highbd_variance_sse2.c 24 KB
Newer Older
Johann's avatar
Johann committed
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52
/*
 *  Copyright (c) 2014 The WebM project authors. All Rights Reserved.
 *
 *  Use of this source code is governed by a BSD-style license
 *  that can be found in the LICENSE file in the root of the source
 *  tree. An additional intellectual property rights grant can be found
 *  in the file PATENTS.  All contributing project authors may
 *  be found in the AUTHORS file in the root of the source tree.
 */
#include "./vpx_config.h"

#include "vpx_ports/mem.h"

typedef uint32_t (*high_variance_fn_t) (const uint16_t *src, int src_stride,
                                        const uint16_t *ref, int ref_stride,
                                        uint32_t *sse, int *sum);

uint32_t vpx_highbd_calc8x8var_sse2(const uint16_t *src, int src_stride,
                                    const uint16_t *ref, int ref_stride,
                                    uint32_t *sse, int *sum);

uint32_t vpx_highbd_calc16x16var_sse2(const uint16_t *src, int src_stride,
                                      const uint16_t *ref, int ref_stride,
                                      uint32_t *sse, int *sum);

static void highbd_8_variance_sse2(const uint16_t *src, int src_stride,
                                   const uint16_t *ref, int ref_stride,
                                   int w, int h, uint32_t *sse, int *sum,
                                   high_variance_fn_t var_fn, int block_size) {
  int i, j;

  *sse = 0;
  *sum = 0;

  for (i = 0; i < h; i += block_size) {
    for (j = 0; j < w; j += block_size) {
      unsigned int sse0;
      int sum0;
      var_fn(src + src_stride * i + j, src_stride,
             ref + ref_stride * i + j, ref_stride, &sse0, &sum0);
      *sse += sse0;
      *sum += sum0;
    }
  }
}

static void highbd_10_variance_sse2(const uint16_t *src, int src_stride,
                                    const uint16_t *ref, int ref_stride,
                                    int w, int h, uint32_t *sse, int *sum,
                                    high_variance_fn_t var_fn, int block_size) {
  int i, j;
  uint64_t sse_long = 0;
Yaowu Xu's avatar
Yaowu Xu committed
53
  int32_t sum_long = 0;
Johann's avatar
Johann committed
54 55 56 57 58 59 60 61 62 63 64 65

  for (i = 0; i < h; i += block_size) {
    for (j = 0; j < w; j += block_size) {
      unsigned int sse0;
      int sum0;
      var_fn(src + src_stride * i + j, src_stride,
             ref + ref_stride * i + j, ref_stride, &sse0, &sum0);
      sse_long += sse0;
      sum_long += sum0;
    }
  }
  *sum = ROUND_POWER_OF_TWO(sum_long, 2);
Yaowu Xu's avatar
Yaowu Xu committed
66
  *sse = (uint32_t)ROUND_POWER_OF_TWO(sse_long, 4);
Johann's avatar
Johann committed
67 68 69 70 71 72 73 74
}

static void highbd_12_variance_sse2(const uint16_t *src, int src_stride,
                                    const uint16_t *ref, int ref_stride,
                                    int w, int h, uint32_t *sse, int *sum,
                                    high_variance_fn_t var_fn, int block_size) {
  int i, j;
  uint64_t sse_long = 0;
Yaowu Xu's avatar
Yaowu Xu committed
75
  int32_t sum_long = 0;
Johann's avatar
Johann committed
76 77 78 79 80 81 82 83 84 85 86 87

  for (i = 0; i < h; i += block_size) {
    for (j = 0; j < w; j += block_size) {
      unsigned int sse0;
      int sum0;
      var_fn(src + src_stride * i + j, src_stride,
             ref + ref_stride * i + j, ref_stride, &sse0, &sum0);
      sse_long += sse0;
      sum_long += sum0;
    }
  }
  *sum = ROUND_POWER_OF_TWO(sum_long, 4);
Yaowu Xu's avatar
Yaowu Xu committed
88
  *sse = (uint32_t)ROUND_POWER_OF_TWO(sse_long, 8);
Johann's avatar
Johann committed
89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243
}


