aom_dsp/aom_dsp_rtcd_defs.pl

   1 ##
   2 ## Copyright (c) 2017, Alliance for Open Media. All rights reserved
   3 ##
   4 ## This source code is subject to the terms of the BSD 2 Clause License and
   5 ## the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
   6 ## was not distributed with this source code in the LICENSE file, you can
   7 ## obtain it at www.aomedia.org/license/software. If the Alliance for Open
   8 ## Media Patent License 1.0 was not distributed with this source code in the
   9 ## PATENTS file, you can obtain it at www.aomedia.org/license/patent.
  10 ##
  11 sub aom_dsp_forward_decls() {
  12 print <<EOF
  13 /*
  14  * DSP
  15  */
  16
  17 #include "aom/aom_integer.h"
  18 #include "aom_dsp/aom_dsp_common.h"
  19 #include "av1/common/enums.h"
  20 #include "av1/common/blockd.h"
  21
  22 EOF
  23 }
  24 forward_decls qw/aom_dsp_forward_decls/;
  25
  26 # optimizations which depend on multiple features
  27 $avx2_ssse3 = '';
  28 if ((aom_config("HAVE_AVX2") eq "yes") && (aom_config("HAVE_SSSE3") eq "yes")) {
  29   $avx2_ssse3 = 'avx2';
  30 }
  31
  32 # functions that are 64 bit only.
  33 $mmx_x86_64 = $sse2_x86_64 = $ssse3_x86_64 = $avx_x86_64 = $avx2_x86_64 = '';
  34 if ($opts{arch} eq "x86_64") {
  35   $mmx_x86_64 = 'mmx';
  36   $sse2_x86_64 = 'sse2';
  37   $ssse3_x86_64 = 'ssse3';
  38   $avx_x86_64 = 'avx';
  39   $avx2_x86_64 = 'avx2';
  40 }
  41
  42 @block_widths = (4, 8, 16, 32, 64, 128);
  43
  44 @block_sizes = ();
  45 foreach $w (@block_widths) {
  46   foreach $h (@block_widths) {
  47     push @block_sizes, [$w, $h] if ($w <= 2*$h && $h <= 2*$w) ;
  48   }
  49 }
  50 push @block_sizes, [4, 16];
  51 push @block_sizes, [16, 4];
  52 push @block_sizes, [8, 32];
  53 push @block_sizes, [32, 8];
  54 push @block_sizes, [16, 64];
  55 push @block_sizes, [64, 16];
  56
  57 @tx_dims = (2, 4, 8, 16, 32, 64);
  58 @tx_sizes = ();
  59 foreach $w (@tx_dims) {
  60   push @tx_sizes, [$w, $w];
  61   foreach $h (@tx_dims) {
  62     push @tx_sizes, [$w, $h] if ($w >=4 && $h >=4 && ($w == 2*$h || $h == 2*$w));
  63     push @tx_sizes, [$w, $h] if ($w >=4 && $h >=4 && ($w == 4*$h || $h == 4*$w));
  64   }
  65 }
  66
  67 @pred_names = qw/dc dc_top dc_left dc_128 v h paeth smooth smooth_v smooth_h/;
  68
  69 #
  70 # Intra prediction
  71 #
  72
  73 foreach (@tx_sizes) {
  74   ($w, $h) = @$_;
  75   foreach $pred_name (@pred_names) {
  76     add_proto "void", "aom_${pred_name}_predictor_${w}x${h}",
  77               "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
  78     add_proto "void", "aom_highbd_${pred_name}_predictor_${w}x${h}",
  79               "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
  80   }
  81 }
  82
  83 specialize qw/aom_dc_top_predictor_4x4 msa neon sse2/;
  84 specialize qw/aom_dc_top_predictor_4x8 sse2/;
  85 specialize qw/aom_dc_top_predictor_8x4 sse2/;
  86 specialize qw/aom_dc_top_predictor_8x8 neon msa sse2/;
  87 specialize qw/aom_dc_top_predictor_8x16 sse2/;
  88 specialize qw/aom_dc_top_predictor_16x8 sse2/;
  89 specialize qw/aom_dc_top_predictor_16x16 neon msa sse2/;
  90 specialize qw/aom_dc_top_predictor_16x32 sse2/;
  91 specialize qw/aom_dc_top_predictor_16x64 sse2/;
  92 specialize qw/aom_dc_top_predictor_32x16 sse2 avx2/;
  93 specialize qw/aom_dc_top_predictor_32x32 msa neon sse2 avx2/;
  94 specialize qw/aom_dc_top_predictor_32x64 sse2 avx2/;
  95 specialize qw/aom_dc_top_predictor_64x64 sse2 avx2/;
  96 specialize qw/aom_dc_top_predictor_64x32 sse2 avx2/;
  97 specialize qw/aom_dc_top_predictor_64x16 sse2 avx2/;
  98 specialize qw/aom_dc_left_predictor_4x4 msa neon sse2/;
  99 specialize qw/aom_dc_left_predictor_4x8 sse2/;
 100 specialize qw/aom_dc_left_predictor_8x4 sse2/;
 101 specialize qw/aom_dc_left_predictor_8x8 neon msa sse2/;
 102 specialize qw/aom_dc_left_predictor_8x16 sse2/;
 103 specialize qw/aom_dc_left_predictor_16x8 sse2/;
 104 specialize qw/aom_dc_left_predictor_16x16 neon msa sse2/;
 105 specialize qw/aom_dc_left_predictor_16x32 sse2/;
 106 specialize qw/aom_dc_left_predictor_16x64 sse2/;
 107 specialize qw/aom_dc_left_predictor_32x16 sse2 avx2/;
 108 specialize qw/aom_dc_left_predictor_32x32 msa neon sse2 avx2/;
 109 specialize qw/aom_dc_left_predictor_32x64 sse2 avx2/;
 110 specialize qw/aom_dc_left_predictor_64x64 sse2 avx2/;
 111 specialize qw/aom_dc_left_predictor_64x32 sse2 avx2/;
 112 specialize qw/aom_dc_left_predictor_64x16 sse2 avx2/;
 113 specialize qw/aom_dc_128_predictor_4x4 msa neon sse2/;
 114 specialize qw/aom_dc_128_predictor_4x8 sse2/;
 115 specialize qw/aom_dc_128_predictor_8x4 sse2/;
 116 specialize qw/aom_dc_128_predictor_8x8 neon msa sse2/;
 117 specialize qw/aom_dc_128_predictor_8x16 sse2/;
 118 specialize qw/aom_dc_128_predictor_16x8 sse2/;
 119 specialize qw/aom_dc_128_predictor_16x16 neon msa sse2/;
 120 specialize qw/aom_dc_128_predictor_16x32 sse2/;
 121 specialize qw/aom_dc_128_predictor_16x64 sse2/;
 122 specialize qw/aom_dc_128_predictor_32x16 sse2 avx2/;
 123 specialize qw/aom_dc_128_predictor_32x32 msa neon sse2 avx2/;
 124 specialize qw/aom_dc_128_predictor_32x64 sse2 avx2/;
 125 specialize qw/aom_dc_128_predictor_64x64 sse2 avx2/;
 126 specialize qw/aom_dc_128_predictor_64x32 sse2 avx2/;
 127 specialize qw/aom_dc_128_predictor_64x16 sse2 avx2/;
 128 specialize qw/aom_v_predictor_4x4 neon msa sse2/;
 129 specialize qw/aom_v_predictor_4x8 sse2/;
 130 specialize qw/aom_v_predictor_8x4 sse2/;
 131 specialize qw/aom_v_predictor_8x8 neon msa sse2/;
 132 specialize qw/aom_v_predictor_8x16 sse2/;
 133 specialize qw/aom_v_predictor_16x8 sse2/;
 134 specialize qw/aom_v_predictor_16x16 neon msa sse2/;
 135 specialize qw/aom_v_predictor_16x32 sse2/;
 136 specialize qw/aom_v_predictor_16x64 sse2/;
 137 specialize qw/aom_v_predictor_32x16 sse2 avx2/;
 138 specialize qw/aom_v_predictor_32x32 neon msa sse2 avx2/;
 139 specialize qw/aom_v_predictor_32x64 sse2 avx2/;
 140 specialize qw/aom_v_predictor_64x64 sse2 avx2/;
 141 specialize qw/aom_v_predictor_64x32 sse2 avx2/;
 142 specialize qw/aom_v_predictor_64x16 sse2 avx2/;
 143 specialize qw/aom_h_predictor_4x8 sse2/;
 144 specialize qw/aom_h_predictor_4x4 neon dspr2 msa sse2/;
 145 specialize qw/aom_h_predictor_8x4 sse2/;
 146 specialize qw/aom_h_predictor_8x8 neon dspr2 msa sse2/;
 147 specialize qw/aom_h_predictor_8x16 sse2/;
 148 specialize qw/aom_h_predictor_16x8 sse2/;
 149 specialize qw/aom_h_predictor_16x16 neon dspr2 msa sse2/;
 150 specialize qw/aom_h_predictor_16x32 sse2/;
 151 specialize qw/aom_h_predictor_16x64 sse2/;
 152 specialize qw/aom_h_predictor_32x16 sse2/;
 153 specialize qw/aom_h_predictor_32x32 neon msa sse2 avx2/;
 154 specialize qw/aom_h_predictor_32x64 sse2/;
 155 specialize qw/aom_h_predictor_64x64 sse2/;
 156 specialize qw/aom_h_predictor_64x32 sse2/;
 157 specialize qw/aom_h_predictor_64x16 sse2/;
 158 specialize qw/aom_paeth_predictor_4x4 ssse3/;
 159 specialize qw/aom_paeth_predictor_4x8 ssse3/;
 160 specialize qw/aom_paeth_predictor_8x4 ssse3/;
 161 specialize qw/aom_paeth_predictor_8x8 ssse3/;
 162 specialize qw/aom_paeth_predictor_8x16 ssse3/;
 163 specialize qw/aom_paeth_predictor_16x8 ssse3 avx2/;
 164 specialize qw/aom_paeth_predictor_16x16 ssse3 avx2/;
 165 specialize qw/aom_paeth_predictor_16x32 ssse3 avx2/;
 166 specialize qw/aom_paeth_predictor_16x64 ssse3 avx2/;
 167 specialize qw/aom_paeth_predictor_32x16 ssse3 avx2/;
 168 specialize qw/aom_paeth_predictor_32x32 ssse3 avx2/;
 169 specialize qw/aom_paeth_predictor_32x64 ssse3 avx2/;
 170 specialize qw/aom_paeth_predictor_64x32 ssse3 avx2/;
 171 specialize qw/aom_paeth_predictor_64x64 ssse3 avx2/;
 172 specialize qw/aom_paeth_predictor_64x16 ssse3 avx2/;
 173 specialize qw/aom_paeth_predictor_16x8 ssse3/;
 174 specialize qw/aom_paeth_predictor_16x16 ssse3/;
 175 specialize qw/aom_paeth_predictor_16x32 ssse3/;
 176 specialize qw/aom_paeth_predictor_32x16 ssse3/;
 177 specialize qw/aom_paeth_predictor_32x32 ssse3/;
 178 specialize qw/aom_smooth_predictor_4x4 ssse3/;
 179 specialize qw/aom_smooth_predictor_4x8 ssse3/;
 180 specialize qw/aom_smooth_predictor_8x4 ssse3/;
 181 specialize qw/aom_smooth_predictor_8x8 ssse3/;
 182 specialize qw/aom_smooth_predictor_8x16 ssse3/;
 183 specialize qw/aom_smooth_predictor_16x8 ssse3/;
 184 specialize qw/aom_smooth_predictor_16x16 ssse3/;
 185 specialize qw/aom_smooth_predictor_16x32 ssse3/;
 186 specialize qw/aom_smooth_predictor_16x64 ssse3/;
 187 specialize qw/aom_smooth_predictor_32x16 ssse3/;
 188 specialize qw/aom_smooth_predictor_32x32 ssse3/;
 189 specialize qw/aom_smooth_predictor_32x64 ssse3/;
 190 specialize qw/aom_smooth_predictor_64x64 ssse3/;
 191 specialize qw/aom_smooth_predictor_64x32 ssse3/;
 192 specialize qw/aom_smooth_predictor_64x16 ssse3/;
 193
 194 specialize qw/aom_smooth_v_predictor_4x4 ssse3/;
 195 specialize qw/aom_smooth_v_predictor_4x8 ssse3/;
 196 specialize qw/aom_smooth_v_predictor_4x16 ssse3/;
 197 specialize qw/aom_smooth_v_predictor_8x4 ssse3/;
 198 specialize qw/aom_smooth_v_predictor_8x8 ssse3/;
 199 specialize qw/aom_smooth_v_predictor_8x16 ssse3/;
 200 specialize qw/aom_smooth_v_predictor_8x32 ssse3/;
 201 specialize qw/aom_smooth_v_predictor_16x4 ssse3/;
 202 specialize qw/aom_smooth_v_predictor_16x8 ssse3/;
 203 specialize qw/aom_smooth_v_predictor_16x16 ssse3/;
 204 specialize qw/aom_smooth_v_predictor_16x32 ssse3/;
 205 specialize qw/aom_smooth_v_predictor_16x64 ssse3/;
 206 specialize qw/aom_smooth_v_predictor_32x8 ssse3/;
 207 specialize qw/aom_smooth_v_predictor_32x16 ssse3/;
 208 specialize qw/aom_smooth_v_predictor_32x32 ssse3/;
 209 specialize qw/aom_smooth_v_predictor_32x64 ssse3/;
 210 specialize qw/aom_smooth_v_predictor_64x64 ssse3/;
 211 specialize qw/aom_smooth_v_predictor_64x32 ssse3/;
 212 specialize qw/aom_smooth_v_predictor_64x16 ssse3/;
 213
 214 specialize qw/aom_smooth_h_predictor_4x4 ssse3/;
 215 specialize qw/aom_smooth_h_predictor_4x8 ssse3/;
 216 specialize qw/aom_smooth_h_predictor_4x16 ssse3/;
 217 specialize qw/aom_smooth_h_predictor_8x4 ssse3/;
 218 specialize qw/aom_smooth_h_predictor_8x8 ssse3/;
 219 specialize qw/aom_smooth_h_predictor_8x16 ssse3/;
 220 specialize qw/aom_smooth_h_predictor_8x32 ssse3/;
 221 specialize qw/aom_smooth_h_predictor_16x4 ssse3/;
 222 specialize qw/aom_smooth_h_predictor_16x8 ssse3/;
 223 specialize qw/aom_smooth_h_predictor_16x16 ssse3/;
 224 specialize qw/aom_smooth_h_predictor_16x32 ssse3/;
 225 specialize qw/aom_smooth_h_predictor_16x64 ssse3/;
 226 specialize qw/aom_smooth_h_predictor_32x8 ssse3/;
 227 specialize qw/aom_smooth_h_predictor_32x16 ssse3/;
 228 specialize qw/aom_smooth_h_predictor_32x32 ssse3/;
 229 specialize qw/aom_smooth_h_predictor_32x64 ssse3/;
 230 specialize qw/aom_smooth_h_predictor_64x64 ssse3/;
 231 specialize qw/aom_smooth_h_predictor_64x32 ssse3/;
 232 specialize qw/aom_smooth_h_predictor_64x16 ssse3/;
 233
 234 # TODO(yunqingwang): optimize rectangular DC_PRED to replace division
 235 # by multiply and shift.
