[libcamera-devel] [PATCH v2 12/14] ipa: ipu3: Cache line duration at configure call

[Jean-Michel Hautbois]

Hi Kieran,

On 11/11/2021 00:45, Kieran Bingham wrote:
> Quoting Kieran Bingham (2021-11-10 23:10:49)
>> Quoting Jean-Michel Hautbois (2021-11-10 19:58:59)
>>> We use the line duration several times in the IPAIPU3. Instead of
>>> recalculating it each time, cache the value as a utils::Duration. When
>>> the value is needed in double, call the intern count() function to get
>>> it.
>>
>> caching it I like ... using the internal count() ... sounds scary...
>>
>> Oh ... in fact, it does look like the 'correct' way to do this.
>> But I think that's a limitation on our implementation of
>> utils::Duration.
>>
>> I think we can do:
>>          maxShutterSpeed = maxExposure * lineDuration.get<std::nano>();
>>
>> But that essentially just casts a std::nano to a std::nano and then
>> calls .count() on it ... so that's a bit redundant compared to what
>> you're already doing...
>>
>> Perhaps we should have our Duration class implement a default .get()
>> which always returns nanoseconds.... (and simply calls .count() or an
>> explicit operator double() ... ?)
>>
>> Anyway, that isn't a fault of this patch as far as I can tell - so lets
>> consider that on top.
>>
>> If you prefer the explicit named cast with .get<std::nano>() I think
>> it's ok to use that as the compiler will probably optimise it ... but
>> either way is up to you I think...
>>
>> Reviewed-by: Kieran Bingham <kieran.bingham at ideasonboard.com>
> 
> Hrm... perhaps except for one more thing to check first ...
> 
> 
>>
>>> Signed-off-by: Jean-Michel Hautbois <jeanmichel.hautbois at ideasonboard.com>
>>> ---
>>>   src/ipa/ipu3/ipu3.cpp | 17 +++++++++--------
>>>   1 file changed, 9 insertions(+), 8 deletions(-)
>>>
>>> diff --git a/src/ipa/ipu3/ipu3.cpp b/src/ipa/ipu3/ipu3.cpp
>>> index eb8c0dc4..68a97e8e 100644
>>> --- a/src/ipa/ipu3/ipu3.cpp
>>> +++ b/src/ipa/ipu3/ipu3.cpp
>>> @@ -174,6 +174,8 @@ private:
>>>          uint32_t minGain_;
>>>          uint32_t maxGain_;
>>>   
>>> +       utils::Duration lineDuration_;
>>> +
>>>          /* Interface to the Camera Helper */
>>>          std::unique_ptr<CameraSensorHelper> camHelper_;
>>>   
>>> @@ -195,8 +197,8 @@ void IPAIPU3::updateSessionConfiguration(const IPACameraSensorInfo &sensorInfo,
>>>          int32_t minExposure = v4l2Exposure.min().get<int32_t>();
>>>          int32_t maxExposure = v4l2Exposure.max().get<int32_t>();
>>>   
>>> -       utils::Duration lineDuration = sensorInfo.lineLength * 1.0s
>>> -                                    / sensorInfo.pixelRate;
>>> +       lineDuration_ = sensorInfo.lineLength * 1.0s
>>> +                     / sensorInfo.pixelRate;
>>>   
>>>          const ControlInfo &v4l2Gain = sensorControls.find(V4L2_CID_ANALOGUE_GAIN)->second;
>>>          int32_t minGain = v4l2Gain.min().get<int32_t>();
>>> @@ -209,8 +211,8 @@ void IPAIPU3::updateSessionConfiguration(const IPACameraSensorInfo &sensorInfo,
>>>           *
>>>           * \todo take VBLANK into account for maximum shutter speed
>>>           */
>>> -       context_.configuration.agc.minShutterSpeed = minExposure * lineDuration;
>>> -       context_.configuration.agc.maxShutterSpeed = maxExposure * lineDuration;
>>> +       context_.configuration.agc.minShutterSpeed = minExposure * lineDuration_;
>>> +       context_.configuration.agc.maxShutterSpeed = maxExposure * lineDuration_;
>>>          context_.configuration.agc.minAnalogueGain = camHelper_->gain(minGain);
>>>          context_.configuration.agc.maxAnalogueGain = camHelper_->gain(maxGain);
>>>   }
>>> @@ -239,11 +241,10 @@ void IPAIPU3::updateControls(const IPACameraSensorInfo &sensorInfo,
>>>           * exposure min, max and default and convert it from lines to
>>>           * microseconds.
>>>           */
>>> -       double lineDuration = sensorInfo.lineLength / (sensorInfo.pixelRate / 1e6);
>>
>> I like that we get rid of this 1e6 though ;-)
> 
> It 'might' still be right - but needs a correct/thorough checking (or
> use the explicit .get<std::nano>() / .get<std::micro>() ....
> 
> The 1e6 is I assume taking us to micro's?... but...
> 
> I believe the Duration.count() is going to be returning nano seconds...
> can you validate/check that this calculation is using the correct units
> as expected please?

It returns the number of ticks for this duration in a number of ns 
indeed. In our case:

lineDuration_ = sensorInfo_.lineLength * 1.0s / sensorInfo_.pixelRate;
LOG(IPAIPU3, Debug) << "Line duration: " << lineDuration_
   << " is " << lineDuration_.count() << " ticks"
   << " and " << lineDuration.get<std::micro>() << " microseconds";

=> output is:
Line duration: 16.80us is 16800 ticks and 16.8 microseconds

So, I guess you are right ;-), we should use .get<std::micro>().


> 
> I think this makes me favour the explicit units / period in the get<>()
> call to make it clear what the calculation does..
> 
> 
> 
>>
>>>          const ControlInfo &v4l2Exposure = sensorControls.find(V4L2_CID_EXPOSURE)->second;
>>> -       int32_t minExposure = v4l2Exposure.min().get<int32_t>() * lineDuration;
>>> -       int32_t maxExposure = v4l2Exposure.max().get<int32_t>() * lineDuration;
>>> -       int32_t defExposure = v4l2Exposure.def().get<int32_t>() * lineDuration;
>>> +       int32_t minExposure = v4l2Exposure.min().get<int32_t>() * lineDuration_.count();
>>> +       int32_t maxExposure = v4l2Exposure.max().get<int32_t>() * lineDuration_.count();
>>> +       int32_t defExposure = v4l2Exposure.def().get<int32_t>() * lineDuration_.count();
>>>          controls[&controls::ExposureTime] = ControlInfo(minExposure, maxExposure,
>>>                                                          defExposure);
>>>   
>>> -- 
>>> 2.32.0
>>>