Introduce new internal driver color format
V4L2_PIX_FMT_NV12_128 and uapi color format
COLOR_FMT_NV12_128 to map with NV12 128x32
alignment. Also, clean up intermediate HAL
color formats.
Change-Id: Ia28d0f422f2777bda865c2fc6c7499cce9dabb54
Signed-off-by: Priyanka Gujjula <pgujjula@codeaurora.org>
Currently uapi header files are included from core
kernel include path since core kernel makefile also
includes them. Hence making changes to include path
in techpack makefile such that uapi header files are
included from techpack project.
Change-Id: I604b8bbb3058473b3e301cbbce112c8c93a0e963
Signed-off-by: Priyanka Gujjula <pgujjula@codeaurora.org>
Update VP9 decoder capabilities for shima to be inline
with lahaina.
Change-Id: Ieab73987d92bb093fd06391086935900b5e01768
Signed-off-by: Govindaraj Rajagopal <grajagop@codeaurora.org>
In existing approach, dynamic low latency i.e low latency hint
is restricted via specific low latency components in userspace.
This change adds one more checkpoint to ensure that video
firmware is configured for low latency only for desired codecs.
Change-Id: Iad36c3b84ddc3b94033c4bf5f0afdd3a18600111
Signed-off-by: Vikash Garodia <vgarodia@codeaurora.org>
Revise VP9 decoder capabilities and limits to
match the updated Lahaina PRD.
Change-Id: I7a92c88b21e19ab44b43a0c829491f58aa1ce492
Signed-off-by: Mihir Ganu <mganu@codeaurora.org>
Set chroma_qp_offset hfi to firmware only for client set cases.
Supported value: 0 & -12
If hfi not set, firmware proceeds with its own default value.
10-bit: default: 0
8-bit : default: Adaptive chroma_qp
Change-Id: I40262aae87b2a385c6cd7d60faa19b6adfe8b151
Signed-off-by: Govindaraj Rajagopal <grajagop@codeaurora.org>
Updated VSP base cycle entries for Shima to inline
with vperf.
Change-Id: I26e7a521434765b97cc1a508b714652d14adf2aa
Signed-off-by: Govindaraj Rajagopal <grajagop@codeaurora.org>
Revise internal timestamp data type from u64
to s64 in case of negative timestamps.
Change-Id: Ib026d504a08ce5ac67d1d962b671c37637761321
Signed-off-by: Shi Zhongbo <zhongbos@codeaurora.org>
Fix the incorrect VP9 decoder scratch buffer calculation
and match it with firmware definition of scratch buffer size.
Change-Id: Ie901082760d675153d1eb898cbe665cc6ad59bd1
Signed-off-by: Mihir Ganu <mganu@codeaurora.org>
1. Add a new EOS flag to denote EOS pending status;
2. Set all existed time stamps with EOS flag as true
after inserting EOS time stamp;
3. Ignore time stamp sorting to buffers with EOS pending
flag as true;
Change-Id: I72b041150973e2a28d1039c5f906fdb222a86081
Signed-off-by: Shi Zhongbo <zhongbos@codeaurora.org>
[1] Only one vidc node is maintained, so "sku-index" entry is
removed in dtsi.
[2] In video driver, always read fuse value if efuse_data entry
is populated
Change-Id: I86b8c38f1e71491678b54a11783292e1281f90d3
Signed-off-by: Govindaraj Rajagopal <grajagop@codeaurora.org>
1. Sufficient event should use fw min count for DCVS.
2. Input buffer count calculation is independent of Realtime/Non-Realtime.
3. Batching should be disabled if it is non-realtime session.
4. Removed misleading logs and changed buffer count logs to
convey the correct meaning.
Change-Id: Ieba6fc022549f5e767a833410ead5aa7fb52e14b
Signed-off-by: Chinmay Sawarkar <chinmays@codeaurora.org>
Fix to skip time stamp store and fetch for buffers
with codec config flag.
Change-Id: I867db519c45751ef2dadef5a323c7a4ad2741385
Signed-off-by: Shi Zhongbo <zhongbos@codeaurora.org>
Replace the obsolete do_gettimeofday API with ktime API.
Change-Id: Id105d2c531f8b9aa3315800ab3312bc6a619b792
Signed-off-by: Mihir Ganu <mganu@codeaurora.org>
For less_or_equal check, need to meet all the
conditions when return true, not only one condition.