 236 specialize qw/aom_dc_predictor_4x4 dspr2 msa neon sse2/;
 237 specialize qw/aom_dc_predictor_4x8 sse2/;
 238 specialize qw/aom_dc_predictor_8x4 sse2/;
 239 specialize qw/aom_dc_predictor_8x8 dspr2 neon msa sse2/;
 240 specialize qw/aom_dc_predictor_8x16 sse2/;
 241 specialize qw/aom_dc_predictor_16x8 sse2/;
 242 specialize qw/aom_dc_predictor_16x16 dspr2 neon msa sse2/;
 243 specialize qw/aom_dc_predictor_16x32 sse2/;
 244 specialize qw/aom_dc_predictor_16x64 sse2/;
 245 specialize qw/aom_dc_predictor_32x16 sse2 avx2/;
 246 specialize qw/aom_dc_predictor_32x32 msa neon sse2 avx2/;
 247 specialize qw/aom_dc_predictor_32x64 sse2 avx2/;
 248 specialize qw/aom_dc_predictor_64x64 sse2 avx2/;
 249 specialize qw/aom_dc_predictor_64x32 sse2 avx2/;
 250 specialize qw/aom_dc_predictor_64x16 sse2 avx2/;
 251
 252   specialize qw/aom_highbd_v_predictor_4x4 sse2/;
 253   specialize qw/aom_highbd_v_predictor_4x8 sse2/;
 254   specialize qw/aom_highbd_v_predictor_8x4 sse2/;
 255   specialize qw/aom_highbd_v_predictor_8x8 sse2/;
 256   specialize qw/aom_highbd_v_predictor_8x16 sse2/;
 257   specialize qw/aom_highbd_v_predictor_16x8 sse2/;
 258   specialize qw/aom_highbd_v_predictor_16x16 sse2/;
 259   specialize qw/aom_highbd_v_predictor_16x32 sse2/;
 260   specialize qw/aom_highbd_v_predictor_32x16 sse2/;
 261   specialize qw/aom_highbd_v_predictor_32x32 sse2/;
 262
 263   # TODO(yunqingwang): optimize rectangular DC_PRED to replace division
 264   # by multiply and shift.
 265   specialize qw/aom_highbd_dc_predictor_4x4 sse2/;
 266   specialize qw/aom_highbd_dc_predictor_4x8 sse2/;
 267   specialize qw/aom_highbd_dc_predictor_8x4 sse2/;;
 268   specialize qw/aom_highbd_dc_predictor_8x8 sse2/;;
 269   specialize qw/aom_highbd_dc_predictor_8x16 sse2/;;
 270   specialize qw/aom_highbd_dc_predictor_16x8 sse2/;
 271   specialize qw/aom_highbd_dc_predictor_16x16 sse2/;
 272   specialize qw/aom_highbd_dc_predictor_16x32 sse2/;
 273   specialize qw/aom_highbd_dc_predictor_32x16 sse2/;
 274   specialize qw/aom_highbd_dc_predictor_32x32 sse2/;
 275
 276   specialize qw/aom_highbd_h_predictor_4x4 sse2/;
 277   specialize qw/aom_highbd_h_predictor_4x8 sse2/;
 278   specialize qw/aom_highbd_h_predictor_8x4 sse2/;
 279   specialize qw/aom_highbd_h_predictor_8x8 sse2/;
 280   specialize qw/aom_highbd_h_predictor_8x16 sse2/;
 281   specialize qw/aom_highbd_h_predictor_16x8 sse2/;
 282   specialize qw/aom_highbd_h_predictor_16x16 sse2/;
 283   specialize qw/aom_highbd_h_predictor_16x32 sse2/;
 284   specialize qw/aom_highbd_h_predictor_32x16 sse2/;
 285   specialize qw/aom_highbd_h_predictor_32x32 sse2/;
 286   specialize qw/aom_highbd_dc_left_predictor_4x4 sse2/;
 287   specialize qw/aom_highbd_dc_top_predictor_4x4 sse2/;
 288   specialize qw/aom_highbd_dc_128_predictor_4x4 sse2/;
 289   specialize qw/aom_highbd_dc_left_predictor_4x8 sse2/;
 290   specialize qw/aom_highbd_dc_top_predictor_4x8 sse2/;
 291   specialize qw/aom_highbd_dc_128_predictor_4x8 sse2/;
 292   specialize qw/aom_highbd_dc_left_predictor_8x4 sse2/;
 293   specialize qw/aom_highbd_dc_top_predictor_8x4 sse2/;
 294   specialize qw/aom_highbd_dc_128_predictor_8x4 sse2/;
 295   specialize qw/aom_highbd_dc_left_predictor_8x8 sse2/;
 296   specialize qw/aom_highbd_dc_top_predictor_8x8 sse2/;
 297   specialize qw/aom_highbd_dc_128_predictor_8x8 sse2/;
 298   specialize qw/aom_highbd_dc_left_predictor_8x16 sse2/;
 299   specialize qw/aom_highbd_dc_top_predictor_8x16 sse2/;
 300   specialize qw/aom_highbd_dc_128_predictor_8x16 sse2/;
 301   specialize qw/aom_highbd_dc_left_predictor_16x8 sse2/;
 302   specialize qw/aom_highbd_dc_top_predictor_16x8 sse2/;
 303   specialize qw/aom_highbd_dc_128_predictor_16x8 sse2/;
 304   specialize qw/aom_highbd_dc_left_predictor_16x16 sse2/;
 305   specialize qw/aom_highbd_dc_top_predictor_16x16 sse2/;
 306   specialize qw/aom_highbd_dc_128_predictor_16x16 sse2/;
 307   specialize qw/aom_highbd_dc_left_predictor_16x32 sse2/;
 308   specialize qw/aom_highbd_dc_top_predictor_16x32 sse2/;
 309   specialize qw/aom_highbd_dc_128_predictor_16x32 sse2/;
 310   specialize qw/aom_highbd_dc_left_predictor_32x16 sse2/;
 311   specialize qw/aom_highbd_dc_top_predictor_32x16 sse2/;
 312   specialize qw/aom_highbd_dc_128_predictor_32x16 sse2/;
 313   specialize qw/aom_highbd_dc_left_predictor_32x32 sse2/;
 314   specialize qw/aom_highbd_dc_top_predictor_32x32 sse2/;
 315   specialize qw/aom_highbd_dc_128_predictor_32x32 sse2/;
 316
 317 #
 318 # Sub Pixel Filters
 319 #
 320 add_proto qw/void aom_convolve_copy/,             "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h";
 321 add_proto qw/void aom_convolve8_horiz/,           "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h";
 322 add_proto qw/void aom_convolve8_vert/,            "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h";
 323
 324 specialize qw/aom_convolve_copy       sse2      /;
 325 specialize qw/aom_convolve8_horiz     sse2 ssse3/, "$avx2_ssse3";
 326 specialize qw/aom_convolve8_vert      sse2 ssse3/, "$avx2_ssse3";
 327
 328 # TODO(any): These need to be extended to up to 128x128 block sizes
 329 if (!(aom_config("CONFIG_AV1") eq "yes")) {
 330   specialize qw/aom_convolve_copy       neon dspr2 msa/;
 331   specialize qw/aom_convolve8_horiz     neon dspr2 msa/;
 332   specialize qw/aom_convolve8_vert      neon dspr2 msa/;
 333 }
 334
 335 add_proto qw/void aom_highbd_convolve_copy/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h, int bps";
 336 specialize qw/aom_highbd_convolve_copy sse2 avx2/;
 337
 338 add_proto qw/void aom_highbd_convolve8_horiz/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h, int bps";
 339 specialize qw/aom_highbd_convolve8_horiz avx2/, "$sse2_x86_64";
 340
 341 add_proto qw/void aom_highbd_convolve8_vert/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h, int bps";
 342 specialize qw/aom_highbd_convolve8_vert avx2/, "$sse2_x86_64";
 343
 344 #
 345 # Loopfilter
 346 #
 347 add_proto qw/void aom_lpf_vertical_14/, "uint8_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh";
 348 specialize qw/aom_lpf_vertical_14 sse2/;
 349
 350 add_proto qw/void aom_lpf_vertical_14_dual/, "uint8_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh";
 351 specialize qw/aom_lpf_vertical_14_dual sse2/;
 352
 353 add_proto qw/void aom_lpf_vertical_6/, "uint8_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh";
 354 specialize qw/aom_lpf_vertical_6 sse2/;
 355
 356 add_proto qw/void aom_lpf_vertical_8/, "uint8_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh";
 357 specialize qw/aom_lpf_vertical_8 sse2/;
 358
 359 add_proto qw/void aom_lpf_vertical_8_dual/, "uint8_t *s, int pitch, const uint8_t *blimit0, const uint8_t *limit0, const uint8_t *thresh0, const uint8_t *blimit1, const uint8_t *limit1, const uint8_t *thresh1";
 360
 361 add_proto qw/void aom_lpf_vertical_4/, "uint8_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh";
 362 specialize qw/aom_lpf_vertical_4 sse2/;
 363
 364 add_proto qw/void aom_lpf_vertical_4_dual/, "uint8_t *s, int pitch, const uint8_t *blimit0, const uint8_t *limit0, const uint8_t *thresh0, const uint8_t *blimit1, const uint8_t *limit1, const uint8_t *thresh1";
 365
 366 add_proto qw/void aom_lpf_horizontal_14/, "uint8_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh";
 367 specialize qw/aom_lpf_horizontal_14 sse2/;
 368
 369 add_proto qw/void aom_lpf_horizontal_14_dual/, "uint8_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh";
 370 specialize qw/aom_lpf_horizontal_14_dual sse2/;
 371
 372 add_proto qw/void aom_lpf_horizontal_6/, "uint8_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh";
 373 specialize qw/aom_lpf_horizontal_6 sse2/;
 374
 375 add_proto qw/void aom_lpf_horizontal_8/, "uint8_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh";
 376 specialize qw/aom_lpf_horizontal_8 sse2/;
 377
 378 add_proto qw/void aom_lpf_horizontal_8_dual/, "uint8_t *s, int pitch, const uint8_t *blimit0, const uint8_t *limit0, const uint8_t *thresh0, const uint8_t *blimit1, const uint8_t *limit1, const uint8_t *thresh1";
 379
 380 add_proto qw/void aom_lpf_horizontal_4/, "uint8_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh";
 381 specialize qw/aom_lpf_horizontal_4 sse2/;
 382
 383 add_proto qw/void aom_lpf_horizontal_4_dual/, "uint8_t *s, int pitch, const uint8_t *blimit0, const uint8_t *limit0, const uint8_t *thresh0, const uint8_t *blimit1, const uint8_t *limit1, const uint8_t *thresh1";
 384
 385 add_proto qw/void aom_highbd_lpf_vertical_14/, "uint16_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh, int bd";
 386 specialize qw/aom_highbd_lpf_vertical_14 sse2/;
 387
 388 add_proto qw/void aom_highbd_lpf_vertical_14_dual/, "uint16_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh,   int bd";
 389 specialize qw/aom_highbd_lpf_vertical_14_dual sse2 avx2/;
 390
 391 add_proto qw/void aom_highbd_lpf_vertical_8/, "uint16_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh, int bd";
 392 specialize qw/aom_highbd_lpf_vertical_8 sse2/;
 393
 394 add_proto qw/void aom_highbd_lpf_vertical_6/, "uint16_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh, int bd";
 395 specialize qw/aom_highbd_lpf_vertical_6 sse2/;
 396
 397 add_proto qw/void aom_highbd_lpf_vertical_8_dual/, "uint16_t *s, int pitch, const uint8_t *blimit0, const uint8_t *limit0, const uint8_t *thresh0, const uint8_t *blimit1, const uint8_t *limit1, const uint8_t *thresh1, int bd";
 398 specialize qw/aom_highbd_lpf_vertical_8_dual sse2 avx2/;
 399
 400 add_proto qw/void aom_highbd_lpf_vertical_4/, "uint16_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh, int bd";
 401 specialize qw/aom_highbd_lpf_vertical_4 sse2/;
 402
 403 add_proto qw/void aom_highbd_lpf_vertical_4_dual/, "uint16_t *s, int pitch, const uint8_t *blimit0, const uint8_t *limit0, const uint8_t *thresh0, const uint8_t *blimit1, const uint8_t *limit1, const uint8_t *thresh1, int bd";
 404 specialize qw/aom_highbd_lpf_vertical_4_dual sse2 avx2/;
 405
 406 add_proto qw/void aom_highbd_lpf_horizontal_14/, "uint16_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh, int bd";
 407 specialize qw/aom_highbd_lpf_horizontal_14 sse2/;
 408
 409 add_proto qw/void aom_highbd_lpf_horizontal_14_dual/, "uint16_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh, int bd";
 410 specialize qw/aom_highbd_lpf_horizontal_14_dual sse2 avx2/;
 411
 412 add_proto qw/void aom_highbd_lpf_horizontal_6/, "uint16_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh, int bd";
 413 specialize qw/aom_highbd_lpf_horizontal_6 sse2/;
 414
 415 add_proto qw/void aom_highbd_lpf_horizontal_8/, "uint16_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh, int bd";
 416 specialize qw/aom_highbd_lpf_horizontal_8 sse2/;
 417
 418 add_proto qw/void aom_highbd_lpf_horizontal_8_dual/, "uint16_t *s, int pitch, const uint8_t *blimit0, const uint8_t *limit0, const uint8_t *thresh0, const uint8_t *blimit1, const uint8_t *limit1, const uint8_t *thresh1, int bd";
 419 specialize qw/aom_highbd_lpf_horizontal_8_dual sse2 avx2/;
 420
 421 add_proto qw/void aom_highbd_lpf_horizontal_4/, "uint16_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh, int bd";
 422 specialize qw/aom_highbd_lpf_horizontal_4 sse2/;
 423
 424 add_proto qw/void aom_highbd_lpf_horizontal_4_dual/, "uint16_t *s, int pitch, const uint8_t *blimit0, const uint8_t *limit0, const uint8_t *thresh0, const uint8_t *blimit1, const uint8_t *limit1, const uint8_t *thresh1, int bd";
 425 specialize qw/aom_highbd_lpf_horizontal_4_dual sse2 avx2/;
 426
 427 # Helper functions.
