Robin Gareus
47d34bfbfb
FMA may return *more precise* results since there is no intermediate rounding when computing (a + b * c). (b * c) is evaluate with infinite precision and only the final result after adding a is rounded to float32. This allows for a FLT_EPSILON difference compared to rounding (b * c) first. |
||
---|---|---|
.. | ||
appleutility | ||
ardour | ||
ardouralsautil | ||
audiographer | ||
auscan | ||
backends | ||
canvas | ||
clearlooks-newer | ||
evoral | ||
fluidsynth | ||
fst | ||
gtkmm2ext | ||
hidapi | ||
libltc | ||
lua | ||
midi++2 | ||
panners | ||
pbd | ||
plugins | ||
ptformat | ||
qm-dsp | ||
surfaces | ||
temporal | ||
vamp-plugins | ||
vamp-pyin | ||
vfork | ||
vst3 | ||
waveview | ||
widgets | ||
zita-convolver | ||
zita-resampler |