Performance tests with the new SlidingDFT class were quite disappointing. Even in the worst scenario, fftSize=32768 and stepSize=128, the plain FFT approach is still 7.5 times faster than the sliding DFT variant... @version 0.70, 28-Jun-08 @author Hanns Holger Rutz @todo editing (addAllDep) @todo delay compensation (fftSize data seems to be missing,the necessary pre-delay seems to be fftSize / 2) @todo cache management @todo cost-calculation of potential sliding DFT @todo there's an efficient multi-channel FFT? @todo the AsyncEvent should have Span info so DocumentFramedoesn't always need to repaint
|
|
|
|
|
|