shithub: cstory

Info  •  Files  •  Log  •  Branches

Optimise the Lanczos resampler a little

Apprently floating-point divisions are slower than multiplications

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

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

@@ -170,7 +170,7 @@

 						const double nx = 3.14159265358979323846 * kernel_input;

 						const double nxa = nx / kernel_radius;

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

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

 					}

 				}