| Age | Commit message (Collapse) | Author |
|---|
|
The mainline Unicam driver unconditionally creates the embedded data
video device. Create the corresponding stream unconditionally. Drop the
warning in case of mismatch between Unicam and the CamHelper, as this
becomes a normal situation when the sensor doesn't support emebedded
data.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
|
|
Now that the CameraSensor class implements support for embedded data,
use the corresponding API instead of retrieving the embedded data format
from the subdev manually.
This changes requires sensor kernel drivers that implement the upstream
embedded data API based on V4L2 streams. As the API is under development
and not merged in the upstream kernel yet, this breaks compatibility
with the downstream Raspberry Pi kernel.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
|
|
The Vc4CameraData::platformConfigure() function configures inicam and
the ISP. The corresponding configuration steps are interleaved, making
it more difficult to read the code and follow the flow. Reorganize the
function to improve readability.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
|
|
Use the stream formatting operator for V4L2VideoFormat instead of
calling .toString() manually. As the .toString() member function is a
wrapper around the stream formatting operator, this makes the code
slightly more efficient.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
|
|
Now that the CameraSensor class implements support for embedded data,
use the corresponding API to enable or disable the embedded data stream.
This changes requires sensor kernel drivers that implement the upstream
embedded data API based on V4L2 streams. As the API is under development
and not merged in the upstream kernel yet, this breaks compatibility
with the downstream Raspberry Pi kernel.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
|
|
Some sensors support producing and transmitting embedded data over a
stream separate from the image stream. Add support for this feature in
the CameraSensor interface, and implement it for the CameraSensorRaw
class. The CameraSensorLegacy uses the default stub implementation, as
the corresponding kernel drivers don't support embedded data.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
|
|
Add a new CameraSensorRaw implementation of the CameraSensor interface
tailored to devices that implement the new V4L2 raw camera sensors API.
This new class duplicates code from the CameraSensorLegacy class. The
two classes will be refactored to share code.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
---
Changes since combined RFC:
- Set factory priority explicitly
- Set flipsAlterBayerOrder_
- Fix build warning due to missing default case in cfa switch
- Check for read-only hblank using V4L2_CTRL_FLAG_READ_ONLY
|
|
In order to support a default implementation for camera sensors when no
better implementation matches, libcamera needs to try "specialized"
implementations first and pick the default last. Make this possible by
adding a priority value for factories. Newly registered factories are
inserted in the factories list sorted by descending priority, and the
default factory uses a negative priority to be inserted as the last
element.
This mechanism may be a bit overkill in the sense that there is no
expected use cases for priorities other than trying the default last,
but the implementation is simple and easy to understand.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Stefan Klug <stefan.klug@ideasonboard.com>
---
Changes since combined RFC:
- Make the factory priority mandatory
|
|
With a camera sensor factory in place, the next step is to create an
abstract base class that all camera sensors implement, providing a
uniform API to pipeline handler. Turn all public functions of the
CameraSensor class into pure virtual functions, and move the
implementation to the CameraSensorLegacy class.
Part of the code is likely worth keeping as common helpers in a base
class. However, to follow the principle of not designing helpers with a
single user, this commit moves the whole implementation. Common helpers
will be introduced later, along with other CameraSensor subclasses.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Stefan Klug <stefan.klug@ideasonboard.com>
|
|
Introduce a factory to create CameraSensor derived classes instances by
inspecting the sensor media entity name and provide a convenience macro
to register specialized sensor handlers.
Signed-off-by: Jacopo Mondi <jacopo@jmondi.org>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Stefan Klug <stefan.klug@ideasonboard.com>
---
Changes since combined RFC:
- Fix indentation in REGISTER_CAMERA_SENSOR() macro
|
|
Update kernel headers to the rpi/v6.8/unicam/next branch at commit
1dfbc8d9a4e2. This pulls in the new V4L2 internal pads and
device-specific metadata APIs that are under development, and should not
be merged in libcamera until the APIs stabilize.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
|
|
With support for metadata in the streams API, the v4l2_meta_format
structure has been extended with width, height and bytesperline fields.
Support them in the V4L2VideoDevice getFormat() and setFormat()
functions.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
|
|
Support the newly introduced V4L2 media bus formats for metadata. This
includes generic metadata formats, and two sensor-specific embedded data
formats.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
|
|
This commit moves the check that determines whether the mode argument of
`open*()` exists into a separate function.
With that, the check is fixed because previously it failed to account
for the fact that `O_TMPFILE` is not a power of two.
Furthermore, add `assert()`s in the fortified variants that ensure that
no mode is required by the specified flags.
Signed-off-by: Barnabás Pőcze <pobrn@protonmail.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
|
|
To avoid confusion, have `__open64_2()` and `__openat64_2()` delegate to
`open64()` and `openat64()`, respectively, instead of `open()` and
`openat()`.