#define HIGH_GET_VAR(S) \
void vpx_highbd_get##S##x##S##var_sse2(const uint8_t *src8, int src_stride, \
                                       const uint8_t *ref8, int ref_stride, \
                                       uint32_t *sse, int *sum) { \
  uint16_t *src = CONVERT_TO_SHORTPTR(src8); \
  uint16_t *ref = CONVERT_TO_SHORTPTR(ref8); \
  vpx_highbd_calc##S##x##S##var_sse2(src, src_stride, ref, ref_stride, \
                                     sse, sum); \
} \
\
void vpx_highbd_10_get##S##x##S##var_sse2(const uint8_t *src8, int src_stride, \
                                          const uint8_t *ref8, int ref_stride, \
                                          uint32_t *sse, int *sum) { \
  uint16_t *src = CONVERT_TO_SHORTPTR(src8); \
  uint16_t *ref = CONVERT_TO_SHORTPTR(ref8); \
  vpx_highbd_calc##S##x##S##var_sse2(src, src_stride, ref, ref_stride, \
                                     sse, sum); \
  *sum = ROUND_POWER_OF_TWO(*sum, 2); \
  *sse = ROUND_POWER_OF_TWO(*sse, 4); \
} \
\
void vpx_highbd_12_get##S##x##S##var_sse2(const uint8_t *src8, int src_stride, \
                                          const uint8_t *ref8, int ref_stride, \
                                          uint32_t *sse, int *sum) { \
  uint16_t *src = CONVERT_TO_SHORTPTR(src8); \
  uint16_t *ref = CONVERT_TO_SHORTPTR(ref8); \
  vpx_highbd_calc##S##x##S##var_sse2(src, src_stride, ref, ref_stride, \
                                     sse, sum); \
  *sum = ROUND_POWER_OF_TWO(*sum, 4); \
  *sse = ROUND_POWER_OF_TWO(*sse, 8); \
}

HIGH_GET_VAR(16);
HIGH_GET_VAR(8);

#undef HIGH_GET_VAR

#define VAR_FN(w, h, block_size, shift) \
uint32_t vpx_highbd_8_variance##w##x##h##_sse2( \
    const uint8_t *src8, int src_stride, \
    const uint8_t *ref8, int ref_stride, uint32_t *sse) { \
  int sum; \
  uint16_t *src = CONVERT_TO_SHORTPTR(src8); \
  uint16_t *ref = CONVERT_TO_SHORTPTR(ref8); \
  highbd_8_variance_sse2(src, src_stride, ref, ref_stride, w, h, sse, &sum, \
                         vpx_highbd_calc##block_size##x##block_size##var_sse2, \
                         block_size); \
  return *sse - (((int64_t)sum * sum) >> shift); \
} \
\
uint32_t vpx_highbd_10_variance##w##x##h##_sse2( \
    const uint8_t *src8, int src_stride, \
    const uint8_t *ref8, int ref_stride, uint32_t *sse) { \
  int sum; \
  uint16_t *src = CONVERT_TO_SHORTPTR(src8); \
  uint16_t *ref = CONVERT_TO_SHORTPTR(ref8); \
  highbd_10_variance_sse2( \
      src, src_stride, ref, ref_stride, w, h, sse, &sum, \
      vpx_highbd_calc##block_size##x##block_size##var_sse2, block_size); \
  return *sse - (((int64_t)sum * sum) >> shift); \
} \
\
uint32_t vpx_highbd_12_variance##w##x##h##_sse2( \
    const uint8_t *src8, int src_stride, \
    const uint8_t *ref8, int ref_stride, uint32_t *sse) { \
  int sum; \
  uint16_t *src = CONVERT_TO_SHORTPTR(src8); \
  uint16_t *ref = CONVERT_TO_SHORTPTR(ref8); \
  highbd_12_variance_sse2( \
      src, src_stride, ref, ref_stride, w, h, sse, &sum, \
      vpx_highbd_calc##block_size##x##block_size##var_sse2, block_size); \
  return *sse - (((int64_t)sum * sum) >> shift); \
}

VAR_FN(64, 64, 16, 12);
VAR_FN(64, 32, 16, 11);
VAR_FN(32, 64, 16, 11);
VAR_FN(32, 32, 16, 10);
VAR_FN(32, 16, 16, 9);
VAR_FN(16, 32, 16, 9);
VAR_FN(16, 16, 16, 8);
VAR_FN(16, 8, 8, 7);
VAR_FN(8, 16, 8, 7);
VAR_FN(8, 8, 8, 6);

