[libcamera-devel] [PATCH 2/2] libcamera: control: explicitly define default control values
Christian Rauch
Rauch.Christian at gmx.de
Tue Aug 23 19:50:40 CEST 2022
Hi Kieran,
Am 23.08.22 um 01:01 schrieb Kieran Bingham:
> Quoting Christian Rauch via libcamera-devel (2022-08-22 23:23:45)
>> Hi Kieran,
>>
>> Regarding all your comments, I simply replaced the "default" 0 for "def"
>> from the old constructor with explicitly setting it 0 (or its binary
>> counterpart 'false') to match the types. If 0 is the wrong "def" value
>> for those controls, then they were already wrong before :-) They were
>> just implicitly wrong. Now you have to explicitly specify a "def" value,
>> and I kept 0 for backward compatibility.
>
> Yes, they may have been implicitly incorrect before, but we shouldn't
> set them to be explictly incorrect. So this proposed series means we
> need to consider each one carefully.
>
> As mentioned though, I think we need to work out if all controls make
> sense to have a default value, and if not - how to express that it isn't
> available. I suspect that previously the omision of a default value was
> the expression that the default was 'redundant' (or unnecessary), thus
> by making it required, you need to explicitly set each one.
Deliberating the right choice of default values will take a lot of time
and I am probably also not the right person to make these choices. My
intention with this patch series was to not change the current
behaviour, but just to make it a requirement that the min/max/def values
are either explicitly set fully, or not set at all.
If the current (and proposed) behaviour of initialising by 0 is not
desired, then I would suggest that this should be handled in a separate
patch series that involves more stakeholders that know better than me
which default values are sensible.
Best,
Christian
>
> --
> Kieran
>
>
>
>> Feel free to change the default values. You should also be able to merge
>> this patch (2/2) without the first (1/2).
>>
>> Best,
>> Christian
>>
>>
>> Am 22.08.22 um 23:52 schrieb Kieran Bingham:
>>> Quoting Christian Rauch via libcamera-devel (2022-08-21 13:43:43)
>>>> Explicitly set the default value for all ControlInfo to 0, false or minimum.
>>>>
>>>> Signed-off-by: Christian Rauch <Rauch.Christian at gmx.de>
>>>> ---
>>>> src/ipa/raspberrypi/raspberrypi.cpp | 22 ++++++++++---------
>>>> src/ipa/rkisp1/rkisp1.cpp | 10 ++++-----
>>>> src/libcamera/pipeline/ipu3/ipu3.cpp | 2 +-
>>>> .../pipeline/raspberrypi/raspberrypi.cpp | 2 +-
>>>> 4 files changed, 19 insertions(+), 17 deletions(-)
>>>>
>>>> diff --git a/src/ipa/raspberrypi/raspberrypi.cpp b/src/ipa/raspberrypi/raspberrypi.cpp
>>>> index 69c73f8c..c6360a51 100644
>>>> --- a/src/ipa/raspberrypi/raspberrypi.cpp
>>>> +++ b/src/ipa/raspberrypi/raspberrypi.cpp
>>>> @@ -74,23 +74,23 @@ constexpr Duration controllerMinFrameDuration = 1.0s / 30.0;
>>>>
>>>> /* List of controls handled by the Raspberry Pi IPA */
>>>> static const ControlInfoMap::Map ipaControls{
>>>> - { &controls::AeEnable, ControlInfo(false, true) },
>>>> - { &controls::ExposureTime, ControlInfo(0, 66666) },
>>>> - { &controls::AnalogueGain, ControlInfo(1.0f, 16.0f) },
>>>> + { &controls::AeEnable, ControlInfo(false, true, false) },
>>>
>>> I would expect most of the time AutoExposure would be enabled ... so a
>>> default of false seems misleading... Yet - perhaps that's what the 0 was
>>> already implying? So - does that mean a default is unreasonable here?
>>>
>>>
>>>> + { &controls::ExposureTime, ControlInfo(0, 66666, 0) },
>>>> + { &controls::AnalogueGain, ControlInfo(1.0f, 16.0f, 0.0f) },
>>>
>>> This default is less than the minimum.
>>>
>>>
>>>> { &controls::AeMeteringMode, ControlInfo(controls::AeMeteringModeValues) },
>>>> { &controls::AeConstraintMode, ControlInfo(controls::AeConstraintModeValues) },
>>>> { &controls::AeExposureMode, ControlInfo(controls::AeExposureModeValues) },
>>>> { &controls::ExposureValue, ControlInfo(-8.0f, 8.0f, 0.0f) },
>>>> - { &controls::AwbEnable, ControlInfo(false, true) },
>>>> - { &controls::ColourGains, ControlInfo(0.0f, 32.0f) },
>>>> + { &controls::AwbEnable, ControlInfo(false, true, false) },
>>>
>>> Same as AeEnable.... it doesn't seem reasonble to say the default is
>>> disabled? I think applications probably expect the default IPA behaviour
>>> to have AE/AWB enabled, unless it is disabled explicitly ?
