]> granicus.if.org Git - libvpx/commit
Add optimized vpx_sum_squares_2d_i16 for vp10.
authorGeza Lore <gezalore@gmail.com>
Fri, 12 Feb 2016 16:04:35 +0000 (16:04 +0000)
committerGeza Lore <gezalore@gmail.com>
Mon, 15 Feb 2016 16:54:52 +0000 (16:54 +0000)
commitabd00505d1c658cc106bad51369197270a299f92
tree7569efa9d28966a886ad7cce0d01e7f8cdebc496
parentd1cad9c3f5dfbc4ac0f8111439402cb60837ef00
Add optimized vpx_sum_squares_2d_i16 for vp10.

Using this we can eliminate large numbers of calls to predict intra,
and is also faster than most of the variance functions it replaces.
This is an equivalence transform so coding performance is unaffected.

Encoder speedup is approx 7% when var_tx, super_tx and ext_tx are all
enabled.

Change-Id: I0d4c83afc4a97a1826f3abd864bd68e41bb504fb
test/sum_squares_test.cc [new file with mode: 0644]
test/test.mk
vp10/encoder/rdopt.c
vpx_dsp/sum_squares.c [new file with mode: 0644]
vpx_dsp/vpx_dsp.mk
vpx_dsp/vpx_dsp_common.h
vpx_dsp/vpx_dsp_rtcd_defs.pl
vpx_dsp/x86/sum_squares_sse2.c [new file with mode: 0644]