shithub: cstory

Info  •  Files  •  Log  •  Branches

Convert Lanczos resampler from double to float

Forgot that C++98 has a float version of `sin`.

--- a/src/Backends/Audio/SoftwareMixer.cpp

+++ b/src/Backends/Audio/SoftwareMixer.cpp

@@ -153,22 +153,22 @@

 				// Perform Lanczos resampling

 				const int kernel_radius = 2;

-				double accumulator = 0;

+				float accumulator = 0;

 				for (int i = -MIN(kernel_radius - 1, sound->position); i <= kernel_radius; ++i)

 				{

 					const signed char input_sample = sound->samples[(sound->position + i) % sound->frames];

-					const double kernel_input = ((double)sound->position_subsample / 0x10000) - i;

+					const float kernel_input = ((float)sound->position_subsample / 0x10000) - i;

-					if (kernel_input == 0.0)

+					if (kernel_input == 0.0f)

 					{

 						accumulator += input_sample;

 					}

 					else

 					{

-						const double nx = 3.14159265358979323846 * kernel_input;

-						const double nxa = nx / kernel_radius;

+						const float nx = 3.14159265358979323846f * kernel_input;

+						const float nxa = nx / kernel_radius;

 						accumulator += input_sample * (sin(nx) * sin(nxa) / (nx * nxa));

 					}