>>>
>>>
>>>> + { &controls::ColourGains, ControlInfo(0.0f, 32.0f, 0.0f) },
>>>> { &controls::AwbMode, ControlInfo(controls::AwbModeValues) },
>>>> { &controls::Brightness, ControlInfo(-1.0f, 1.0f, 0.0f) },
>>>> { &controls::Contrast, ControlInfo(0.0f, 32.0f, 1.0f) },
>>>> { &controls::Saturation, ControlInfo(0.0f, 32.0f, 1.0f) },
>>>> { &controls::Sharpness, ControlInfo(0.0f, 16.0f, 1.0f) },
>>>> - { &controls::ColourCorrectionMatrix, ControlInfo(-16.0f, 16.0f) },
>>>> + { &controls::ColourCorrectionMatrix, ControlInfo(-16.0f, 16.0f, 0.0f) },
>>>> { &controls::ScalerCrop, ControlInfo(Rectangle{}, Rectangle(65535, 65535, 65535, 65535), Rectangle{}) },
>>>> - { &controls::FrameDurationLimits, ControlInfo(INT64_C(33333), INT64_C(120000)) },
>>>> + { &controls::FrameDurationLimits, ControlInfo(INT64_C(33333), INT64_C(120000), INT64_C(0)) },
>>>
>>> Again, this is less than the minimum. A default should probably be
>>> specific to the mode of the sensor too ?
>>>
>>>
>>>> { &controls::draft::NoiseReductionMode, ControlInfo(controls::draft::NoiseReductionModeValues) }
>>>> };
>>>>
>>>> @@ -463,10 +463,11 @@ int IPARPi::configure(const IPACameraSensorInfo &sensorInfo,
>>>> const Duration maxSensorFrameDuration = mode_.maxFrameLength * mode_.lineLength;
>>>> ctrlMap[&controls::FrameDurationLimits] =
>>>> ControlInfo(static_cast<int64_t>(minSensorFrameDuration.get<std::micro>()),
>>>> - static_cast<int64_t>(maxSensorFrameDuration.get<std::micro>()));
>>>> + static_cast<int64_t>(maxSensorFrameDuration.get<std::micro>()),
>>>> + int64_t(0));
>>>
>>> This is smaller than the minimum. Perhaps at least here during configure
>>> it could be set to something more specific about the camera mode
>>> capability - or something that is more 'default' like 30 FPS ? (33333 ?)
>>>
>>>>
>>>> ctrlMap[&controls::AnalogueGain] =
>>>> - ControlInfo(1.0f, static_cast<float>(helper_->gain(maxSensorGainCode_)));
>>>> + ControlInfo(1.0f, static_cast<float>(helper_->gain(maxSensorGainCode_)), 0.0f);
>>>
>>> Less than minimum. I think 1.0f makes sense here too as that's unity
>>> gain.
>>>
>>>>
>>>> /*
>>>> * Calculate the max exposure limit from the frame duration limit as V4L2
>>>> @@ -478,7 +479,8 @@ int IPARPi::configure(const IPACameraSensorInfo &sensorInfo,
>>>>
>>>> ctrlMap[&controls::ExposureTime] =
>>>> ControlInfo(static_cast<int32_t>(helper_->exposure(exposureMin).get<std::micro>()),
>>>> - static_cast<int32_t>(maxShutter.get<std::micro>()));
>>>> + static_cast<int32_t>(maxShutter.get<std::micro>()),
>>>> + int32_t(0));
>>>
>>> I'm not sure an exposure time of 0 makes sense ?, but I'm not sure what
>>> a reasonable default is here either. Maybe something like the maximum of the frame
>>> duration, and maxShutter ?
