Re: [RFC 0/5] Split iMX8MQ Hantro VPU into G1 and G2 with blk-ctrl support

[Adam Ford <aford173@gmail.com>]

On Mon, Dec 6, 2021 at 4:30 PM Adam Ford <aford173@gmail.com> wrote:
>
> On Mon, Dec 6, 2021 at 2:51 PM Nicolas Dufresne <nicolas@ndufresne.ca> wrote:
> >
> > Le dimanche 05 décembre 2021 à 13:51 -0600, Adam Ford a écrit :
> > > On Sun, Dec 5, 2021 at 12:16 PM Adam Ford <aford173@gmail.com> wrote:
> > > >
> > > > Currently, the VPU in the i.MQ8MQ is appearing as one codec, but in reality, it's
> > > > two IP blocks called G1 and G2.  There is initialization code in VPU code to
> > > > pull some clocks, resets and other features which has been integrated into
> > > > the vpu-blk-ctrl for the i.MX8M Mini and a similar method can be used to make
> > > > the VPU codec's operate as stand-alone cores without having to know the details
> > > > of each other or the quirks unique to the i.MX8MQ, so the remaining code can be
> > > > left more generic.
> > > >
> > > > This series was started by Lucas Stach, and picked up by me so some patches have
> > > > his s-o-b and mine where I might have changed a few minor items.  It's in an RFC state
> > > > because the G2 VP9 operations appear to hang, but the parent code from which I started doesn't
> > > > appear to show VP9 support, and it looks like it should.
> > > >
> > > > since the g-streamer and media trees are in a constant state of change, this series is based on
> > > > git://linuxtv.org/hverkuil/media_tree.git for-v5.17e
> > > >
> > >
> > > I forgot to post Fluster results.
> > >
> > > Before the patches to this branch:
> > >
> > > 2gst-main] root@localhost:~/gstreamer/fluster# ./fluster.py list -c |grep -i v4l
> > >
> > >     GStreamer-H.264-V4L2-Gst1.0: GStreamer H.264 V4L2 decoder for
> > > GStreamer 1.0... ❌
> > >     GStreamer-H.264-V4L2SL-Gst1.0: GStreamer H.264 V4L2SL decoder for
> > > GStreamer 1.0... ✔️
> > >     GStreamer-H.265-V4L2-Gst1.0: GStreamer H.265 V4L2 decoder for
> > > GStreamer 1.0... ❌
> > >     GStreamer-H.265-V4L2SL-Gst1.0: GStreamer H.265 V4L2SL decoder for
> > > GStreamer 1.0... ❌
> > >     GStreamer-VP8-V4L2-Gst1.0: GStreamer VP8 V4L2 decoder for GStreamer 1.0... ❌
> > >     GStreamer-VP8-V4L2SL-Gst1.0: GStreamer VP8 V4L2SL decoder for
> > > GStreamer 1.0... ✔️
> > >     GStreamer-VP9-V4L2SL-Gst1.0: GStreamer VP9 V4L2SL decoder for
> > > GStreamer 1.0... ❌
> > >     GStreamer-AV1-V4L2SL-Gst1.0: GStreamer AV1 V4L2SL decoder for
> > > GStreamer 1.0... ❌
> > > [gst-main] root@localhost:~/gstreamer/fluster#
> > >
> > > There was no VP9 support.
> > >
> > > ./fluster.py run -dGStreamer-H.264-V4L2SL-Gst1.0
> > > GStreamer-H.264-V4L2SL-Gst1.0
> > > Ran 90/135 tests successfully               in 58.787 secs
> > >
> > > AFTER this series, two decoders appear:
> > >
> > > [   15.919137] hantro-vpu 38300000.video-codec: registered
> > > nxp,imx8mq-vpu-g1-dec as /dev/video0
> > > [   15.983579] hantro-vpu 38310000.video-codec: registered
> > > nxp,imx8mq-vpu-g2-dec as /dev/video1
> > >
> > > VP9 was listed:
> > >
> > >     GStreamer-H.264-V4L2-Gst1.0: GStreamer H.264 V4L2 decoder for
> > > GStreamer 1.0... ❌
> > >     GStreamer-H.264-V4L2SL-Gst1.0: GStreamer H.264 V4L2SL decoder for
> > > GStreamer 1.0... ✔️
> > >     GStreamer-H.265-V4L2-Gst1.0: GStreamer H.265 V4L2 decoder for
> > > GStreamer 1.0... ❌
> > >     GStreamer-H.265-V4L2SL-Gst1.0: GStreamer H.265 V4L2SL decoder for
> > > GStreamer 1.0... ❌
> > >     GStreamer-VP8-V4L2-Gst1.0: GStreamer VP8 V4L2 decoder for GStreamer 1.0... ❌
> > >     GStreamer-VP8-V4L2SL-Gst1.0: GStreamer VP8 V4L2SL decoder for
> > > GStreamer 1.0... ✔️
> > >     GStreamer-VP9-V4L2SL-Gst1.0: GStreamer VP9 V4L2SL decoder for
> > > GStreamer 1.0... ✔️
> > >     GStreamer-AV1-V4L2SL-Gst1.0: GStreamer AV1 V4L2SL decoder for
> > > GStreamer 1.0... ❌
> > >
> > > ./fluster.py run -d GStreamer-VP8-V4L2SL-Gst1.0
> > > Ran 55/61 tests successfully               in 8.565 secs
> > >
> > >
> > > ./fluster.py run -dGStreamer-H.264-V4L2SL-Gst1.0
> > > Ran 90/135 tests successfully               in 60.269 secs
> > >
> > > Same results for H.264
> > >
> > > VP9 Hangs, where it didn't even appear as available before:
> > >
> > > ./fluster.py run -d GStreamer-VP9-V4L2SL-Gst1.0
> > > ****************************************************************************************************
> > > Running test suite VP9-TEST-VECTORS with decoder GStreamer-VP9-V4L2SL-Gst1.0
> > > Using 4 parallel job(s)
> > > ****************************************************************************************************
> > >
> > > [TEST SUITE      ] (DECODER                    ) TEST VECTOR
> > >                           ... RESULT
> > > ----------------------------------------------------------------------
> > > [VP9-TEST-VECTORS] (GStreamer-VP9-V4L2SL-Gst1.0)
> > > vp90-2-00-quantizer-00.webm                     ... Success
> > > [VP9-TEST-VECTORS] (GStreamer-VP9-V4L2SL-Gst1.0)
> > > vp90-2-00-quantizer-01.webm                     ... Success
> > > [VP9-TEST-VECTORS] (GStreamer-VP9-V4L2SL-Gst1.0)
> > > vp90-2-00-quantizer-02.webm                     ... Succes
> > >
> > > <hang > - and yes, 'Success' didnt' finish writing to the serial port.
> >
> > Looks like hope to me ! Do you get further with -j 1 (one concurrent decode) ?
>
> Anything after the first attempt appears to fail.  I haven't figured
> out if it's the G2 code or the power domain stuff.  Since VP9 didn't
> work I migrated it to use the vpu-blk-ctrl, I can't really compare to
> know if it was a regression or an improvement.
>
> :-)
>
I managed to get the vpu-blk-ctrl updated to address the G2 hanging.

