From e3c817b8cf95bce4649cfa133d835821b9389413 Mon Sep 17 00:00:00 2001 From: Chris Cannam Date: Thu, 29 Sep 2022 08:27:29 +0100 Subject: [PATCH] Make this terminology a bit more consistent --- COMPILING.md | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-) diff --git a/COMPILING.md b/COMPILING.md index 0cfa42e..557aa13 100644 --- a/COMPILING.md +++ b/COMPILING.md @@ -240,7 +240,7 @@ resampler or libsamplerate. ### FFT libraries supported The choice of FFT library makes no difference to output quality, only -to CPU usage. +to processing speed. ``` Library Build option CPP define Notes @@ -248,11 +248,12 @@ Library Build option CPP define Notes Built-in -Dfft=builtin -DUSE_BUILTIN_FFT Default except on macOS/iOS. -Accelerate -Dfft=vdsp -DHAVE_VDSP Default on macOS/iOS. +vDSP -Dfft=vdsp -DHAVE_VDSP Default on macOS/iOS (in + the Accelerate framework). Best option on these platforms. FFTW3 -Dfft=fftw -DHAVE_FFTW3 A bit faster than built-in, - a bit slower than Accelerate. + a bit slower than vDSP. GPL licence. SLEEF -Dfft=sleef -DHAVE_SLEEF Usually very fast. Not as widely