 428 add_proto qw/void av1_round_shift_array/, "int32_t *arr, int size, int bit";
 429 specialize "av1_round_shift_array", qw/sse4_1/;
 430
 431 #
 432 # Encoder functions.
 433 #
 434
 435 #
 436 # Forward transform
 437 #
 438 if (aom_config("CONFIG_AV1_ENCODER") eq "yes"){
 439     add_proto qw/void aom_fdct4x4/, "const int16_t *input, tran_low_t *output, int stride";
 440     specialize qw/aom_fdct4x4 sse2/;
 441
 442     add_proto qw/void aom_fdct4x4_1/, "const int16_t *input, tran_low_t *output, int stride";
 443     specialize qw/aom_fdct4x4_1 sse2/;
 444
 445     add_proto qw/void aom_fdct8x8/, "const int16_t *input, tran_low_t *output, int stride";
 446     specialize qw/aom_fdct8x8 sse2/, "$ssse3_x86_64";
 447
 448     # High bit depth
 449     add_proto qw/void aom_highbd_fdct4x4/, "const int16_t *input, tran_low_t *output, int stride";
 450     specialize qw/aom_highbd_fdct4x4 sse2/;
 451
 452     add_proto qw/void aom_highbd_fdct8x8/, "const int16_t *input, tran_low_t *output, int stride";
 453     specialize qw/aom_highbd_fdct8x8 sse2/;
 454
 455 }  # CONFIG_AV1_ENCODER
 456
 457 #
 458 # Inverse transform
 459 if (aom_config("CONFIG_AV1") eq "yes") {
 460   add_proto qw/void aom_highbd_iwht4x4_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
 461
 462   add_proto qw/void aom_highbd_iwht4x4_16_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
 463 }  # CONFIG_AV1
 464
 465 #
 466 # Quantization
 467 #
 468 if (aom_config("CONFIG_AV1_ENCODER") eq "yes") {
 469   add_proto qw/void aom_quantize_b/, "const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan";
 470   specialize qw/aom_quantize_b sse2/, "$ssse3_x86_64", "$avx_x86_64";
 471
 472   add_proto qw/void aom_quantize_b_32x32/, "const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan";
 473   specialize qw/aom_quantize_b_32x32/, "$ssse3_x86_64", "$avx_x86_64";
 474
 475   add_proto qw/void aom_quantize_b_64x64/, "const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan";
 476 }  # CONFIG_AV1_ENCODER
 477
 478 if (aom_config("CONFIG_AV1_ENCODER") eq "yes") {
 479   add_proto qw/void aom_highbd_quantize_b/, "const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan";
 480   specialize qw/aom_highbd_quantize_b sse2 avx2/;
 481
 482   add_proto qw/void aom_highbd_quantize_b_32x32/, "const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan";
 483   specialize qw/aom_highbd_quantize_b_32x32 sse2/;
 484
 485   add_proto qw/void aom_highbd_quantize_b_64x64/, "const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan";
 486
 487 }  # CONFIG_AV1_ENCODER
 488 if (aom_config("CONFIG_AV1") eq "yes") {
 489   #
 490   # Alpha blending with mask
 491   #
 492   add_proto qw/void aom_lowbd_blend_a64_d16_mask/, "uint8_t *dst, uint32_t dst_stride, const CONV_BUF_TYPE *src0, uint32_t src0_stride, const CONV_BUF_TYPE *src1, uint32_t src1_stride, const uint8_t *mask, uint32_t mask_stride, int h, int w, int suby, int subx, ConvolveParams *conv_params";
 493   add_proto qw/void aom_highbd_blend_a64_d16_mask/, "uint8_t *dst, uint32_t dst_stride, const CONV_BUF_TYPE *src0, uint32_t src0_stride, const CONV_BUF_TYPE *src1, uint32_t src1_stride, const uint8_t *mask, uint32_t mask_stride, int h, int w, int suby, int subx, ConvolveParams *conv_params, const int bd";
 494   add_proto qw/void aom_blend_a64_mask/, "uint8_t *dst, uint32_t dst_stride, const uint8_t *src0, uint32_t src0_stride, const uint8_t *src1, uint32_t src1_stride, const uint8_t *mask, uint32_t mask_stride, int h, int w, int suby, int subx";
 495   add_proto qw/void aom_blend_a64_hmask/, "uint8_t *dst, uint32_t dst_stride, const uint8_t *src0, uint32_t src0_stride, const uint8_t *src1, uint32_t src1_stride, const uint8_t *mask, int h, int w";
 496   add_proto qw/void aom_blend_a64_vmask/, "uint8_t *dst, uint32_t dst_stride, const uint8_t *src0, uint32_t src0_stride, const uint8_t *src1, uint32_t src1_stride, const uint8_t *mask, int h, int w";
 497   specialize "aom_blend_a64_mask", qw/sse4_1/;
 498   specialize "aom_blend_a64_hmask", qw/sse4_1/;
 499   specialize "aom_blend_a64_vmask", qw/sse4_1/;
 500
 501   add_proto qw/void aom_highbd_blend_a64_mask/, "uint8_t *dst, uint32_t dst_stride, const uint8_t *src0, uint32_t src0_stride, const uint8_t *src1, uint32_t src1_stride, const uint8_t *mask, uint32_t mask_stride, int h, int w, int suby, int subx, int bd";
 502   add_proto qw/void aom_highbd_blend_a64_hmask/, "uint8_t *dst, uint32_t dst_stride, const uint8_t *src0, uint32_t src0_stride, const uint8_t *src1, uint32_t src1_stride, const uint8_t *mask, int h, int w, int bd";
 503   add_proto qw/void aom_highbd_blend_a64_vmask/, "uint8_t *dst, uint32_t dst_stride, const uint8_t *src0, uint32_t src0_stride, const uint8_t *src1, uint32_t src1_stride, const uint8_t *mask, int h, int w, int bd";
 504   specialize "aom_highbd_blend_a64_mask", qw/sse4_1/;
 505   specialize "aom_highbd_blend_a64_hmask", qw/sse4_1/;
 506   specialize "aom_highbd_blend_a64_vmask", qw/sse4_1/;
 507
 508 }  # CONFIG_AV1
 509
 510 if (aom_config("CONFIG_AV1_ENCODER") eq "yes") {
 511   #
 512   # Block subtraction
 513   #
 514   add_proto qw/void aom_subtract_block/, "int rows, int cols, int16_t *diff_ptr, ptrdiff_t diff_stride, const uint8_t *src_ptr, ptrdiff_t src_stride, const uint8_t *pred_ptr, ptrdiff_t pred_stride";
 515   specialize qw/aom_subtract_block neon msa sse2/;
 516
 517   if (aom_config("CONFIG_AV1_ENCODER") eq "yes") {
 518     #
 519     # Sum of Squares
 520     #
 521     add_proto qw/uint64_t aom_sum_squares_2d_i16/, "const int16_t *src, int stride, int width, int height";
 522     specialize qw/aom_sum_squares_2d_i16 sse2/;
 523
 524     add_proto qw/uint64_t aom_sum_squares_i16/, "const int16_t *src, uint32_t N";
 525     specialize qw/aom_sum_squares_i16 sse2/;
 526   }
 527
 528
 529   #
 530   # Avg
 531   #
 532   if (aom_config("CONFIG_AV1_ENCODER") eq "yes") {
 533     #
 534     # Avg
 535     #
 536     specialize qw/aom_avg_8x8 sse2 neon msa/;
 537       add_proto qw/void aom_highbd_subtract_block/, "int rows, int cols, int16_t *diff_ptr, ptrdiff_t diff_stride, const uint8_t *src_ptr, ptrdiff_t src_stride, const uint8_t *pred_ptr, ptrdiff_t pred_stride, int bd";
 538       specialize qw/aom_highbd_subtract_block sse2/;
 539
 540     #
 541     # Minmax
 542     #
 543     add_proto qw/void aom_minmax_8x8/, "const uint8_t *s, int p, const uint8_t *d, int dp, int *min, int *max";
 544     specialize qw/aom_minmax_8x8 sse2 neon/;
 545       add_proto qw/void aom_highbd_minmax_8x8/, "const uint8_t *s, int p, const uint8_t *d, int dp, int *min, int *max";
 546
 547     add_proto qw/void aom_hadamard_8x8/, "const int16_t *src_diff, int src_stride, int16_t *coeff";
 548     specialize qw/aom_hadamard_8x8 sse2 neon/, "$ssse3_x86_64";
 549
 550     add_proto qw/void aom_hadamard_16x16/, "const int16_t *src_diff, int src_stride, int16_t *coeff";
 551     specialize qw/aom_hadamard_16x16 sse2 neon/;
 552
 553     add_proto qw/int aom_satd/, "const int16_t *coeff, int length";
 554     specialize qw/aom_satd sse2 neon/;
 555
 556     add_proto qw/void aom_int_pro_row/, "int16_t *hbuf, const uint8_t *ref, int ref_stride, int height";
 557     specialize qw/aom_int_pro_row sse2 neon/;
 558
 559     add_proto qw/int16_t aom_int_pro_col/, "const uint8_t *ref, int width";
 560     specialize qw/aom_int_pro_col sse2 neon/;
 561
 562     add_proto qw/int aom_vector_var/, "const int16_t *ref, const int16_t *src, int bwl";
 563     specialize qw/aom_vector_var neon sse2/;
 564   }  # CONFIG_AV1_ENCODER
 565
 566   #
 567   # Single block SAD / Single block Avg SAD
 568   #
 569   foreach (@block_sizes) {
 570     ($w, $h) = @$_;
 571     add_proto qw/unsigned int/, "aom_sad${w}x${h}", "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride";
 572     add_proto qw/unsigned int/, "aom_sad${w}x${h}_avg", "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred";
 573     add_proto qw/unsigned int/, "aom_jnt_sad${w}x${h}_avg", "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred, const JNT_COMP_PARAMS *jcp_param";
 574   }
 575
 576   specialize qw/aom_sad128x128    avx2          sse2/;
 577   specialize qw/aom_sad128x64     avx2          sse2/;
 578   specialize qw/aom_sad64x128     avx2          sse2/;
 579   specialize qw/aom_sad64x64      avx2 neon msa sse2/;
 580   specialize qw/aom_sad64x32      avx2      msa sse2/;
 581   specialize qw/aom_sad32x64      avx2      msa sse2/;
 582   specialize qw/aom_sad32x32      avx2 neon msa sse2/;
 583   specialize qw/aom_sad32x16      avx2      msa sse2/;
 584   specialize qw/aom_sad16x32                msa sse2/;
 585   specialize qw/aom_sad16x16           neon msa sse2/;
 586   specialize qw/aom_sad16x8            neon msa sse2/;
 587   specialize qw/aom_sad8x16            neon msa sse2/;
 588   specialize qw/aom_sad8x8             neon msa sse2/;
 589   specialize qw/aom_sad8x4                  msa sse2/;
 590   specialize qw/aom_sad4x8                  msa sse2/;
 591   specialize qw/aom_sad4x4             neon msa sse2/;
 592
 593   specialize qw/aom_sad128x128_avg avx2     sse2/;
 594   specialize qw/aom_sad128x64_avg  avx2     sse2/;
 595   specialize qw/aom_sad64x128_avg  avx2     sse2/;
 596   specialize qw/aom_sad64x64_avg   avx2 msa sse2/;
 597   specialize qw/aom_sad64x32_avg   avx2 msa sse2/;
 598   specialize qw/aom_sad32x64_avg   avx2 msa sse2/;
 599   specialize qw/aom_sad32x32_avg   avx2 msa sse2/;
 600   specialize qw/aom_sad32x16_avg   avx2 msa sse2/;
 601   specialize qw/aom_sad16x32_avg        msa sse2/;
 602   specialize qw/aom_sad16x16_avg        msa sse2/;
 603   specialize qw/aom_sad16x8_avg         msa sse2/;
 604   specialize qw/aom_sad8x16_avg         msa sse2/;
 605   specialize qw/aom_sad8x8_avg          msa sse2/;
 606   specialize qw/aom_sad8x4_avg          msa sse2/;
 607   specialize qw/aom_sad4x8_avg          msa sse2/;
 608   specialize qw/aom_sad4x4_avg          msa sse2/;
 609
 610   specialize qw/aom_sad4x16      sse2/;
 611   specialize qw/aom_sad16x4      sse2/;
 612   specialize qw/aom_sad8x32      sse2/;
 613   specialize qw/aom_sad32x8      sse2/;
 614   specialize qw/aom_sad16x64     sse2/;
 615   specialize qw/aom_sad64x16     sse2/;
 616
 617   specialize qw/aom_sad4x16_avg  sse2/;
 618   specialize qw/aom_sad16x4_avg  sse2/;
 619   specialize qw/aom_sad8x32_avg  sse2/;
 620   specialize qw/aom_sad32x8_avg  sse2/;
 621   specialize qw/aom_sad16x64_avg sse2/;
 622   specialize qw/aom_sad64x16_avg sse2/;
 623
 624   specialize qw/aom_jnt_sad128x128_avg ssse3/;
 625   specialize qw/aom_jnt_sad128x64_avg  ssse3/;
 626   specialize qw/aom_jnt_sad64x128_avg  ssse3/;
 627   specialize qw/aom_jnt_sad64x64_avg   ssse3/;
 628   specialize qw/aom_jnt_sad64x32_avg   ssse3/;
 629   specialize qw/aom_jnt_sad32x64_avg   ssse3/;
 630   specialize qw/aom_jnt_sad32x32_avg   ssse3/;
 631   specialize qw/aom_jnt_sad32x16_avg   ssse3/;
 632   specialize qw/aom_jnt_sad16x32_avg   ssse3/;
 633   specialize qw/aom_jnt_sad16x16_avg   ssse3/;
 634   specialize qw/aom_jnt_sad16x8_avg    ssse3/;
 635   specialize qw/aom_jnt_sad8x16_avg    ssse3/;
 636   specialize qw/aom_jnt_sad8x8_avg     ssse3/;
 637   specialize qw/aom_jnt_sad8x4_avg     ssse3/;
 638   specialize qw/aom_jnt_sad4x8_avg     ssse3/;
 639   specialize qw/aom_jnt_sad4x4_avg     ssse3/;
 640
 641   specialize qw/aom_jnt_sad4x16_avg     ssse3/;
 642   specialize qw/aom_jnt_sad16x4_avg     ssse3/;
