[Ffmpeg-devel] [PATCH] Altivec vorbis optimizations

[Luca Barbato]

Luca Barbato wrote:
[snip]

vector_fmul
vector_fmul_reverse
vector_fmul_add_add (just the step = 1 case, step = 2 still pending
correction)

5% speedup, the output is byte identical

float_to_int16

10% speedup, the output isn't byte identical due the different rounding.

The code is more or less clean.

-- 

Luca Barbato

Gentoo/linux Gentoo/PPC
http://dev.gentoo.org/~lu_zero