[<prev] [next>] [thread-next>] [day] [month] [year] [list]
Message-Id: <cover.1650453924.git.ming.qian@nxp.com>
Date: Wed, 20 Apr 2022 19:34:52 +0800
From: Ming Qian <ming.qian@....com>
To: mchehab@...nel.org, shawnguo@...nel.org
Cc: hverkuil-cisco@...all.nl, robh+dt@...nel.org,
s.hauer@...gutronix.de, kernel@...gutronix.de, festevam@...il.com,
linux-imx@....com, aisheng.dong@....com,
linux-media@...r.kernel.org, linux-kernel@...r.kernel.org,
linux-arm-kernel@...ts.infradead.org
Subject: [PATCH v3 0/2] media: amphion: copy timestamp from output to
v3
- add helper function vpu_skip_frame() and vpu_next_src_buf()
v2
- reduce the min buffers count
There is a timestamp manager in firmware,
so in previous, driver will pass the timestamp to firmware,
and when a frame is decoded or encoded,
firmware will pass the timestamp to driver,
then driver set the timestamp to the capture vb2_buffer.
But there is problem that the timestamp manager will ensure that the
timestamps are monotonically increasing.
So if there are reorder frames, the capture buffer's timestamp is
inconsistent with the timestamp of the corresponding output buffer.
It goes against the description of V4L2_BUF_FLAG_TIMESTAMP_COPY in the
specification
In order to comply with the specification, driver should copy timestamp
from output buffer to capture buffer, and there is an helper function
v4l2_m2m_buf_copy_metadata to handle it.
Ming Qian (2):
media: amphion: decoder copy timestamp from output to capture
media: amphion: encoder copy timestamp from output to capture
drivers/media/platform/amphion/vdec.c | 76 ++++++++-------------
drivers/media/platform/amphion/venc.c | 47 +++++--------
drivers/media/platform/amphion/vpu_malone.c | 2 +-
drivers/media/platform/amphion/vpu_v4l2.c | 56 +++++++++++++++
drivers/media/platform/amphion/vpu_v4l2.h | 3 +
5 files changed, 103 insertions(+), 81 deletions(-)
--
2.35.1
Powered by blists - more mailing lists