 643   specialize qw/aom_jnt_sad8x32_avg     ssse3/;
 644   specialize qw/aom_jnt_sad32x8_avg     ssse3/;
 645   specialize qw/aom_jnt_sad16x64_avg     ssse3/;
 646   specialize qw/aom_jnt_sad64x16_avg     ssse3/;
 647
 648   add_proto qw/unsigned int/, "aom_sad4xh", "const uint8_t *a, int a_stride, const uint8_t *b, int b_stride, int width, int height";
 649   add_proto qw/unsigned int/, "aom_sad8xh", "const uint8_t *a, int a_stride, const uint8_t *b, int b_stride, int width, int height";
 650   add_proto qw/unsigned int/, "aom_sad16xh", "const uint8_t *a, int a_stride, const uint8_t *b, int b_stride, int width, int height";
 651   add_proto qw/unsigned int/, "aom_sad32xh", "const uint8_t *a, int a_stride, const uint8_t *b, int b_stride, int width, int height";
 652   add_proto qw/unsigned int/, "aom_sad64xh", "const uint8_t *a, int a_stride, const uint8_t *b, int b_stride, int width, int height";
 653   add_proto qw/unsigned int/, "aom_sad128xh", "const uint8_t *a, int a_stride, const uint8_t *b, int b_stride, int width, int height";
 654
 655   specialize qw/aom_sad4xh   sse2/;
 656   specialize qw/aom_sad8xh   sse2/;
 657   specialize qw/aom_sad16xh  sse2/;
 658   specialize qw/aom_sad32xh  sse2/;
 659   specialize qw/aom_sad64xh  sse2/;
 660   specialize qw/aom_sad128xh sse2/;
 661
 662
 663     foreach (@block_sizes) {
 664       ($w, $h) = @$_;
 665       add_proto qw/unsigned int/, "aom_highbd_sad${w}x${h}", "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride";
 666       add_proto qw/unsigned int/, "aom_highbd_sad${w}x${h}_avg", "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred";
 667       if ($w != 128 && $h != 128 && $w != 4) {
 668         specialize "aom_highbd_sad${w}x${h}", qw/sse2/;
 669         specialize "aom_highbd_sad${w}x${h}_avg", qw/sse2/;
 670       }
 671       add_proto qw/unsigned int/, "aom_highbd_jnt_sad${w}x${h}_avg", "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred, const JNT_COMP_PARAMS* jcp_param";
 672     }
 673     specialize qw/aom_highbd_sad128x128 avx2/;
 674     specialize qw/aom_highbd_sad128x64  avx2/;
 675     specialize qw/aom_highbd_sad64x128  avx2/;
 676     specialize qw/aom_highbd_sad64x64   avx2 sse2/;
 677     specialize qw/aom_highbd_sad64x32   avx2 sse2/;
 678     specialize qw/aom_highbd_sad32x64   avx2 sse2/;
 679     specialize qw/aom_highbd_sad32x32   avx2 sse2/;
 680     specialize qw/aom_highbd_sad32x16   avx2 sse2/;
 681     specialize qw/aom_highbd_sad16x32   avx2 sse2/;
 682     specialize qw/aom_highbd_sad16x16   avx2 sse2/;
 683     specialize qw/aom_highbd_sad16x8    avx2 sse2/;
 684     specialize qw/aom_highbd_sad8x4     sse2/;
 685
 686     specialize qw/aom_highbd_sad128x128_avg avx2/;
 687     specialize qw/aom_highbd_sad128x64_avg  avx2/;
 688     specialize qw/aom_highbd_sad64x128_avg  avx2/;
 689     specialize qw/aom_highbd_sad64x64_avg   avx2 sse2/;
 690     specialize qw/aom_highbd_sad64x32_avg   avx2 sse2/;
 691     specialize qw/aom_highbd_sad32x64_avg   avx2 sse2/;
 692     specialize qw/aom_highbd_sad32x32_avg   avx2 sse2/;
 693     specialize qw/aom_highbd_sad32x16_avg   avx2 sse2/;
 694     specialize qw/aom_highbd_sad16x32_avg   avx2 sse2/;
 695     specialize qw/aom_highbd_sad16x16_avg   avx2 sse2/;
 696     specialize qw/aom_highbd_sad16x8_avg    avx2 sse2/;
 697     specialize qw/aom_highbd_sad8x4_avg     sse2/;
 698
 699     specialize qw/aom_highbd_sad16x4       sse2/;
 700     specialize qw/aom_highbd_sad8x32       sse2/;
 701     specialize qw/aom_highbd_sad32x8       sse2/;
 702     specialize qw/aom_highbd_sad16x64      sse2/;
 703     specialize qw/aom_highbd_sad64x16      sse2/;
 704
 705     specialize qw/aom_highbd_sad16x4_avg   sse2/;
 706     specialize qw/aom_highbd_sad8x32_avg   sse2/;
 707     specialize qw/aom_highbd_sad32x8_avg   sse2/;
 708     specialize qw/aom_highbd_sad16x64_avg  sse2/;
 709     specialize qw/aom_highbd_sad64x16_avg  sse2/;
 710
 711   #
 712   # Masked SAD
 713   #
 714   foreach (@block_sizes) {
 715     ($w, $h) = @$_;
 716     add_proto qw/unsigned int/, "aom_masked_sad${w}x${h}", "const uint8_t *src, int src_stride, const uint8_t *ref, int ref_stride, const uint8_t *second_pred, const uint8_t *msk, int msk_stride, int invert_mask";
 717     specialize "aom_masked_sad${w}x${h}", qw/ssse3/;
 718   }
 719
 720
 721     foreach (@block_sizes) {
 722       ($w, $h) = @$_;
 723       add_proto qw/unsigned int/, "aom_highbd_masked_sad${w}x${h}", "const uint8_t *src8, int src_stride, const uint8_t *ref8, int ref_stride, const uint8_t *second_pred8, const uint8_t *msk, int msk_stride, int invert_mask";
 724       specialize "aom_highbd_masked_sad${w}x${h}", qw/ssse3/;
 725     }
 726
 727
 728   #
 729   # OBMC SAD
 730   #
 731   foreach (@block_sizes) {
 732     ($w, $h) = @$_;
 733     add_proto qw/unsigned int/, "aom_obmc_sad${w}x${h}", "const uint8_t *pre, int pre_stride, const int32_t *wsrc, const int32_t *mask";
 734     if (! (($w == 128 && $h == 32) || ($w == 32 && $h == 128))) {
 735        specialize "aom_obmc_sad${w}x${h}", qw/sse4_1/;
 736     }
 737   }
 738
 739
 740     foreach (@block_sizes) {
 741       ($w, $h) = @$_;
 742       add_proto qw/unsigned int/, "aom_highbd_obmc_sad${w}x${h}", "const uint8_t *pre, int pre_stride, const int32_t *wsrc, const int32_t *mask";
 743       if (! (($w == 128 && $h == 32) || ($w == 32 && $h == 128))) {
 744         specialize "aom_highbd_obmc_sad${w}x${h}", qw/sse4_1/;
 745       }
 746     }
 747
 748
 749   #
 750   # Multi-block SAD, comparing a reference to N blocks 1 pixel apart horizontally
 751   #
 752   # Blocks of 3
 753   foreach $s (@block_widths) {
 754     add_proto qw/void/, "aom_sad${s}x${s}x3", "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint32_t *sad_array";
 755   }
 756   specialize qw/aom_sad64x64x3            msa/;
 757   specialize qw/aom_sad32x32x3            msa/;
 758   specialize qw/aom_sad16x16x3 sse3 ssse3 msa/;
 759   specialize qw/aom_sad8x8x3   sse3       msa/;
 760   specialize qw/aom_sad4x4x3   sse3       msa/;
 761
 762   add_proto qw/void/, "aom_sad16x8x3", "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint32_t *sad_array";
 763   specialize qw/aom_sad16x8x3 sse3 ssse3 msa/;
 764   add_proto qw/void/, "aom_sad8x16x3", "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint32_t *sad_array";
 765   specialize qw/aom_sad8x16x3 sse3 msa/;
 766
 767   # Blocks of 8
 768   foreach $s (@block_widths) {
 769     add_proto qw/void/, "aom_sad${s}x${s}x8", "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint32_t *sad_array";
 770   }
 771   specialize qw/aom_sad64x64x8        msa/;
 772   specialize qw/aom_sad32x32x8        msa/;
 773   specialize qw/aom_sad16x16x8 sse4_1 msa/;
 774   specialize qw/aom_sad8x8x8   sse4_1 msa/;
 775   specialize qw/aom_sad4x4x8   sse4_1 msa/;
 776
 777   add_proto qw/void/, "aom_sad16x8x8", "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint32_t *sad_array";
 778   specialize qw/aom_sad16x8x8 sse4_1 msa/;
 779   add_proto qw/void/, "aom_sad8x16x8", "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint32_t *sad_array";
 780   specialize qw/aom_sad8x16x8 sse4_1 msa/;
 781   add_proto qw/void/, "aom_sad8x4x8", "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint32_t *sad_array";
 782   specialize qw/aom_sad8x4x8 msa/;
 783   add_proto qw/void/, "aom_sad4x8x8", "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint32_t *sad_array";
 784   specialize qw/aom_sad4x8x8 msa/;
 785
 786
 787   foreach $s (@block_widths) {
 788     # Blocks of 3
 789     add_proto qw/void/, "aom_highbd_sad${s}x${s}x3", "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint32_t *sad_array";
 790     # Blocks of 8
 791     add_proto qw/void/, "aom_highbd_sad${s}x${s}x8", "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint32_t *sad_array";
 792   }
 793   # Blocks of 3
 794   add_proto qw/void/, "aom_highbd_sad16x8x3", "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint32_t *sad_array";
 795   add_proto qw/void/, "aom_highbd_sad8x16x3", "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint32_t *sad_array";
 796   # Blocks of 8
 797   add_proto qw/void/, "aom_highbd_sad16x8x8", "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint32_t *sad_array";
 798   add_proto qw/void/, "aom_highbd_sad8x16x8", "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint32_t *sad_array";
 799   add_proto qw/void/, "aom_highbd_sad8x4x8", "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint32_t *sad_array";
 800   add_proto qw/void/, "aom_highbd_sad4x8x8", "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint32_t *sad_array";
 801
 802
 803   #
 804   # Multi-block SAD, comparing a reference to N independent blocks
 805   #
 806   foreach (@block_sizes) {
 807     ($w, $h) = @$_;
 808     add_proto qw/void/, "aom_sad${w}x${h}x4d", "const uint8_t *src_ptr, int src_stride, const uint8_t * const ref_ptr[], int ref_stride, uint32_t *sad_array";
 809   }
 810
 811   specialize qw/aom_sad128x128x4d avx2          sse2/;
 812   specialize qw/aom_sad128x64x4d  avx2          sse2/;
 813   specialize qw/aom_sad64x128x4d  avx2          sse2/;
 814   specialize qw/aom_sad64x64x4d   avx2 neon msa sse2/;
 815   specialize qw/aom_sad64x32x4d   avx2      msa sse2/;
 816   specialize qw/aom_sad32x64x4d   avx2      msa sse2/;
 817   specialize qw/aom_sad32x32x4d   avx2 neon msa sse2/;
 818   specialize qw/aom_sad32x16x4d             msa sse2/;
 819   specialize qw/aom_sad16x32x4d             msa sse2/;
 820   specialize qw/aom_sad16x16x4d        neon msa sse2/;
 821   specialize qw/aom_sad16x8x4d              msa sse2/;
 822   specialize qw/aom_sad8x16x4d              msa sse2/;
 823   specialize qw/aom_sad8x8x4d               msa sse2/;
 824   specialize qw/aom_sad8x4x4d               msa sse2/;
 825   specialize qw/aom_sad4x8x4d               msa sse2/;
 826   specialize qw/aom_sad4x4x4d               msa sse2/;
 827
 828   specialize qw/aom_sad4x16x4d  sse2/;
 829   specialize qw/aom_sad16x4x4d  sse2/;
 830   specialize qw/aom_sad8x32x4d  sse2/;
 831   specialize qw/aom_sad32x8x4d  sse2/;
 832   specialize qw/aom_sad16x64x4d sse2/;
 833   specialize qw/aom_sad64x16x4d sse2/;
 834
 835   #
 836   # Multi-block SAD, comparing a reference to N independent blocks
 837   #
 838   foreach (@block_sizes) {
 839     ($w, $h) = @$_;
 840     add_proto qw/void/, "aom_highbd_sad${w}x${h}x4d", "const uint8_t *src_ptr, int src_stride, const uint8_t * const ref_ptr[], int ref_stride, uint32_t *sad_array";
 841     if ($w != 128 && $h != 128) {
 842       specialize "aom_highbd_sad${w}x${h}x4d", qw/sse2/;
 843     }
 844   }
 845   specialize qw/aom_highbd_sad128x128x4d avx2/;
 846   specialize qw/aom_highbd_sad128x64x4d  avx2/;
 847   specialize qw/aom_highbd_sad64x128x4d  avx2/;
 848   specialize qw/aom_highbd_sad64x64x4d   sse2 avx2/;
 849   specialize qw/aom_highbd_sad64x32x4d   sse2 avx2/;
 850   specialize qw/aom_highbd_sad32x64x4d   sse2 avx2/;
 851   specialize qw/aom_highbd_sad32x32x4d   sse2 avx2/;
 852   specialize qw/aom_highbd_sad32x16x4d   sse2 avx2/;
 853   specialize qw/aom_highbd_sad16x32x4d   sse2 avx2/;
 854   specialize qw/aom_highbd_sad16x16x4d   sse2 avx2/;
 855   specialize qw/aom_highbd_sad16x8x4d    sse2 avx2/;
 856   specialize qw/aom_highbd_sad8x16x4d    sse2/;
 857   specialize qw/aom_highbd_sad8x8x4d     sse2/;
 858   specialize qw/aom_highbd_sad8x4x4d     sse2/;