#undef VAR_FN

unsigned int vpx_highbd_8_mse16x16_sse2(const uint8_t *src8, int src_stride,
                                      const uint8_t *ref8, int ref_stride,
                                      unsigned int *sse) {
  int sum;
  uint16_t *src = CONVERT_TO_SHORTPTR(src8);
  uint16_t *ref = CONVERT_TO_SHORTPTR(ref8);
  highbd_8_variance_sse2(src, src_stride, ref, ref_stride, 16, 16,
                         sse, &sum, vpx_highbd_calc16x16var_sse2, 16);
  return *sse;
}

unsigned int vpx_highbd_10_mse16x16_sse2(const uint8_t *src8, int src_stride,
                                         const uint8_t *ref8, int ref_stride,
                                         unsigned int *sse) {
  int sum;
  uint16_t *src = CONVERT_TO_SHORTPTR(src8);
  uint16_t *ref = CONVERT_TO_SHORTPTR(ref8);
  highbd_10_variance_sse2(src, src_stride, ref, ref_stride, 16, 16,
                          sse, &sum, vpx_highbd_calc16x16var_sse2, 16);
  return *sse;
}

unsigned int vpx_highbd_12_mse16x16_sse2(const uint8_t *src8, int src_stride,
                                         const uint8_t *ref8, int ref_stride,
                                         unsigned int *sse) {
  int sum;
  uint16_t *src = CONVERT_TO_SHORTPTR(src8);
  uint16_t *ref = CONVERT_TO_SHORTPTR(ref8);
  highbd_12_variance_sse2(src, src_stride, ref, ref_stride, 16, 16,
                          sse, &sum, vpx_highbd_calc16x16var_sse2, 16);
  return *sse;
}

unsigned int vpx_highbd_8_mse8x8_sse2(const uint8_t *src8, int src_stride,
                                    const uint8_t *ref8, int ref_stride,
                                    unsigned int *sse) {
  int sum;
  uint16_t *src = CONVERT_TO_SHORTPTR(src8);
  uint16_t *ref = CONVERT_TO_SHORTPTR(ref8);
  highbd_8_variance_sse2(src, src_stride, ref, ref_stride, 8, 8,
                         sse, &sum, vpx_highbd_calc8x8var_sse2, 8);
  return *sse;
}

unsigned int vpx_highbd_10_mse8x8_sse2(const uint8_t *src8, int src_stride,
                                       const uint8_t *ref8, int ref_stride,
                                       unsigned int *sse) {
  int sum;
  uint16_t *src = CONVERT_TO_SHORTPTR(src8);
  uint16_t *ref = CONVERT_TO_SHORTPTR(ref8);
  highbd_10_variance_sse2(src, src_stride, ref, ref_stride, 8, 8,
                          sse, &sum, vpx_highbd_calc8x8var_sse2, 8);
  return *sse;
}

unsigned int vpx_highbd_12_mse8x8_sse2(const uint8_t *src8, int src_stride,
                                       const uint8_t *ref8, int ref_stride,
                                       unsigned int *sse) {
  int sum;
  uint16_t *src = CONVERT_TO_SHORTPTR(src8);
  uint16_t *ref = CONVERT_TO_SHORTPTR(ref8);
  highbd_12_variance_sse2(src, src_stride, ref, ref_stride, 8, 8,
                          sse, &sum, vpx_highbd_calc8x8var_sse2, 8);
  return *sse;
}
Johann's avatar
Johann committed
244 245

#if CONFIG_USE_X86INC
246 247 248
// The 2 unused parameters are place holders for PIC enabled build.
// These definitions are for functions defined in
// highbd_subpel_variance_impl_sse2.asm
Johann's avatar
Johann committed
249 250 251 252 253 254
#define DECL(w, opt) \
  int vpx_highbd_sub_pixel_variance##w##xh_##opt(const uint16_t *src, \
                                                 ptrdiff_t src_stride, \
                                                 int x_offset, int y_offset, \
                                                 const uint16_t *dst, \
                                                 ptrdiff_t dst_stride, \
255 256 257
                                                 int height, \
                                                 unsigned int *sse, \
                                                 void *unused0, void *unused);
Johann's avatar
Johann committed
258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281
#define DECLS(opt1, opt2) \
  DECL(8, opt1); \
  DECL(16, opt1)