This does not change the behaviour because
`V4L2CompatManager::instance()->openat()` calls `openat64()` internally,
and that adds the `O_LARGEFILE` flag unconditionally.
Fixes: 1023107b6405 ("v4l2: v4l2_compat: Intercept open64, openat64, and mmap64")
Signed-off-by: Barnabás Pőcze <pobrn@protonmail.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
|
|
The matrixVlidateYaml() function is declared in the libcamera::ipa::
namespace, but defined in the libcamera:: namespace. This causes a
dynamic linking error at runtime. Fix it by moving the function
definition.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Tested-by: Jacopo Mondi <jacopo.mondi@ideasonboard.com>
Reviewed-by: Paul Elder <paul.elder@ideasonboard.com>
|
|
The YamlObject::get<T>() function template has a specialization for
double but not for float. When used in an IPA module, the issue is
caught at module load time only, when dynamic links are resolved,
causing errors such as
Failed to open IPA module shared object: /usr/lib/libcamera/ipa_rkisp1.so: undefined symbol: _ZNK9libcamera10YamlObject6GetterIfE3getERK_
Fix it by adding a float specialization. The alternative would be to use
double only in IPA modules, but the lack of enforcement at compile time
makes this dangerous.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Tested-by: Jacopo Mondi <jacopo.mondi@ideasonboard.com>
Reviewed-by: Stefan Klug <stefan.klug@ideasonboard.com>
|
|
The frame context agc.update variable is used to indicate if the ISP
histogram metering parameters need to be updated. Rename it to
updateMetering to make usage more explicit.
Suggested-by: Paul Elder <paul.elder@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
|
|
In order to be more compatible with modern hardware and APIs. This
notably allows GL implementations to directly import the buffers more
often and seems to be required for Wayland.
Further more, as we already enforce a 8 byte stride, these formats work
better for clients that don't support padding - such as libwebrtc at the
time of writing.
Tested devices:
- Librem5
- PinePhone
- Thinkpad X13s
Signed-off-by: Robert Mader <robert.mader@collabora.com>
Tested-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: Milan Zamazal <mzamazal@redhat.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Signed-off-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
|
|
All users of the Pwl::readYaml() function have been removed. The
function is not used, and is deprecated in favour of YamlObject::get().
Drop it.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
|
|
Now that deserializing a Pwl object from YAML data is possible using the
YamlObject::get() function, replace all usage of Pwl::readYaml() to
prepare for its removal.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: David Plowman <david.plowman@raspberrypi.com>
Tested-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com> # On Raspberry Pi 4
|
|
The AGC algorithm implements the AeEnable control at runtime. Move the
declaration of the control from the IPA module to the algorithm.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
|
|
The sensor's maximum shutter speed is clamped by the maximum frame
duration specified in requests. If the requested maximum frame duration
is lower than the sensor's minimum shutter speed, the Agc::process()
function will pass a minimum value higher than the maximum to the
setLimits() function, resulting in an assertion failure. Fix it by
clamping the value to both the lower and the upper bounds.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
|
|
The AGC active state and frame context both contain a variable named
maxShutterSpeed. The variable is used to limit the maximum shutter speed
when computing the exposure time and gains, but stores the maximum frame
duration, not clamped by the sensor's maximum shutter speed. Rename it
to maxFrameDuration.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: Paul Elder <paul.elder@ideasonboard.com>
|
|
The effective exposure value for each frame is split into shutter time,
analog gain and digital gain based on the AGC constraint mode and
exposure mode. The algorithm uses the modes from the active state, which
tracks the latest queued request, instead of the frame context, which
tracks the value of the controls requested for that frame. Fix it by
using the correct modes.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: Paul Elder <paul.elder@ideasonboard.com>
|
|
The condition
if (std::pow(std::floor(root), 2) < factor)
predivider = static_cast<uint8_t>(std::ceil(root));
else
predivider = static_cast<uint8_t>(std::floor(root));
can only be false when the factor's root is an integer. In that case,
std::ceil(root) and std::floor(root) will be equal. The computation can
thus be simplified by always rounding up.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
|
|
The ISP histogram parameters depends on the AE metering mode, but not on
the other AE algorithm controls. The exposure mode, constraints mode and
frame duration limits influence the behaviour of the algorithm, but not
the histogram computation parameters. Update the histogram parameters
only when AE metering mode changes.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: Paul Elder <paul.elder@ideasonboard.com>
|
|
The Agc::computeHistogramPredivider() function doesn't need to modify
its size parameter. Make it const.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: Paul Elder <paul.elder@ideasonboard.com>
|
|
The IPAFrameContext AGC documentation is lagging behind the
implementation and misses many variables. Document them.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: Paul Elder <paul.elder@ideasonboard.com>
|
|
The IPAActiveState AGC documentation is lagging behind the
implementation and misses many variables. Document them.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: Paul Elder <paul.elder@ideasonboard.com>
|
|
agc_mean_luminance.h uses utils::Duration, include the corresponding
header.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Daniel Scally <dan.scally@ideasonboard.com>
|
|
Enumerators in libcamera start with an upper case letter. Fix the
AgcConstraint::Bound enumerators.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Daniel Scally <dan.scally@ideasonboard.com>
|
|
When no metering modes are specified in the tuning file, the AGC
initialzation fails with
[0:00:46.148508875] [209] ERROR RkISP1Agc agc.cpp:46 'AeMeteringMode' parameter not found in tuning file
which results in a camera initialization failure. Fix it by downgrading
the error into a warning, and continuing the AGC initialization with the
default metering mode.
