[v6,23/24] libcamera: software_isp: lut: Skip calculation lookup tables if gpuIspEnabled is true.

25332 diff mbox series

On my reference platform Qualcomm RB5 sm8520 the qcam application CPU
occupancy drops from ~100% to about 95% of a single core so this one change
sheds aprox 5% CPU usage.

Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
---
 src/ipa/simple/algorithms/lut.cpp | 70 ++++++++++++++++---------------
 1 file changed, 37 insertions(+), 33 deletions(-)

Bryan O'Donoghue <bryan.odonoghue@linaro.org> writes:

> On my reference platform Qualcomm RB5 sm8520 the qcam application CPU
> occupancy drops from ~100% to about 95% of a single core so this one change
> sheds aprox 5% CPU usage.
>
> Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>

No dot at the end of the commit title.

Reviewed-by: Milan Zamazal <mzamazal@redhat.com>

> ---
>  src/ipa/simple/algorithms/lut.cpp | 70 ++++++++++++++++---------------
>  1 file changed, 37 insertions(+), 33 deletions(-)
>
> diff --git a/src/ipa/simple/algorithms/lut.cpp b/src/ipa/simple/algorithms/lut.cpp
> index 7f223e812..618219b17 100644
> --- a/src/ipa/simple/algorithms/lut.cpp
> +++ b/src/ipa/simple/algorithms/lut.cpp
> @@ -56,32 +56,34 @@ void Lut::queueRequest(typename Module::Context &context,
>  
>  void Lut::updateGammaTable(IPAContext &context)
>  {
> -	auto &gammaTable = context.activeState.gamma.gammaTable;
>  	const auto blackLevel = context.activeState.blc.level;
> -	const unsigned int blackIndex = blackLevel * gammaTable.size() / 256;
>  	const auto contrast = context.activeState.knobs.contrast.value_or(1.0);
>  
> -	const float divisor = gammaTable.size() - blackIndex - 1.0;
> -	for (unsigned int i = blackIndex; i < gammaTable.size(); i++) {
> -		double normalized = (i - blackIndex) / divisor;
> -		/* Convert 0..2 to 0..infinity; avoid actual inifinity at tan(pi/2) */
> -		double contrastExp = tan(std::clamp(contrast * M_PI_4, 0.0, M_PI_2 - 0.00001));
> -		/* Apply simple S-curve */
> -		if (normalized < 0.5)
> -			normalized = 0.5 * std::pow(normalized / 0.5, contrastExp);
> -		else
> -			normalized = 1.0 - 0.5 * std::pow((1.0 - normalized) / 0.5, contrastExp);
> -		gammaTable[i] = UINT8_MAX *
> -				std::pow(normalized, context.configuration.gamma);
> +	if (!context.gpuIspEnabled) {
> +		auto &gammaTable = context.activeState.gamma.gammaTable;
> +		const unsigned int blackIndex = blackLevel * gammaTable.size() / 256;
> +		const float divisor = gammaTable.size() - blackIndex - 1.0;
> +		for (unsigned int i = blackIndex; i < gammaTable.size(); i++) {
> +			double normalized = (i - blackIndex) / divisor;
> +			/* Convert 0..2 to 0..infinity; avoid actual inifinity at tan(pi/2) */
> +			double contrastExp = tan(std::clamp(contrast * M_PI_4, 0.0, M_PI_2 - 0.00001));
> +			/* Apply simple S-curve */
> +			if (normalized < 0.5)
> +				normalized = 0.5 * std::pow(normalized / 0.5, contrastExp);
> +			else
> +				normalized = 1.0 - 0.5 * std::pow((1.0 - normalized) / 0.5, contrastExp);
> +			gammaTable[i] = UINT8_MAX *
> +					std::pow(normalized, context.configuration.gamma);
> +		}
> +		/*
> +		 * Due to CCM operations, the table lookup may reach indices below the black
> +		 * level. Let's set the table values below black level to the minimum
> +		 * non-black value to prevent problems when the minimum value is
> +		 * significantly non-zero (for example, when the image should be all grey).
> +		 */
> +		std::fill(gammaTable.begin(), gammaTable.begin() + blackIndex,
> +			  gammaTable[blackIndex]);
>  	}
> -	/*
> -	 * Due to CCM operations, the table lookup may reach indices below the black
> -	 * level. Let's set the table values below black level to the minimum
> -	 * non-black value to prevent problems when the minimum value is
> -	 * significantly non-zero (for example, when the image should be all grey).
> -	 */
> -	std::fill(gammaTable.begin(), gammaTable.begin() + blackIndex,
> -		  gammaTable[blackIndex]);
>  
>  	context.activeState.gamma.blackLevel = blackLevel;
>  	context.activeState.gamma.contrast = contrast;
> @@ -134,17 +136,19 @@ void Lut::prepare(IPAContext &context,
>  		auto &green = params->greenCcm;
>  		auto &blue = params->blueCcm;
>  		params->ccm = ccm;
> -		for (unsigned int i = 0; i < DebayerParams::kRGBLookupSize; i++) {
> -			red[i].r = ccmValue(i, ccm[0][0]);
> -			red[i].g = ccmValue(i, ccm[1][0]);
> -			red[i].b = ccmValue(i, ccm[2][0]);
> -			green[i].r = ccmValue(i, ccm[0][1]);
> -			green[i].g = ccmValue(i, ccm[1][1]);
> -			green[i].b = ccmValue(i, ccm[2][1]);
> -			blue[i].r = ccmValue(i, ccm[0][2]);
> -			blue[i].g = ccmValue(i, ccm[1][2]);
> -			blue[i].b = ccmValue(i, ccm[2][2]);
> -			params->gammaLut[i] = gammaTable[i / div];
> +		if (!context.gpuIspEnabled) {
> +			for (unsigned int i = 0; i < DebayerParams::kRGBLookupSize; i++) {
> +				red[i].r = ccmValue(i, ccm[0][0]);
> +				red[i].g = ccmValue(i, ccm[1][0]);
> +				red[i].b = ccmValue(i, ccm[2][0]);
> +				green[i].r = ccmValue(i, ccm[0][1]);
> +				green[i].g = ccmValue(i, ccm[1][1]);
> +				green[i].b = ccmValue(i, ccm[2][1]);
> +				blue[i].r = ccmValue(i, ccm[0][2]);
> +				blue[i].g = ccmValue(i, ccm[1][2]);
> +				blue[i].b = ccmValue(i, ccm[2][2]);
> +				params->gammaLut[i] = gammaTable[i / div];
> +			}
>  		}
>  	}