminal.org>
Thanks for the update Carlos,
On 10/15/2013 11:49 AM, Carlos Rafael Giani wrote:
> Hi,
>
> I have been working on these plugins in my spare time. Currently, there
> are two IPU plugins (a videotransform element and a sink), a GLES based
> sink using Vivante's direct textures (for smooth HD playback), a VPU
> decoder plugin, and several VPU encoder plugins. These work, but there
> are several things left to do. The encoders are relatively unfinished
> (they can encode, but need coniderably more testing), and h264
> reordering and input<->output frame association is not established yet,
> potentially messing up the timestamps.
>
Nice work getting this far!
> That said, I can reliably playback 1080p video with this. This is _not_
> a port of the existing 0.10 plugins, but written from scratch. The 0.10
> plugins have several conceptual flaws, and most importantly, are not
> built on top of the GStreamer video en/decoder base classes. A rewrite
> was just easier.
>
It's funny how that works.
> One detail that was very important to me was to avoid buffer copies as
> much as possible. With GStreamer 1.0 , defining custom allocators and
> attaching metadata to buffers is much easier to do. So for example a
> pipeline which decodes MPEG2, rotates the frame with the IPU, and
> encodes this to h264 automatically ensures the data is not unnecessarily
> copied around by the CPU. It directly wanders from VPU to IPU and VPU
> again through DMA.
>
> I am running into problems with the existing VPU wrapper library there
> (I am not directly using imx-lib). I need some way to pass user data
> through the VPU en- and decoder. That is, when I for example specify
> input data for the decoder, I need a way to also give it a user-defined
> void pointer, that is then passed through and placed into the output
> frame that corresponds to the input frame I just specified. If the
> authors of the VPU wrapper could be contacted, it would be ideal. I
> hesitate to fork it and make my fork a dependency.
>
> Also, a documentation for the VPU wrapper would be very welcome :)
>
Have you seen the file "i.MX_6Dual6Quad_VPU_API_Reference_Manual.pdf"
within the 'docs' package here (item #1):
https://community.freescale.com/docs/DOC-94809
next prev parent reply other threads:[~2013-10-15 19:16 UTC|newest]
Thread overview: 12+ messages / expand[flat|nested] mbox.gz Atom feed top
2013-10-15 13:30 gstreamer 1.x freescale plugings Alok Kumar
2013-10-15 17:03 ` Daiane Angolini
2013-10-15 17:13 ` Alok Kumar
2013-10-15 17:44 ` Daiane Angolini
2013-10-15 18:19 ` Eric Nelson
2013-10-15 18:20 ` Daiane Angolini
2013-10-15 18:39 ` Otavio Salvador
2013-10-15 18:49 ` Carlos Rafael Giani
2013-10-15 19:16 ` Eric Nelson [this message]
2013-10-15 20:09 ` Carlos Rafael Giani
2013-10-15 23:55 ` Philip Craig
2013-10-16 15:28 ` Haakon Stende
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=525D947D.4060703@boundarydevices.com \
--to=eric.nelson@boundarydevices.com \
--cc=akcooper8@gmail.com \
--cc=dv@pseudoterminal.org \
--cc=meta-freescale@yoctoproject.org \
--cc=otavio@ossystems.com.br \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.