Fixes: 35233938ee5d ("ipa: rkisp1: agc: Read histogram weights from tuning file")
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: Paul Elder <paul.elder@ideasonboard.com>
|
|
Add an operation for multiplying a matrix with a vector.
Signed-off-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
|
|
Add an algorithm module to the rkisp1 IPA for crosstalk correction.
Signed-off-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Stefan Klug <stefan.klug@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
|
|
Add a class to encapsulate the functionality of fetching a matrix based
on an integer key, and interpolating if there is no exact match. This is
expected to be used by both color correction matrices / crosstalk
correction as well as lens shading correction.
A cache is included only for exact matches of the key. The caller is
expected to decide the tolererance for rounding.
Signed-off-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Stefan Klug <stefan.klug@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
|
|
Add a class to represent a Matrix object and operations for adding
matrices, multipling a matrix by a scalar, and multiplying two matrices.
Signed-off-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Stefan Klug <stefan.klug@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
|
|
Implement a specialization of the YamlObject::Getter structure to
support deserializing ipa::Pwl objects from YAML data.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
|
|
The Pwl::Pwl(const std::vector<Point> &) constructor is inefficient as
it makes a copy of the given points vector. Add a second constructor
that takes an rvalue reference to a points vector to provide move
semantics.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
|
|
Add a size() function to the Pwl class to return the number of points in
the piecewise linear function. This is useful, for instance, to validate
that all points added with append() or prepend() have been taken into
account.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
|
|
The Pwl::empty() function is a one-liner that can be easily optimized by
the compiler given the chance. Make it inline.
While at it, move the function documentation block to match the class
declaration order.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
|
|
Suffix the Doxygen \param commands with the direction for all
parameters.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
|
|
Now that YAML deserialization of Vector instances is supported through
YamlObject::get(), remove the Vector::readYaml() function. It turns out
not to be used.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
|
|
Implement a specialization of the YamlObject::Getter structure to
support deserializing ipa::Vector objects from YAML data.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Paul Elder <paul.elder@ideasonboard.com>
|
|
The YamlObject::get() function is a function template that gets fully
specialized for various types. This works fine for non-template types,
but specializing it for template types (e.g. a std::vector<U>) would
require partial template specialization, which C++ allows for classes
and variables but not functions.
To work around this problem, delegate the implementation to a new
YamlObject::Getter structure template, which will support partial
specialization.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
|
|
The YamlObject::get() function template is implemented for a set of
basic types through template specializations. The function declaration
uses std::enable_if_t<> guards to signal incorrect usage at compilation
time. This however prevents the API to be extended with additional
specializations in other compilation units.
To prepare for new specializations of the function for the ipa::Vector
and ipa::Pwl classes types, implemented in their respective compilation
units, drop the std::enable_it_t<> guards. Incorrect usage will still be
reported as link errors.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
|
|
Add the skeletal AGC module to the rkisp1 tuning script. For now it just
spits out hardcoded values.
Signed-off-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Stefan Klug <stefan.klug@ideasonboard.com>
Reviewed-by: Daniel Scally <dan.scally@ideasonboard.com>
|
|
Add a skeletal AGC module just so that we can have some AGC tuning
values that we can use to test during development of AGC in the IPAs. As
rkisp1 is the main target, we only add support for rkisp1 for now.
The parameters are mostly copied from the hardcoded values in ctt,
except for the metering modes.
Signed-off-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Stefan Klug <stefan.klug@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
|
|
Plumb controls for setting metering mode, exposure mode, constraint
mode, and frame duration limits. Also report them as available controls,
as well as in metadata.
While at it, add the missing #include for tuple, as a std::tie is used.
Signed-off-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Stefan Klug <stefan.klug@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
|
|
Add support to the rkisp1 AGC to read histogram weights from the tuning
file. As controls for selecting the metering mode are not yet supported,
for now hardcode the matrix metering mode, which is the same as what the
AGC previously hardcoded.
Signed-off-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Stefan Klug <stefan.klug@ideasonboard.com>
Reviewed-by: Daniel Scally <dan.scally@ideasonboard.com>
|