./fluster.py run -d GStreamer-VP9-V4L2SL-Gst1.0
Ran 144/303 tests successfully               in 224.847 secs

Since I couldn't get it to work before the split, I am not sure if
this is good or not.  I also wasn't sure if I needed to allocate more
CMA for this.

If/when people can give me some feedback on the RFC's, I'll submit an
updated formal patch for both the i.MQ8MQ and the i.MQ8M Mini as one
series based on  "for-v5.17e" from
git://linuxtv.org/hverkuil/media_tree.git

I wasn't sure if I needed to break up the migration from a combined
VPU to split G1 and G2.  I might need some help/suggestions on the
YAML, because I struggle with that part a little, but I'd like the 8MM
to use the 8MQ dt-binding yaml file since they are so similar.

I hope to have some more time this week because my weekends are
completely booked for the rest of the year, and leave for London, on
23 Dec 2021 and I won't return until 5 Jan 2022.


> adam
>
> >
> > >
> > >
> > >
> > > > Adam Ford (2):
> > > >   media: hantro: split i.MX8MQ G1 and G2 code
> > > >   arm64: dts: imx8mq: Split i.MX8MQ G1 and G2 with vpu-blk-ctrl
> > > >
> > > > Lucas Stach (3):
> > > >   dt-bindings: power: imx8mq: add defines for VPU blk-ctrl domains
> > > >   dt-bindings: soc: add binding for i.MX8MQ VPU blk-ctrl
> > > >   soc: imx: imx8m-blk-ctrl: add i.MX8MQ VPU blk-ctrl
> > > >
> > > >  .../soc/imx/fsl,imx8mq-vpu-blk-ctrl.yaml      |  71 +++++++++++
> > > >  arch/arm64/boot/dts/freescale/imx8mq.dtsi     |  69 ++++++----
> > > >  drivers/soc/imx/imx8m-blk-ctrl.c              |  67 ++++++++++
> > > >  drivers/staging/media/hantro/hantro_drv.c     |   4 +-
> > > >  drivers/staging/media/hantro/hantro_hw.h      |   2 +-
> > > >  drivers/staging/media/hantro/imx8m_vpu_hw.c   | 119 +++---------------
> > > >  include/dt-bindings/power/imx8mq-power.h      |   3 +
> > > >  7 files changed, 205 insertions(+), 130 deletions(-)
> > > >  create mode 100644 Documentation/devicetree/bindings/soc/imx/fsl,imx8mq-vpu-blk-ctrl.yaml
> > > >
> > > > --
> > > > 2.32.0
> > > >
> >