 859   specialize qw/aom_highbd_sad4x8x4d     sse2/;
 860   specialize qw/aom_highbd_sad4x4x4d     sse2/;
 861
 862   specialize qw/aom_highbd_sad4x16x4d  sse2/;
 863   specialize qw/aom_highbd_sad16x4x4d  sse2/;
 864   specialize qw/aom_highbd_sad8x32x4d  sse2/;
 865   specialize qw/aom_highbd_sad32x8x4d  sse2/;
 866   specialize qw/aom_highbd_sad16x64x4d sse2/;
 867   specialize qw/aom_highbd_sad64x16x4d sse2/;
 868
 869
 870   #
 871   # Structured Similarity (SSIM)
 872   #
 873   if (aom_config("CONFIG_INTERNAL_STATS") eq "yes") {
 874     add_proto qw/void aom_ssim_parms_8x8/, "const uint8_t *s, int sp, const uint8_t *r, int rp, uint32_t *sum_s, uint32_t *sum_r, uint32_t *sum_sq_s, uint32_t *sum_sq_r, uint32_t *sum_sxr";
 875     specialize qw/aom_ssim_parms_8x8/, "$sse2_x86_64";
 876
 877     add_proto qw/void aom_ssim_parms_16x16/, "const uint8_t *s, int sp, const uint8_t *r, int rp, uint32_t *sum_s, uint32_t *sum_r, uint32_t *sum_sq_s, uint32_t *sum_sq_r, uint32_t *sum_sxr";
 878     specialize qw/aom_ssim_parms_16x16/, "$sse2_x86_64";
 879
 880     add_proto qw/void aom_highbd_ssim_parms_8x8/, "const uint16_t *s, int sp, const uint16_t *r, int rp, uint32_t *sum_s, uint32_t *sum_r, uint32_t *sum_sq_s, uint32_t *sum_sq_r, uint32_t *sum_sxr";
 881
 882   }
 883 }  # CONFIG_AV1_ENCODER
 884
 885 if (aom_config("CONFIG_AV1_ENCODER") eq "yes") {
 886
 887   #
 888   # Specialty Variance
 889   #
 890   add_proto qw/void aom_get16x16var/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse, int *sum";
 891
 892   add_proto qw/void aom_get8x8var/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse, int *sum";
 893
 894   specialize qw/aom_get16x16var sse2 avx2 neon msa/;
 895   specialize qw/aom_get8x8var   sse2      neon msa/;
 896
 897
 898   add_proto qw/unsigned int aom_mse16x16/, "const uint8_t *src_ptr, int  source_stride, const uint8_t *ref_ptr, int  recon_stride, unsigned int *sse";
 899   add_proto qw/unsigned int aom_mse16x8/, "const uint8_t *src_ptr, int  source_stride, const uint8_t *ref_ptr, int  recon_stride, unsigned int *sse";
 900   add_proto qw/unsigned int aom_mse8x16/, "const uint8_t *src_ptr, int  source_stride, const uint8_t *ref_ptr, int  recon_stride, unsigned int *sse";
 901   add_proto qw/unsigned int aom_mse8x8/, "const uint8_t *src_ptr, int  source_stride, const uint8_t *ref_ptr, int  recon_stride, unsigned int *sse";
 902
 903   specialize qw/aom_mse16x16          sse2 avx2 neon msa/;
 904   specialize qw/aom_mse16x8           sse2           msa/;
 905   specialize qw/aom_mse8x16           sse2           msa/;
 906   specialize qw/aom_mse8x8            sse2           msa/;
 907
 908     foreach $bd (8, 10, 12) {
 909       add_proto qw/void/, "aom_highbd_${bd}_get16x16var", "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse, int *sum";
 910       add_proto qw/void/, "aom_highbd_${bd}_get8x8var", "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse, int *sum";
 911
 912       add_proto qw/unsigned int/, "aom_highbd_${bd}_mse16x16", "const uint8_t *src_ptr, int  source_stride, const uint8_t *ref_ptr, int  recon_stride, unsigned int *sse";
 913       add_proto qw/unsigned int/, "aom_highbd_${bd}_mse16x8", "const uint8_t *src_ptr, int  source_stride, const uint8_t *ref_ptr, int  recon_stride, unsigned int *sse";
 914       add_proto qw/unsigned int/, "aom_highbd_${bd}_mse8x16", "const uint8_t *src_ptr, int  source_stride, const uint8_t *ref_ptr, int  recon_stride, unsigned int *sse";
 915       add_proto qw/unsigned int/, "aom_highbd_${bd}_mse8x8", "const uint8_t *src_ptr, int  source_stride, const uint8_t *ref_ptr, int  recon_stride, unsigned int *sse";
 916
 917       specialize "aom_highbd_${bd}_mse16x16", qw/sse2/;
 918       specialize "aom_highbd_${bd}_mse8x8", qw/sse2/;
 919     }
 920
 921
 922   #
 923   #
 924   #
 925   add_proto qw/void aom_upsampled_pred/, "uint8_t *comp_pred, int width, int height, int subsample_x_q3, int subsample_y_q3, const uint8_t *ref, int ref_stride";
 926   specialize qw/aom_upsampled_pred sse2/;
 927
 928   add_proto qw/void aom_comp_avg_upsampled_pred/, "uint8_t *comp_pred, const uint8_t *pred, int width, int height, int subsample_x_q3, int subsample_y_q3, const uint8_t *ref, int ref_stride";
 929   specialize qw/aom_comp_avg_upsampled_pred sse2/;
 930
 931   add_proto qw/void aom_jnt_comp_avg_upsampled_pred/, "uint8_t *comp_pred, const uint8_t *pred, int width, int height, int subsample_x_q3, int subsample_y_q3, const uint8_t *ref, int ref_stride, const JNT_COMP_PARAMS *jcp_param";
 932   specialize qw/aom_jnt_comp_avg_upsampled_pred ssse3/;
 933
 934
 935   add_proto qw/void aom_highbd_upsampled_pred/, "uint16_t *comp_pred, int width, int height, int subsample_x_q3, int subsample_y_q3, const uint8_t *ref8, int ref_stride, int bd";
 936   specialize qw/aom_highbd_upsampled_pred sse2/;
 937   add_proto qw/void aom_highbd_comp_avg_upsampled_pred/, "uint16_t *comp_pred, const uint8_t *pred8, int width, int height, int subsample_x_q3, int subsample_y_q3, const uint8_t *ref8, int ref_stride, int bd";
 938   specialize qw/aom_highbd_comp_avg_upsampled_pred sse2/;
 939
 940   add_proto qw/void aom_highbd_jnt_comp_avg_upsampled_pred/, "uint16_t *comp_pred, const uint8_t *pred8, int width, int height, int subsample_x_q3, int subsample_y_q3, const uint8_t *ref8, int ref_stride, int bd, const JNT_COMP_PARAMS *jcp_param";
 941   specialize qw/aom_highbd_jnt_comp_avg_upsampled_pred sse2/;
 942
 943
 944   #
 945   #
 946   #
 947   add_proto qw/unsigned int aom_get_mb_ss/, "const int16_t *";
 948   add_proto qw/unsigned int aom_get4x4sse_cs/, "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int ref_stride";
 949
 950   specialize qw/aom_get_mb_ss sse2 msa/;
 951   specialize qw/aom_get4x4sse_cs neon msa/;
 952
 953   #
 954   # Variance / Subpixel Variance / Subpixel Avg Variance
 955   #
 956   add_proto qw/unsigned int/, "aom_variance2x2", "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
 957
 958   add_proto qw/unsigned int/, "aom_variance2x4", "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
 959
 960   add_proto qw/unsigned int/, "aom_variance4x2", "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
 961
 962   foreach (@block_sizes) {
 963     ($w, $h) = @$_;
 964     add_proto qw/unsigned int/, "aom_variance${w}x${h}", "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
 965     add_proto qw/uint32_t/, "aom_sub_pixel_variance${w}x${h}", "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse";
 966     add_proto qw/uint32_t/, "aom_sub_pixel_avg_variance${w}x${h}", "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred";
 967     add_proto qw/uint32_t/, "aom_jnt_sub_pixel_avg_variance${w}x${h}", "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred, const JNT_COMP_PARAMS *jcp_param";
 968   }
 969   specialize qw/aom_variance128x128   sse2 avx2         /;
 970   specialize qw/aom_variance128x64    sse2 avx2         /;
 971   specialize qw/aom_variance64x128    sse2 avx2         /;
 972   specialize qw/aom_variance64x64     sse2 avx2 neon msa/;
 973   specialize qw/aom_variance64x32     sse2 avx2 neon msa/;
 974   specialize qw/aom_variance32x64     sse2      neon msa/;
 975   specialize qw/aom_variance32x32     sse2 avx2 neon msa/;
 976   specialize qw/aom_variance32x16     sse2 avx2 msa/;
 977   specialize qw/aom_variance16x32     sse2      msa/;
 978   specialize qw/aom_variance16x16     sse2 avx2 neon msa/;
 979   specialize qw/aom_variance16x8      sse2      neon msa/;
 980   specialize qw/aom_variance8x16      sse2      neon msa/;
 981   specialize qw/aom_variance8x8       sse2      neon msa/;
 982   specialize qw/aom_variance8x4       sse2           msa/;
 983   specialize qw/aom_variance4x8       sse2           msa/;
 984   specialize qw/aom_variance4x4       sse2           msa/;
 985
 986   specialize qw/aom_sub_pixel_variance64x64     avx2 neon msa sse2 ssse3/;
 987   specialize qw/aom_sub_pixel_variance64x32               msa sse2 ssse3/;
 988   specialize qw/aom_sub_pixel_variance32x64               msa sse2 ssse3/;
 989   specialize qw/aom_sub_pixel_variance32x32     avx2 neon msa sse2 ssse3/;
 990   specialize qw/aom_sub_pixel_variance32x16               msa sse2 ssse3/;
 991   specialize qw/aom_sub_pixel_variance16x32               msa sse2 ssse3/;
 992   specialize qw/aom_sub_pixel_variance16x16          neon msa sse2 ssse3/;
 993   specialize qw/aom_sub_pixel_variance16x8                msa sse2 ssse3/;
 994   specialize qw/aom_sub_pixel_variance8x16                msa sse2 ssse3/;
 995   specialize qw/aom_sub_pixel_variance8x8            neon msa sse2 ssse3/;
 996   specialize qw/aom_sub_pixel_variance8x4                 msa sse2 ssse3/;
 997   specialize qw/aom_sub_pixel_variance4x8                 msa sse2 ssse3/;
 998   specialize qw/aom_sub_pixel_variance4x4                 msa sse2 ssse3/;
 999
1000   specialize qw/aom_sub_pixel_avg_variance64x64 avx2 msa sse2 ssse3/;
1001   specialize qw/aom_sub_pixel_avg_variance64x32      msa sse2 ssse3/;
1002   specialize qw/aom_sub_pixel_avg_variance32x64      msa sse2 ssse3/;
1003   specialize qw/aom_sub_pixel_avg_variance32x32 avx2 msa sse2 ssse3/;
1004   specialize qw/aom_sub_pixel_avg_variance32x16      msa sse2 ssse3/;
1005   specialize qw/aom_sub_pixel_avg_variance16x32      msa sse2 ssse3/;
1006   specialize qw/aom_sub_pixel_avg_variance16x16      msa sse2 ssse3/;
1007   specialize qw/aom_sub_pixel_avg_variance16x8       msa sse2 ssse3/;
1008   specialize qw/aom_sub_pixel_avg_variance8x16       msa sse2 ssse3/;
1009   specialize qw/aom_sub_pixel_avg_variance8x8        msa sse2 ssse3/;
1010   specialize qw/aom_sub_pixel_avg_variance8x4        msa sse2 ssse3/;
1011   specialize qw/aom_sub_pixel_avg_variance4x8        msa sse2 ssse3/;
1012   specialize qw/aom_sub_pixel_avg_variance4x4        msa sse2 ssse3/;
1013
1014   specialize qw/aom_variance4x16 sse2/;
1015   specialize qw/aom_variance16x4 sse2/;
1016   specialize qw/aom_variance8x32 sse2/;
1017   specialize qw/aom_variance32x8 sse2/;
1018   specialize qw/aom_variance16x64 sse2/;
1019   specialize qw/aom_variance64x16 sse2/;
1020   specialize qw/aom_sub_pixel_variance4x16 sse2 ssse3/;
1021   specialize qw/aom_sub_pixel_variance16x4 sse2 ssse3/;
1022   specialize qw/aom_sub_pixel_variance8x32 sse2 ssse3/;
1023   specialize qw/aom_sub_pixel_variance32x8 sse2 ssse3/;
1024   specialize qw/aom_sub_pixel_variance16x64 sse2 ssse3/;
1025   specialize qw/aom_sub_pixel_variance64x16 sse2 ssse3/;
1026   specialize qw/aom_sub_pixel_avg_variance4x16 sse2 ssse3/;
1027   specialize qw/aom_sub_pixel_avg_variance16x4 sse2 ssse3/;
1028   specialize qw/aom_sub_pixel_avg_variance8x32 sse2 ssse3/;
1029   specialize qw/aom_sub_pixel_avg_variance32x8 sse2 ssse3/;
1030   specialize qw/aom_sub_pixel_avg_variance16x64 sse2 ssse3/;
1031   specialize qw/aom_sub_pixel_avg_variance64x16 sse2 ssse3/;
1032
1033   specialize qw/aom_jnt_sub_pixel_avg_variance64x64 ssse3/;
1034   specialize qw/aom_jnt_sub_pixel_avg_variance64x32 ssse3/;
1035   specialize qw/aom_jnt_sub_pixel_avg_variance32x64 ssse3/;
1036   specialize qw/aom_jnt_sub_pixel_avg_variance32x32 ssse3/;
1037   specialize qw/aom_jnt_sub_pixel_avg_variance32x16 ssse3/;
1038   specialize qw/aom_jnt_sub_pixel_avg_variance16x32 ssse3/;
1039   specialize qw/aom_jnt_sub_pixel_avg_variance16x16 ssse3/;
1040   specialize qw/aom_jnt_sub_pixel_avg_variance16x8  ssse3/;
1041   specialize qw/aom_jnt_sub_pixel_avg_variance8x16  ssse3/;