DECLS(sse2, sse);
// TODO(johannkoenig): enable the ssse3 or delete
// DECLS(ssse3, ssse3);
#undef DECLS
#undef DECL

#define FN(w, h, wf, wlog2, hlog2, opt, cast) \
uint32_t vpx_highbd_8_sub_pixel_variance##w##x##h##_##opt(const uint8_t *src8, \
                                                          int src_stride, \
                                                          int x_offset, \
                                                          int y_offset, \
                                                          const uint8_t *dst8, \
                                                          int dst_stride, \
                                                          uint32_t *sse_ptr) { \
  uint32_t sse; \
  uint16_t *src = CONVERT_TO_SHORTPTR(src8); \
  uint16_t *dst = CONVERT_TO_SHORTPTR(dst8); \
  int se = vpx_highbd_sub_pixel_variance##wf##xh_##opt(src, src_stride, \
                                                       x_offset, y_offset, \
                                                       dst, dst_stride, h, \
282
                                                       &sse, NULL, NULL); \
Johann's avatar
Johann committed
283 284 285 286 287 288 289
  if (w > wf) { \
    unsigned int sse2; \
    int se2 = vpx_highbd_sub_pixel_variance##wf##xh_##opt(src + 16, \
                                                          src_stride, \
                                                          x_offset, y_offset, \
                                                          dst + 16, \
                                                          dst_stride, \
290 291
                                                          h, &sse2, \
                                                          NULL, NULL); \
Johann's avatar
Johann committed
292 293 294 295 296 297
    se += se2; \
    sse += sse2; \
    if (w > wf * 2) { \
      se2 = vpx_highbd_sub_pixel_variance##wf##xh_##opt(src + 32, src_stride, \
                                                        x_offset, y_offset, \
                                                        dst + 32, dst_stride, \
298
                                                        h, &sse2, NULL, NULL); \
Johann's avatar
Johann committed
299 300 301 302
      se += se2; \
      sse += sse2; \
      se2 = vpx_highbd_sub_pixel_variance##wf##xh_##opt( \
          src + 48, src_stride, x_offset, y_offset, \
303
          dst + 48, dst_stride, h, &sse2, NULL, NULL); \
Johann's avatar
Johann committed
304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320
      se += se2; \
      sse += sse2; \
    } \
  } \
  *sse_ptr = sse; \
  return sse - ((cast se * se) >> (wlog2 + hlog2)); \
} \
\
uint32_t vpx_highbd_10_sub_pixel_variance##w##x##h##_##opt( \
    const uint8_t *src8, int src_stride, int x_offset, int y_offset, \
    const uint8_t *dst8, int dst_stride, uint32_t *sse_ptr) { \
  uint32_t sse; \
  uint16_t *src = CONVERT_TO_SHORTPTR(src8); \
  uint16_t *dst = CONVERT_TO_SHORTPTR(dst8); \
  int se = vpx_highbd_sub_pixel_variance##wf##xh_##opt(src, src_stride, \
                                                       x_offset, y_offset, \
                                                       dst, dst_stride, \
321
                                                       h, &sse, NULL, NULL); \
Johann's avatar
Johann committed
322 323 324 325 326 327 328
  if (w > wf) { \
    uint32_t sse2; \
    int se2 = vpx_highbd_sub_pixel_variance##wf##xh_##opt(src + 16, \
                                                          src_stride, \
                                                          x_offset, y_offset, \
                                                          dst + 16, \
                                                          dst_stride, \
329 330
                                                          h, &sse2, \
                                                          NULL, NULL); \