>>>
>>>>
>>>> result->controlInfo = ControlInfoMap(std::move(ctrlMap), controls::controls);
>>>> return 0;
>>>> diff --git a/src/ipa/rkisp1/rkisp1.cpp b/src/ipa/rkisp1/rkisp1.cpp
>>>> index 27b4212b..2121bfd2 100644
>>>> --- a/src/ipa/rkisp1/rkisp1.cpp
>>>> +++ b/src/ipa/rkisp1/rkisp1.cpp
>>>> @@ -91,12 +91,12 @@ namespace {
>>>>
>>>> /* List of controls handled by the RkISP1 IPA */
>>>> const ControlInfoMap::Map rkisp1Controls{
>>>> - { &controls::AeEnable, ControlInfo(false, true) },
>>>> - { &controls::AwbEnable, ControlInfo(false, true) },
>>>> + { &controls::AeEnable, ControlInfo(false, true, false) },
>>>> + { &controls::AwbEnable, ControlInfo(false, true, false) },
>>>
>>> Same issue as in RPi.
>>>
>>>
>>>> { &controls::ColourGains, ControlInfo(0.0f, 3.996f, 1.0f) },
>>>> - { &controls::Brightness, ControlInfo(-1.0f, 0.993f) },
>>>> - { &controls::Contrast, ControlInfo(0.0f, 1.993f) },
>>>> - { &controls::Saturation, ControlInfo(0.0f, 1.993f) },
>>>> + { &controls::Brightness, ControlInfo(-1.0f, 0.993f, 0.0f) },
>>>> + { &controls::Contrast, ControlInfo(0.0f, 1.993f, 0.0f) },
>>>> + { &controls::Saturation, ControlInfo(0.0f, 1.993f, 0.0f) },
>>>
>>> I think 0.0 makes sense here though
>>>
>>>> { &controls::Sharpness, ControlInfo(0.0f, 10.0f, 1.0f) },
>>>> { &controls::draft::NoiseReductionMode, ControlInfo(controls::draft::NoiseReductionModeValues) },
>>>> };
>>>> diff --git a/src/libcamera/pipeline/ipu3/ipu3.cpp b/src/libcamera/pipeline/ipu3/ipu3.cpp
>>>> index 9df24603..a1bcfe49 100644
>>>> --- a/src/libcamera/pipeline/ipu3/ipu3.cpp
>>>> +++ b/src/libcamera/pipeline/ipu3/ipu3.cpp
>>>> @@ -42,7 +42,7 @@ namespace libcamera {
>>>> LOG_DEFINE_CATEGORY(IPU3)
>>>>
>>>> static const ControlInfoMap::Map IPU3Controls = {
>>>> - { &controls::draft::PipelineDepth, ControlInfo(2, 3) },
>>>> + { &controls::draft::PipelineDepth, ControlInfo(2, 3, 0) },
>>>
>>>
>>> This needs looking at. The PipelineDepth isn't something the application
>>> can set. So a default doesn't makes sense, and I'm not sure a min/max
>>> does either.
>>>
>>>
>>>
>>>> };
>>>>
>>>> class IPU3CameraData : public Camera::Private
>>>> diff --git a/src/libcamera/pipeline/raspberrypi/raspberrypi.cpp b/src/libcamera/pipeline/raspberrypi/raspberrypi.cpp
>>>> index e895584d..8fd7634d 100644
>>>> --- a/src/libcamera/pipeline/raspberrypi/raspberrypi.cpp
>>>> +++ b/src/libcamera/pipeline/raspberrypi/raspberrypi.cpp
>>>> @@ -954,7 +954,7 @@ int PipelineHandlerRPi::configure(Camera *camera, CameraConfiguration *config)
>>>> /* Add the ScalerCrop control limits based on the current mode. */
>>>> Rectangle ispMinCrop(data->ispMinCropSize_);
>>>> ispMinCrop.scaleBy(data->sensorInfo_.analogCrop.size(), data->sensorInfo_.outputSize);
>>>> - ctrlMap[&controls::ScalerCrop] = ControlInfo(ispMinCrop, Rectangle(data->sensorInfo_.analogCrop.size()));
>>>> + ctrlMap[&controls::ScalerCrop] = ControlInfo(ispMinCrop, Rectangle(data->sensorInfo_.analogCrop.size()), ispMinCrop);
>>>
>>> I think a ScalerCrop would probably expect a default of the largest size
>>> possible, not the smallest ? (At least if I were to think about
>>> 'resetting' the ScalerCrop that is).
>>>
>>> I suspect forcing a default value might be better than implicitly
>>> leaving it zero - but it means we need to give more thought to what the
>>> defaults mean or represent - or provide a way for the default to be
>>> 'unset' if it's not appropriate to have one.
>>>
>>> --
>>> Kieran
>>>
>>>
>>>>
>>>> data->controlInfo_ = ControlInfoMap(std::move(ctrlMap), result.controlInfo.idmap());
>>>>
>>>> --
>>>> 2.34.1
>>>>
More information about the libcamera-devel
mailing list