1042   specialize qw/aom_jnt_sub_pixel_avg_variance8x8   ssse3/;
1043   specialize qw/aom_jnt_sub_pixel_avg_variance8x4   ssse3/;
1044   specialize qw/aom_jnt_sub_pixel_avg_variance4x8   ssse3/;
1045   specialize qw/aom_jnt_sub_pixel_avg_variance4x4   ssse3/;
1046
1047   specialize qw/aom_jnt_sub_pixel_avg_variance4x16  ssse3/;
1048   specialize qw/aom_jnt_sub_pixel_avg_variance16x4  ssse3/;
1049   specialize qw/aom_jnt_sub_pixel_avg_variance8x32  ssse3/;
1050   specialize qw/aom_jnt_sub_pixel_avg_variance32x8  ssse3/;
1051   specialize qw/aom_jnt_sub_pixel_avg_variance16x64 ssse3/;
1052   specialize qw/aom_jnt_sub_pixel_avg_variance64x16 ssse3/;
1053
1054   specialize qw/aom_jnt_sub_pixel_avg_variance128x128  ssse3/;
1055   specialize qw/aom_jnt_sub_pixel_avg_variance128x64   ssse3/;
1056   specialize qw/aom_jnt_sub_pixel_avg_variance64x128   ssse3/;
1057
1058
1059   foreach $bd (8, 10, 12) {
1060     add_proto qw/unsigned int/, "aom_highbd_${bd}_variance2x2", "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
1061
1062     add_proto qw/unsigned int/, "aom_highbd_${bd}_variance2x4", "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
1063
1064     add_proto qw/unsigned int/, "aom_highbd_${bd}_variance4x2", "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
1065
1066     foreach (@block_sizes) {
1067       ($w, $h) = @$_;
1068       add_proto qw/unsigned int/, "aom_highbd_${bd}_variance${w}x${h}", "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse";
1069       add_proto qw/uint32_t/, "aom_highbd_${bd}_sub_pixel_variance${w}x${h}", "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse";
1070       add_proto qw/uint32_t/, "aom_highbd_${bd}_sub_pixel_avg_variance${w}x${h}", "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred";
1071       if ($w != 128 && $h != 128 && $w != 4 && $h != 4) {
1072         specialize "aom_highbd_${bd}_variance${w}x${h}", "sse2";
1073       }
1074       # TODO(david.barker): When ext-partition-types is enabled, we currently
1075       # don't have vectorized 4x16 highbd variance functions
1076       if ($w == 4 && $h == 4) {
1077           specialize "aom_highbd_${bd}_variance${w}x${h}", "sse4_1";
1078         }
1079       if ($w != 128 && $h != 128 && $w != 4) {
1080         specialize "aom_highbd_${bd}_sub_pixel_variance${w}x${h}", qw/sse2/;
1081         specialize "aom_highbd_${bd}_sub_pixel_avg_variance${w}x${h}", qw/sse2/;
1082       }
1083       if ($w == 4 && $h == 4) {
1084         specialize "aom_highbd_${bd}_sub_pixel_variance${w}x${h}", "sse4_1";
1085         specialize "aom_highbd_${bd}_sub_pixel_avg_variance${w}x${h}", "sse4_1";
1086       }
1087
1088       add_proto qw/uint32_t/, "aom_highbd_${bd}_jnt_sub_pixel_avg_variance${w}x${h}", "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred, const JNT_COMP_PARAMS* jcp_param";
1089     }
1090   }
1091
1092   #
1093   # Masked Variance / Masked Subpixel Variance
1094   #
1095   foreach (@block_sizes) {
1096     ($w, $h) = @$_;
1097     add_proto qw/unsigned int/, "aom_masked_sub_pixel_variance${w}x${h}", "const uint8_t *src, int src_stride, int xoffset, int yoffset, const uint8_t *ref, int ref_stride, const uint8_t *second_pred, const uint8_t *msk, int msk_stride, int invert_mask, unsigned int *sse";
1098     specialize "aom_masked_sub_pixel_variance${w}x${h}", qw/ssse3/;
1099   }
1100
1101
1102     foreach $bd ("_8_", "_10_", "_12_") {
1103       foreach (@block_sizes) {
1104         ($w, $h) = @$_;
1105         add_proto qw/unsigned int/, "aom_highbd${bd}masked_sub_pixel_variance${w}x${h}", "const uint8_t *src, int src_stride, int xoffset, int yoffset, const uint8_t *ref, int ref_stride, const uint8_t *second_pred, const uint8_t *msk, int msk_stride, int invert_mask, unsigned int *sse";
1106         specialize "aom_highbd${bd}masked_sub_pixel_variance${w}x${h}", qw/ssse3/;
1107       }
1108     }
1109
1110
1111   #
1112   # OBMC Variance / OBMC Subpixel Variance
1113   #
1114   foreach (@block_sizes) {
1115     ($w, $h) = @$_;
1116     add_proto qw/unsigned int/, "aom_obmc_variance${w}x${h}", "const uint8_t *pre, int pre_stride, const int32_t *wsrc, const int32_t *mask, unsigned int *sse";
1117     add_proto qw/unsigned int/, "aom_obmc_sub_pixel_variance${w}x${h}", "const uint8_t *pre, int pre_stride, int xoffset, int yoffset, const int32_t *wsrc, const int32_t *mask, unsigned int *sse";
1118     specialize "aom_obmc_variance${w}x${h}", q/sse4_1/;
1119   }
1120
1121
1122     foreach $bd ("_", "_10_", "_12_") {
1123       foreach (@block_sizes) {
1124         ($w, $h) = @$_;
1125         add_proto qw/unsigned int/, "aom_highbd${bd}obmc_variance${w}x${h}", "const uint8_t *pre, int pre_stride, const int32_t *wsrc, const int32_t *mask, unsigned int *sse";
1126         add_proto qw/unsigned int/, "aom_highbd${bd}obmc_sub_pixel_variance${w}x${h}", "const uint8_t *pre, int pre_stride, int xoffset, int yoffset, const int32_t *wsrc, const int32_t *mask, unsigned int *sse";
1127         specialize "aom_highbd${bd}obmc_variance${w}x${h}", qw/sse4_1/;
1128       }
1129     }
1130
1131
1132   add_proto qw/uint32_t aom_sub_pixel_avg_variance64x64/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred";
1133   specialize qw/aom_sub_pixel_avg_variance64x64 avx2 msa sse2 ssse3/;
1134
1135   add_proto qw/uint32_t aom_sub_pixel_avg_variance64x32/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred";
1136   specialize qw/aom_sub_pixel_avg_variance64x32 msa sse2 ssse3/;
1137
1138   add_proto qw/uint32_t aom_sub_pixel_avg_variance32x64/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred";
1139   specialize qw/aom_sub_pixel_avg_variance32x64 msa sse2 ssse3/;
1140
1141   add_proto qw/uint32_t aom_sub_pixel_avg_variance32x32/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred";
1142   specialize qw/aom_sub_pixel_avg_variance32x32 avx2 msa sse2 ssse3/;
1143
1144   add_proto qw/uint32_t aom_sub_pixel_avg_variance32x16/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred";
1145   specialize qw/aom_sub_pixel_avg_variance32x16 msa sse2 ssse3/;
1146
1147   add_proto qw/uint32_t aom_sub_pixel_avg_variance16x32/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred";
1148   specialize qw/aom_sub_pixel_avg_variance16x32 msa sse2 ssse3/;
1149
1150   add_proto qw/uint32_t aom_sub_pixel_avg_variance16x16/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred";
1151   specialize qw/aom_sub_pixel_avg_variance16x16 msa sse2 ssse3/;
1152
1153   add_proto qw/uint32_t aom_sub_pixel_avg_variance16x8/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred";
1154   specialize qw/aom_sub_pixel_avg_variance16x8 msa sse2 ssse3/;
1155
1156   add_proto qw/uint32_t aom_sub_pixel_avg_variance8x16/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred";
1157   specialize qw/aom_sub_pixel_avg_variance8x16 msa sse2 ssse3/;
1158
1159   add_proto qw/uint32_t aom_sub_pixel_avg_variance8x8/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred";
1160   specialize qw/aom_sub_pixel_avg_variance8x8 msa sse2 ssse3/;
1161
1162   add_proto qw/uint32_t aom_sub_pixel_avg_variance8x4/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred";
1163   specialize qw/aom_sub_pixel_avg_variance8x4 msa sse2 ssse3/;
1164
1165   add_proto qw/uint32_t aom_sub_pixel_avg_variance4x8/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred";
1166   specialize qw/aom_sub_pixel_avg_variance4x8 msa sse2 ssse3/;
1167
1168   add_proto qw/uint32_t aom_sub_pixel_avg_variance4x4/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred";
1169   specialize qw/aom_sub_pixel_avg_variance4x4 msa sse2 ssse3/;
1170   #
1171   # Specialty Subpixel
1172   #
1173   add_proto qw/uint32_t aom_variance_halfpixvar16x16_h/, "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int  ref_stride, uint32_t *sse";
1174   specialize qw/aom_variance_halfpixvar16x16_h sse2/;
1175
1176   add_proto qw/uint32_t aom_variance_halfpixvar16x16_v/, "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int  ref_stride, uint32_t *sse";
1177   specialize qw/aom_variance_halfpixvar16x16_v sse2/;
1178
1179   add_proto qw/uint32_t aom_variance_halfpixvar16x16_hv/, "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int  ref_stride, uint32_t *sse";
1180   specialize qw/aom_variance_halfpixvar16x16_hv sse2/;
1181
1182   #
1183   # Comp Avg
1184   #
1185   add_proto qw/void aom_comp_avg_pred/, "uint8_t *comp_pred, const uint8_t *pred, int width, int height, const uint8_t *ref, int ref_stride";
1186
1187   add_proto qw/void aom_jnt_comp_avg_pred/, "uint8_t *comp_pred, const uint8_t *pred, int width, int height, const uint8_t *ref, int ref_stride, const JNT_COMP_PARAMS *jcp_param";
1188   specialize qw/aom_jnt_comp_avg_pred ssse3/;
1189
1190
1191     add_proto qw/unsigned int aom_highbd_12_variance64x64/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
1192     specialize qw/aom_highbd_12_variance64x64 sse2/;
1193
1194     add_proto qw/unsigned int aom_highbd_12_variance64x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
1195     specialize qw/aom_highbd_12_variance64x32 sse2/;
1196
1197     add_proto qw/unsigned int aom_highbd_12_variance32x64/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
1198     specialize qw/aom_highbd_12_variance32x64 sse2/;
1199
1200     add_proto qw/unsigned int aom_highbd_12_variance32x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
1201     specialize qw/aom_highbd_12_variance32x32 sse2/;
1202
1203     add_proto qw/unsigned int aom_highbd_12_variance32x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
1204     specialize qw/aom_highbd_12_variance32x16 sse2/;
1205
1206     add_proto qw/unsigned int aom_highbd_12_variance16x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
1207     specialize qw/aom_highbd_12_variance16x32 sse2/;
1208
1209     add_proto qw/unsigned int aom_highbd_12_variance16x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
1210     specialize qw/aom_highbd_12_variance16x16 sse2/;
1211
1212     add_proto qw/unsigned int aom_highbd_12_variance16x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
1213     specialize qw/aom_highbd_12_variance16x8 sse2/;
1214
1215     add_proto qw/unsigned int aom_highbd_12_variance8x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
1216     specialize qw/aom_highbd_12_variance8x16 sse2/;
1217
1218     add_proto qw/unsigned int aom_highbd_12_variance8x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
1219     specialize qw/aom_highbd_12_variance8x8 sse2/;
1220
1221     add_proto qw/unsigned int aom_highbd_12_variance8x4/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
1222     add_proto qw/unsigned int aom_highbd_12_variance4x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
1223     add_proto qw/unsigned int aom_highbd_12_variance4x4/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
1224
1225     add_proto qw/unsigned int aom_highbd_10_variance64x64/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
1226     specialize qw/aom_highbd_10_variance64x64 sse2/;
1227
1228     add_proto qw/unsigned int aom_highbd_10_variance64x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
1229     specialize qw/aom_highbd_10_variance64x32 sse2/;
1230
1231     add_proto qw/unsigned int aom_highbd_10_variance32x64/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
1232     specialize qw/aom_highbd_10_variance32x64 sse2/;
1233
1234     add_proto qw/unsigned int aom_highbd_10_variance32x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