Change-Id: Icf103721596cdcb6a4a90e1def67f3151140db85
Signed-off-by: Qiwei Liu <qiweil@codeaurora.org>
check for cvp support before any cvp
related operation.
CRs-Fixed: 2741030
Change-Id: I7a457608321f0aa5b93b6f62d7ceb76c0b00fb93
Signed-off-by: Manikanta Kanamarlapudi <kmanikan@codeaurora.org>
copy hfi buf req to hal buf req element by
element. Due to change in structure sizes,
memcpy may lead to memory corruption.
CRs-Fixed: 2741030
Change-Id: Ie4a3aeba7cac78af99355eb766a5ead755cf654b
Signed-off-by: Manikanta Kanamarlapudi <kmanikan@codeaurora.org>
Update power sequence by removing all resets during power down.
Change-Id: I88610e9c71a91a32332626bf6da88f0f64ed1c65
Signed-off-by: Mihir Ganu <mganu@codeaurora.org>
(cherry picked from commit 5cffae52eacaaf6b0698de06c11397339d904f35)
Fix p-frame count during intra-period setting for HEVC session without
layer encoding.
Change-Id: I796cac16fbb0538012f9fc97262e1877285e1803
CRs-Fixed: 2735233
Signed-off-by: Amit Shekhar <ashekhar@codeaurora.org>
Avoid adapting driver framerate to camera qbuf rate in superframe
enabled case to avoid setting incorrect hfr buffer timestamps.
Change-Id: I354e7517b9aeef7f1b01f33089b81bf7122e12e2
Signed-off-by: Akshata Sahukar <asahukar@codeaurora.org>
In steady state, the first entry in the linked list of timestamps is
always invalid. When adding a new timestamp entry, adding the
timestamp which is a duplicate of first entry leads to duplicate
linked list nodes. To avoid this, check for validity of an entry
before inserting duplicates to the linked list.
Change-Id: I47dd8fb06b9e4eb307605df6b0e2b352a2e263fe
Signed-off-by: Mihir Ganu <mganu@codeaurora.org>
Need to check against v4l2 flag instead of
hfi flag.
Change-Id: Ib15974b1cc2ee23861b47425ebe21d25ff6dc694
Signed-off-by: Qiwei Liu <qiweil@codeaurora.org>
Fix uninitialized access issue with a few variables.
Change-Id: Ib79186e0b837ce1d020546c8a3436745c5d50532
CRs-Fixed: 2729628
Signed-off-by: Amit Shekhar <ashekhar@codeaurora.org>
As per the recommended configuration, the max spec
for running an encode session in high quality is
1080p@30. Updating the same.
Change-Id: I649790bbf8a3c2264bfe6c36769717df2d2f2a55
Signed-off-by: Vikash Garodia <vgarodia@codeaurora.org>
Currently video has multiple nodes in dtsi to support/limit
sw capabilities via sku-index. Mostly restriction comes in
terms of allowed-clock-rate. So created static entries in
platform.c to configure clocks based on plaform fuse value.
So that single vidc dtsi node is sufficient and there will
not be a probe defer/failure during bootup.
For ex. Shima supports 3 sku's.
AB - max 364.8 MHz - sku_version(1) - picked from dtsi(default sku)
AA - max 201.6 MHz - sku_version(2) - clock_data_v2
AC - max 444 MHz - sku_version(0) - clock_data_v0(no fuse)
Single vidc node - "aa00000.qcom,vidc"
Change-Id: I870f1e3a24c0cbfdd92c537d3e40c4cf5400326d
Signed-off-by: Govindaraj Rajagopal <grajagop@codeaurora.org>
Update min resolution support from 128x128
to 96x96 for decoder.
CRs-Fixed: 2724675
Change-Id: I230d3c5fbdc1d476a634160de30b20e207a5f315
Signed-off-by: Manikanta Kanamarlapudi <kmanikan@codeaurora.org>
qbuf ioctl acquired bufq[port].lock in one thread and flush
call acquired registeredbufs.lock in another thread. So
thread-1 is waiting for registeredbufs.lock & thread-2 is
waiting for bufq[port].lock i.e leading to deadlock. So
added change to avoid above mentioned deadlock.
Change-Id: Ie21984fdb562ca7a09f801f036f3a78429ceab94
Signed-off-by: Govindaraj Rajagopal <grajagop@codeaurora.org>
