[PATCH v8 0/6] media: verisilicon: HEVC: fix 10bits handling

Benjamin Gaignard benjamin.gaignard at collabora.com
Fri Feb 3 01:16:16 PST 2023


When decoding a 10bits bitstreams HEVC driver should only expose 10bits pixel formats.
To fulfill this requirement it is needed to call hantro_reset_raw_fmt()
and to only change driver internal state in case of success.

Fluster score for HEVC (140/147) doesn't change after this series.
Fluster score for VP9 is 146/303.

version 8:
- Correct patch 4.
- Add a patch for VP9.

version 7:
- Remove unused ctx variable in hantro_try_ctrl().
- Change HANTRO_DEFAULT_BIT_DEPTH value to 8.
- Simplify hantro_check_depth_match logic.
- Keep ctx->bit_depth as integer value because it is use
  to compute buffers size for hevc.

version 6:
- Split the patches in multiple sub-patches.
- Rework hantro_reset_encoded_fmt() usage.

version 5:
- Add Nicolas's review tags
- Add Fixes tags

version 4:
- Split the change in 2 patches.
- Change hantro_check_depth_match() prototype to avoid using
  ctx->bit_depth
- Return the result of hantro_reset_raw_fmt() to the caller.
- Only set ctx->bit_depth when hantro_reset_raw_fmt() returns is ok.

Benjamin Gaignard (6):
  media: verisilicon: Do not set context src/dst formats in reset
    functions
  media: verisilicon: Do not use ctx fields as format storage when
    resetting
  media: verisilicon: Do not set ctx->bit_depth in hantro_try_ctrl()
  media: verisilicon: Do not change context bit depth before validating
    the format
  media: verisilicon: HEVC: Only propose 10 bits compatible pixels
    formats
  media: verisilicon: VP9: Only propose 10 bits compatible pixels
    formats

 .../media/platform/verisilicon/hantro_drv.c   | 49 +++++++---
 .../platform/verisilicon/hantro_postproc.c    |  2 +-
 .../media/platform/verisilicon/hantro_v4l2.c  | 90 +++++++++----------
 .../media/platform/verisilicon/hantro_v4l2.h  |  3 +-
 .../media/platform/verisilicon/imx8m_vpu_hw.c |  2 +
 5 files changed, 86 insertions(+), 60 deletions(-)

-- 
2.34.1




More information about the Linux-rockchip mailing list