1235     specialize qw/aom_highbd_10_variance32x32 sse2/;
1236
1237     add_proto qw/unsigned int aom_highbd_10_variance32x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
1238     specialize qw/aom_highbd_10_variance32x16 sse2/;
1239
1240     add_proto qw/unsigned int aom_highbd_10_variance16x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
1241     specialize qw/aom_highbd_10_variance16x32 sse2/;
1242
1243     add_proto qw/unsigned int aom_highbd_10_variance16x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
1244     specialize qw/aom_highbd_10_variance16x16 sse2/;
1245
1246     add_proto qw/unsigned int aom_highbd_10_variance16x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
1247     specialize qw/aom_highbd_10_variance16x8 sse2/;
1248
1249     add_proto qw/unsigned int aom_highbd_10_variance8x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
1250     specialize qw/aom_highbd_10_variance8x16 sse2/;
1251
1252     add_proto qw/unsigned int aom_highbd_10_variance8x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
1253     specialize qw/aom_highbd_10_variance8x8 sse2/;
1254
1255     add_proto qw/unsigned int aom_highbd_10_variance8x4/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
1256     add_proto qw/unsigned int aom_highbd_10_variance4x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
1257     add_proto qw/unsigned int aom_highbd_10_variance4x4/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
1258
1259     add_proto qw/unsigned int aom_highbd_8_variance64x64/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
1260     specialize qw/aom_highbd_8_variance64x64 sse2/;
1261
1262     add_proto qw/unsigned int aom_highbd_8_variance64x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
1263     specialize qw/aom_highbd_8_variance64x32 sse2/;
1264
1265     add_proto qw/unsigned int aom_highbd_8_variance32x64/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
1266     specialize qw/aom_highbd_8_variance32x64 sse2/;
1267
1268     add_proto qw/unsigned int aom_highbd_8_variance32x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
1269     specialize qw/aom_highbd_8_variance32x32 sse2/;
1270
1271     add_proto qw/unsigned int aom_highbd_8_variance32x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
1272     specialize qw/aom_highbd_8_variance32x16 sse2/;
1273
1274     add_proto qw/unsigned int aom_highbd_8_variance16x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
1275     specialize qw/aom_highbd_8_variance16x32 sse2/;
1276
1277     add_proto qw/unsigned int aom_highbd_8_variance16x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
1278     specialize qw/aom_highbd_8_variance16x16 sse2/;
1279
1280     add_proto qw/unsigned int aom_highbd_8_variance16x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
1281     specialize qw/aom_highbd_8_variance16x8 sse2/;
1282
1283     add_proto qw/unsigned int aom_highbd_8_variance8x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
1284     specialize qw/aom_highbd_8_variance8x16 sse2/;
1285
1286     add_proto qw/unsigned int aom_highbd_8_variance8x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
1287     specialize qw/aom_highbd_8_variance8x8 sse2/;
1288
1289     add_proto qw/unsigned int aom_highbd_8_variance8x4/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
1290     add_proto qw/unsigned int aom_highbd_8_variance4x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
1291     add_proto qw/unsigned int aom_highbd_8_variance4x4/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
1292
1293     add_proto qw/void aom_highbd_8_get16x16var/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse, int *sum";
1294     add_proto qw/void aom_highbd_8_get8x8var/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse, int *sum";
1295
1296     add_proto qw/void aom_highbd_10_get16x16var/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse, int *sum";
1297     add_proto qw/void aom_highbd_10_get8x8var/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse, int *sum";
1298
1299     add_proto qw/void aom_highbd_12_get16x16var/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse, int *sum";
1300     add_proto qw/void aom_highbd_12_get8x8var/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse, int *sum";
1301
1302     add_proto qw/unsigned int aom_highbd_8_mse16x16/, "const uint8_t *src_ptr, int  source_stride, const uint8_t *ref_ptr, int  recon_stride, unsigned int *sse";
1303     specialize qw/aom_highbd_8_mse16x16 sse2/;
1304
1305     add_proto qw/unsigned int aom_highbd_8_mse16x8/, "const uint8_t *src_ptr, int  source_stride, const uint8_t *ref_ptr, int  recon_stride, unsigned int *sse";
1306     add_proto qw/unsigned int aom_highbd_8_mse8x16/, "const uint8_t *src_ptr, int  source_stride, const uint8_t *ref_ptr, int  recon_stride, unsigned int *sse";
1307     add_proto qw/unsigned int aom_highbd_8_mse8x8/, "const uint8_t *src_ptr, int  source_stride, const uint8_t *ref_ptr, int  recon_stride, unsigned int *sse";
1308     specialize qw/aom_highbd_8_mse8x8 sse2/;
1309
1310     add_proto qw/unsigned int aom_highbd_10_mse16x16/, "const uint8_t *src_ptr, int  source_stride, const uint8_t *ref_ptr, int  recon_stride, unsigned int *sse";
1311     specialize qw/aom_highbd_10_mse16x16 sse2/;
1312
1313     add_proto qw/unsigned int aom_highbd_10_mse16x8/, "const uint8_t *src_ptr, int  source_stride, const uint8_t *ref_ptr, int  recon_stride, unsigned int *sse";
1314     add_proto qw/unsigned int aom_highbd_10_mse8x16/, "const uint8_t *src_ptr, int  source_stride, const uint8_t *ref_ptr, int  recon_stride, unsigned int *sse";
1315     add_proto qw/unsigned int aom_highbd_10_mse8x8/, "const uint8_t *src_ptr, int  source_stride, const uint8_t *ref_ptr, int  recon_stride, unsigned int *sse";
1316     specialize qw/aom_highbd_10_mse8x8 sse2/;
1317
1318     add_proto qw/unsigned int aom_highbd_12_mse16x16/, "const uint8_t *src_ptr, int  source_stride, const uint8_t *ref_ptr, int  recon_stride, unsigned int *sse";
1319     specialize qw/aom_highbd_12_mse16x16 sse2/;
1320
1321     add_proto qw/unsigned int aom_highbd_12_mse16x8/, "const uint8_t *src_ptr, int  source_stride, const uint8_t *ref_ptr, int  recon_stride, unsigned int *sse";
1322     add_proto qw/unsigned int aom_highbd_12_mse8x16/, "const uint8_t *src_ptr, int  source_stride, const uint8_t *ref_ptr, int  recon_stride, unsigned int *sse";
1323     add_proto qw/unsigned int aom_highbd_12_mse8x8/, "const uint8_t *src_ptr, int  source_stride, const uint8_t *ref_ptr, int  recon_stride, unsigned int *sse";
1324     specialize qw/aom_highbd_12_mse8x8 sse2/;
1325
1326     add_proto qw/void aom_highbd_comp_avg_pred/, "uint16_t *comp_pred, const uint8_t *pred8, int width, int height, const uint8_t *ref8, int ref_stride";
1327
1328     add_proto qw/void aom_highbd_jnt_comp_avg_pred/, "uint16_t *comp_pred, const uint8_t *pred8, int width, int height, const uint8_t *ref8, int ref_stride, const JNT_COMP_PARAMS *jcp_param";
1329     specialize qw/aom_highbd_jnt_comp_avg_pred c/;
1330
1331     #
1332     # Subpixel Variance
1333     #
1334     add_proto qw/uint32_t aom_highbd_12_sub_pixel_variance64x64/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse";
1335     specialize qw/aom_highbd_12_sub_pixel_variance64x64 sse2/;
1336
1337     add_proto qw/uint32_t aom_highbd_12_sub_pixel_variance64x32/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse";
1338     specialize qw/aom_highbd_12_sub_pixel_variance64x32 sse2/;
1339
1340     add_proto qw/uint32_t aom_highbd_12_sub_pixel_variance32x64/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse";
1341     specialize qw/aom_highbd_12_sub_pixel_variance32x64 sse2/;
1342
1343     add_proto qw/uint32_t aom_highbd_12_sub_pixel_variance32x32/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse";
1344     specialize qw/aom_highbd_12_sub_pixel_variance32x32 sse2/;
1345
1346     add_proto qw/uint32_t aom_highbd_12_sub_pixel_variance32x16/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse";
1347     specialize qw/aom_highbd_12_sub_pixel_variance32x16 sse2/;
1348
1349     add_proto qw/uint32_t aom_highbd_12_sub_pixel_variance16x32/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse";
1350     specialize qw/aom_highbd_12_sub_pixel_variance16x32 sse2/;
1351
1352     add_proto qw/uint32_t aom_highbd_12_sub_pixel_variance16x16/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse";
1353     specialize qw/aom_highbd_12_sub_pixel_variance16x16 sse2/;
1354
1355     add_proto qw/uint32_t aom_highbd_12_sub_pixel_variance16x8/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse";
1356     specialize qw/aom_highbd_12_sub_pixel_variance16x8 sse2/;
1357
1358     add_proto qw/uint32_t aom_highbd_12_sub_pixel_variance8x16/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse";
1359     specialize qw/aom_highbd_12_sub_pixel_variance8x16 sse2/;
1360
1361     add_proto qw/uint32_t aom_highbd_12_sub_pixel_variance8x8/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse";
1362     specialize qw/aom_highbd_12_sub_pixel_variance8x8 sse2/;
1363
1364     add_proto qw/uint32_t aom_highbd_12_sub_pixel_variance8x4/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse";
1365     specialize qw/aom_highbd_12_sub_pixel_variance8x4 sse2/;
1366
1367     add_proto qw/uint32_t aom_highbd_12_sub_pixel_variance4x8/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse";
1368     add_proto qw/uint32_t aom_highbd_12_sub_pixel_variance4x4/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse";
1369
1370     add_proto qw/uint32_t aom_highbd_10_sub_pixel_variance64x64/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse";
1371     specialize qw/aom_highbd_10_sub_pixel_variance64x64 sse2/;
1372
1373     add_proto qw/uint32_t aom_highbd_10_sub_pixel_variance64x32/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse";
1374     specialize qw/aom_highbd_10_sub_pixel_variance64x32 sse2/;
1375
1376     add_proto qw/uint32_t aom_highbd_10_sub_pixel_variance32x64/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse";
1377     specialize qw/aom_highbd_10_sub_pixel_variance32x64 sse2/;
1378
1379     add_proto qw/uint32_t aom_highbd_10_sub_pixel_variance32x32/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse";
1380     specialize qw/aom_highbd_10_sub_pixel_variance32x32 sse2/;
1381
1382     add_proto qw/uint32_t aom_highbd_10_sub_pixel_variance32x16/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse";
1383     specialize qw/aom_highbd_10_sub_pixel_variance32x16 sse2/;
1384
1385     add_proto qw/uint32_t aom_highbd_10_sub_pixel_variance16x32/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse";
1386     specialize qw/aom_highbd_10_sub_pixel_variance16x32 sse2/;
1387
1388     add_proto qw/uint32_t aom_highbd_10_sub_pixel_variance16x16/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse";
1389     specialize qw/aom_highbd_10_sub_pixel_variance16x16 sse2/;
1390
1391     add_proto qw/uint32_t aom_highbd_10_sub_pixel_variance16x8/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse";
1392     specialize qw/aom_highbd_10_sub_pixel_variance16x8 sse2/;
1393
1394     add_proto qw/uint32_t aom_highbd_10_sub_pixel_variance8x16/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse";
1395     specialize qw/aom_highbd_10_sub_pixel_variance8x16 sse2/;
1396
1397     add_proto qw/uint32_t aom_highbd_10_sub_pixel_variance8x8/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse";
