commit 0350496fef61dcb20e853b8bd55486f42936329e
author Venkat <sanampudi.venkatarao@ittiam.com> Tue Jun 26 08:41:26 2018 +0530
committer Harish Mahendrakar <harish.mahendrakar@ittiam.com> Fri Jun 29 16:22:55 2018 +0000
tree 312771cdb7721dd01d7d0f348d235ba846c3edd2
parent b48bc12c8d27b66ad8aa067c7f313b6925e4f433

Add neon optimization for selfguided_restoration

Gains w.r.t. C

8bit  ~ 2.93x
10bit ~ 2.96x
12bit ~ 2.96x

Change-Id: Icb7ba3c2160b4c023059631b696425017b6741f1

av1/common/av1_rtcd_defs.pl

diff --git a/av1/common/av1_rtcd_defs.pl b/av1/common/av1_rtcd_defs.pl
index 6aa9255..49ec496 100755
--- a/av1/common/av1_rtcd_defs.pl
+++ b/av1/common/av1_rtcd_defs.pl
@@ -288,12 +288,12 @@
 # LOOP_RESTORATION functions
 
 add_proto qw/void apply_selfguided_restoration/, "const uint8_t *dat, int width, int height, int stride, int eps, const int *xqd, uint8_t *dst, int dst_stride, int32_t *tmpbuf, int bit_depth, int highbd";
-specialize qw/apply_selfguided_restoration sse4_1 avx2/;
+specialize qw/apply_selfguided_restoration sse4_1 avx2 neon/;
 
 add_proto qw/void av1_selfguided_restoration/, "const uint8_t *dgd8, int width, int height,
                                   int dgd_stride, int32_t *flt0, int32_t *flt1, int flt_stride,
                                   int sgr_params_idx, int bit_depth, int highbd";
-specialize qw/av1_selfguided_restoration sse4_1 avx2/;
+specialize qw/av1_selfguided_restoration sse4_1 avx2 neon/;
 
 # CONVOLVE_ROUND/COMPOUND_ROUND functions