1398     specialize qw/aom_highbd_10_sub_pixel_variance8x8 sse2/;
1399
1400     add_proto qw/uint32_t aom_highbd_10_sub_pixel_variance8x4/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse";
1401     specialize qw/aom_highbd_10_sub_pixel_variance8x4 sse2/;
1402
1403     add_proto qw/uint32_t aom_highbd_10_sub_pixel_variance4x8/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse";
1404     add_proto qw/uint32_t aom_highbd_10_sub_pixel_variance4x4/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse";
1405
1406     add_proto qw/uint32_t aom_highbd_8_sub_pixel_variance64x64/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse";
1407     specialize qw/aom_highbd_8_sub_pixel_variance64x64 sse2/;
1408
1409     add_proto qw/uint32_t aom_highbd_8_sub_pixel_variance64x32/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse";
1410     specialize qw/aom_highbd_8_sub_pixel_variance64x32 sse2/;
1411
1412     add_proto qw/uint32_t aom_highbd_8_sub_pixel_variance32x64/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse";
1413     specialize qw/aom_highbd_8_sub_pixel_variance32x64 sse2/;
1414
1415     add_proto qw/uint32_t aom_highbd_8_sub_pixel_variance32x32/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse";
1416     specialize qw/aom_highbd_8_sub_pixel_variance32x32 sse2/;
1417
1418     add_proto qw/uint32_t aom_highbd_8_sub_pixel_variance32x16/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse";
1419     specialize qw/aom_highbd_8_sub_pixel_variance32x16 sse2/;
1420
1421     add_proto qw/uint32_t aom_highbd_8_sub_pixel_variance16x32/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse";
1422     specialize qw/aom_highbd_8_sub_pixel_variance16x32 sse2/;
1423
1424     add_proto qw/uint32_t aom_highbd_8_sub_pixel_variance16x16/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse";
1425     specialize qw/aom_highbd_8_sub_pixel_variance16x16 sse2/;
1426
1427     add_proto qw/uint32_t aom_highbd_8_sub_pixel_variance16x8/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse";
1428     specialize qw/aom_highbd_8_sub_pixel_variance16x8 sse2/;
1429
1430     add_proto qw/uint32_t aom_highbd_8_sub_pixel_variance8x16/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse";
1431     specialize qw/aom_highbd_8_sub_pixel_variance8x16 sse2/;
1432
1433     add_proto qw/uint32_t aom_highbd_8_sub_pixel_variance8x8/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse";
1434     specialize qw/aom_highbd_8_sub_pixel_variance8x8 sse2/;
1435
1436     add_proto qw/uint32_t aom_highbd_8_sub_pixel_variance8x4/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse";
1437     specialize qw/aom_highbd_8_sub_pixel_variance8x4 sse2/;
1438
1439     add_proto qw/uint32_t aom_highbd_8_sub_pixel_variance4x8/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse";
1440     add_proto qw/uint32_t aom_highbd_8_sub_pixel_variance4x4/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse";
1441
1442     add_proto qw/uint32_t aom_highbd_12_sub_pixel_avg_variance64x64/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred";
1443     specialize qw/aom_highbd_12_sub_pixel_avg_variance64x64 sse2/;
1444
1445     add_proto qw/uint32_t aom_highbd_12_sub_pixel_avg_variance64x32/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred";
1446     specialize qw/aom_highbd_12_sub_pixel_avg_variance64x32 sse2/;
1447
1448     add_proto qw/uint32_t aom_highbd_12_sub_pixel_avg_variance32x64/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred";
1449     specialize qw/aom_highbd_12_sub_pixel_avg_variance32x64 sse2/;
1450
1451     add_proto qw/uint32_t aom_highbd_12_sub_pixel_avg_variance32x32/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred";
1452     specialize qw/aom_highbd_12_sub_pixel_avg_variance32x32 sse2/;
1453
1454     add_proto qw/uint32_t aom_highbd_12_sub_pixel_avg_variance32x16/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred";
1455     specialize qw/aom_highbd_12_sub_pixel_avg_variance32x16 sse2/;
1456
1457     add_proto qw/uint32_t aom_highbd_12_sub_pixel_avg_variance16x32/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred";
1458     specialize qw/aom_highbd_12_sub_pixel_avg_variance16x32 sse2/;
1459
1460     add_proto qw/uint32_t aom_highbd_12_sub_pixel_avg_variance16x16/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred";
1461     specialize qw/aom_highbd_12_sub_pixel_avg_variance16x16 sse2/;
1462
1463     add_proto qw/uint32_t aom_highbd_12_sub_pixel_avg_variance16x8/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred";
1464     specialize qw/aom_highbd_12_sub_pixel_avg_variance16x8 sse2/;
1465
1466     add_proto qw/uint32_t aom_highbd_12_sub_pixel_avg_variance8x16/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred";
1467     specialize qw/aom_highbd_12_sub_pixel_avg_variance8x16 sse2/;
1468
1469     add_proto qw/uint32_t aom_highbd_12_sub_pixel_avg_variance8x8/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred";
1470     specialize qw/aom_highbd_12_sub_pixel_avg_variance8x8 sse2/;
1471
1472     add_proto qw/uint32_t aom_highbd_12_sub_pixel_avg_variance8x4/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred";
1473     specialize qw/aom_highbd_12_sub_pixel_avg_variance8x4 sse2/;
1474
1475     add_proto qw/uint32_t aom_highbd_12_sub_pixel_avg_variance4x8/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred";
1476     add_proto qw/uint32_t aom_highbd_12_sub_pixel_avg_variance4x4/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred";
1477
1478     add_proto qw/uint32_t aom_highbd_10_sub_pixel_avg_variance64x64/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred";
1479     specialize qw/aom_highbd_10_sub_pixel_avg_variance64x64 sse2/;
1480
1481     add_proto qw/uint32_t aom_highbd_10_sub_pixel_avg_variance64x32/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred";
1482     specialize qw/aom_highbd_10_sub_pixel_avg_variance64x32 sse2/;
1483
1484     add_proto qw/uint32_t aom_highbd_10_sub_pixel_avg_variance32x64/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred";
1485     specialize qw/aom_highbd_10_sub_pixel_avg_variance32x64 sse2/;
1486
1487     add_proto qw/uint32_t aom_highbd_10_sub_pixel_avg_variance32x32/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred";
1488     specialize qw/aom_highbd_10_sub_pixel_avg_variance32x32 sse2/;
1489
1490     add_proto qw/uint32_t aom_highbd_10_sub_pixel_avg_variance32x16/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred";
1491     specialize qw/aom_highbd_10_sub_pixel_avg_variance32x16 sse2/;
1492
1493     add_proto qw/uint32_t aom_highbd_10_sub_pixel_avg_variance16x32/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred";
1494     specialize qw/aom_highbd_10_sub_pixel_avg_variance16x32 sse2/;
1495
1496     add_proto qw/uint32_t aom_highbd_10_sub_pixel_avg_variance16x16/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred";
1497     specialize qw/aom_highbd_10_sub_pixel_avg_variance16x16 sse2/;
1498
1499     add_proto qw/uint32_t aom_highbd_10_sub_pixel_avg_variance16x8/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred";
1500     specialize qw/aom_highbd_10_sub_pixel_avg_variance16x8 sse2/;
1501
1502     add_proto qw/uint32_t aom_highbd_10_sub_pixel_avg_variance8x16/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred";
1503     specialize qw/aom_highbd_10_sub_pixel_avg_variance8x16 sse2/;
1504
1505     add_proto qw/uint32_t aom_highbd_10_sub_pixel_avg_variance8x8/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred";
1506     specialize qw/aom_highbd_10_sub_pixel_avg_variance8x8 sse2/;
1507
1508     add_proto qw/uint32_t aom_highbd_10_sub_pixel_avg_variance8x4/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred";
1509     specialize qw/aom_highbd_10_sub_pixel_avg_variance8x4 sse2/;
1510
1511     add_proto qw/uint32_t aom_highbd_10_sub_pixel_avg_variance4x8/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred";
1512     add_proto qw/uint32_t aom_highbd_10_sub_pixel_avg_variance4x4/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred";
1513
1514     add_proto qw/uint32_t aom_highbd_8_sub_pixel_avg_variance64x64/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred";
1515     specialize qw/aom_highbd_8_sub_pixel_avg_variance64x64 sse2/;
1516
1517     add_proto qw/uint32_t aom_highbd_8_sub_pixel_avg_variance64x32/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred";
1518     specialize qw/aom_highbd_8_sub_pixel_avg_variance64x32 sse2/;
1519
1520     add_proto qw/uint32_t aom_highbd_8_sub_pixel_avg_variance32x64/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred";
1521     specialize qw/aom_highbd_8_sub_pixel_avg_variance32x64 sse2/;
1522
1523     add_proto qw/uint32_t aom_highbd_8_sub_pixel_avg_variance32x32/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred";
1524     specialize qw/aom_highbd_8_sub_pixel_avg_variance32x32 sse2/;
1525
1526     add_proto qw/uint32_t aom_highbd_8_sub_pixel_avg_variance32x16/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred";
1527     specialize qw/aom_highbd_8_sub_pixel_avg_variance32x16 sse2/;
1528
1529     add_proto qw/uint32_t aom_highbd_8_sub_pixel_avg_variance16x32/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred";
1530     specialize qw/aom_highbd_8_sub_pixel_avg_variance16x32 sse2/;
1531
1532     add_proto qw/uint32_t aom_highbd_8_sub_pixel_avg_variance16x16/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred";
1533     specialize qw/aom_highbd_8_sub_pixel_avg_variance16x16 sse2/;
1534
1535     add_proto qw/uint32_t aom_highbd_8_sub_pixel_avg_variance16x8/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred";
1536     specialize qw/aom_highbd_8_sub_pixel_avg_variance16x8 sse2/;
1537
1538     add_proto qw/uint32_t aom_highbd_8_sub_pixel_avg_variance8x16/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred";
1539     specialize qw/aom_highbd_8_sub_pixel_avg_variance8x16 sse2/;
1540
1541     add_proto qw/uint32_t aom_highbd_8_sub_pixel_avg_variance8x8/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred";
1542     specialize qw/aom_highbd_8_sub_pixel_avg_variance8x8 sse2/;
1543
1544     add_proto qw/uint32_t aom_highbd_8_sub_pixel_avg_variance8x4/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred";
1545     specialize qw/aom_highbd_8_sub_pixel_avg_variance8x4 sse2/;
1546
1547     add_proto qw/uint32_t aom_highbd_8_sub_pixel_avg_variance4x8/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred";
1548     add_proto qw/uint32_t aom_highbd_8_sub_pixel_avg_variance4x4/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, uint32_t *sse, const uint8_t *second_pred";
1549
1550
1551
1552   add_proto qw/void aom_comp_mask_pred/, "uint8_t *comp_pred, const uint8_t *pred, int width, int height, const uint8_t *ref, int ref_stride, const uint8_t *mask, int mask_stride, int invert_mask";
1553   specialize qw/aom_comp_mask_pred ssse3 avx2/;
1554
1555   add_proto qw/void aom_highbd_comp_mask_pred/, "uint16_t *comp_pred, const uint8_t *pred8, int width, int height, const uint8_t *ref8, int ref_stride, const uint8_t *mask, int mask_stride, int invert_mask";
1556   add_proto qw/void aom_highbd_comp_mask_upsampled_pred/, "uint16_t *comp_pred, const uint8_t *pred8, int width, int height, int subsample_x_q3, int subsample_y_q3, const uint8_t *ref8, int ref_stride, const uint8_t *mask, int mask_stride, int invert_mask, int bd";
1557
1558
1559 }  # CONFIG_AV1_ENCODER
1560
1561 1;