Johann's avatar
Johann committed
331 332 333 334 335 336
    se += se2; \
    sse += sse2; \
    if (w > wf * 2) { \
      se2 = vpx_highbd_sub_pixel_variance##wf##xh_##opt(src + 32, src_stride, \
                                                        x_offset, y_offset, \
                                                        dst + 32, dst_stride, \
337
                                                        h, &sse2, NULL, NULL); \
Johann's avatar
Johann committed
338 339 340 341 342
      se += se2; \
      sse += sse2; \
      se2 = vpx_highbd_sub_pixel_variance##wf##xh_##opt(src + 48, src_stride, \
                                                        x_offset, y_offset, \
                                                        dst + 48, dst_stride, \
343
                                                        h, &sse2, NULL, NULL); \
Johann's avatar
Johann committed
344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368
      se += se2; \
      sse += sse2; \
    } \
  } \
  se = ROUND_POWER_OF_TWO(se, 2); \
  sse = ROUND_POWER_OF_TWO(sse, 4); \
  *sse_ptr = sse; \
  return sse - ((cast se * se) >> (wlog2 + hlog2)); \
} \
\
uint32_t vpx_highbd_12_sub_pixel_variance##w##x##h##_##opt( \
    const uint8_t *src8, int src_stride, int x_offset, int y_offset, \
    const uint8_t *dst8, int dst_stride, uint32_t *sse_ptr) { \
  int start_row; \
  uint32_t sse; \
  int se = 0; \
  uint64_t long_sse = 0; \
  uint16_t *src = CONVERT_TO_SHORTPTR(src8); \
  uint16_t *dst = CONVERT_TO_SHORTPTR(dst8); \
  for (start_row = 0; start_row < h; start_row +=16) { \
    uint32_t sse2; \
    int height = h - start_row < 16 ? h - start_row : 16; \
    int se2 = vpx_highbd_sub_pixel_variance##wf##xh_##opt( \
        src + (start_row * src_stride), src_stride, \
        x_offset, y_offset, dst + (start_row * dst_stride), \
369
        dst_stride, height, &sse2, NULL, NULL); \
Johann's avatar
Johann committed
370 371 372 373 374 375
    se += se2; \
    long_sse += sse2; \
    if (w > wf) { \
      se2 = vpx_highbd_sub_pixel_variance##wf##xh_##opt( \
          src + 16 + (start_row * src_stride), src_stride, \
          x_offset, y_offset, dst + 16 + (start_row * dst_stride), \
376
          dst_stride, height, &sse2, NULL, NULL); \
Johann's avatar
Johann committed
377 378 379 380 381 382
      se += se2; \
      long_sse += sse2; \
      if (w > wf * 2) { \
        se2 = vpx_highbd_sub_pixel_variance##wf##xh_##opt( \
            src + 32 + (start_row * src_stride), src_stride, \
            x_offset, y_offset, dst + 32 + (start_row * dst_stride), \
383
            dst_stride, height, &sse2, NULL, NULL); \
Johann's avatar
Johann committed
384 385 386 387 388
        se += se2; \
        long_sse += sse2; \
        se2 = vpx_highbd_sub_pixel_variance##wf##xh_##opt( \
            src + 48 + (start_row * src_stride), src_stride, \
            x_offset, y_offset, dst + 48 + (start_row * dst_stride), \
389
            dst_stride, height, &sse2, NULL, NULL); \
Johann's avatar
Johann committed
390 391 392 393 394 395
        se += se2; \
        long_sse += sse2; \
      }\
    } \
  } \
  se = ROUND_POWER_OF_TWO(se, 4); \
Yaowu Xu's avatar
Yaowu Xu committed
396
  sse = (uint32_t)ROUND_POWER_OF_TWO(long_sse, 8); \
Johann's avatar
Johann committed
397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419
  *sse_ptr = sse; \
  return sse - ((cast se * se) >> (wlog2 + hlog2)); \
}

#define FNS(opt1, opt2) \
FN(64, 64, 16, 6, 6, opt1, (int64_t)); \
FN(64, 32, 16, 6, 5, opt1, (int64_t)); \
FN(32, 64, 16, 5, 6, opt1, (int64_t)); \
FN(32, 32, 16, 5, 5, opt1, (int64_t)); \
FN(32, 16, 16, 5, 4, opt1, (int64_t)); \
FN(16, 32, 16, 4, 5, opt1, (int64_t)); \
FN(16, 16, 16, 4, 4, opt1, (int64_t)); \
FN(16, 8, 16, 4, 3, opt1, (int64_t)); \
FN(8, 16, 8, 3, 4, opt1, (int64_t)); \
FN(8, 8, 8, 3, 3, opt1, (int64_t)); \
FN(8, 4, 8, 3, 2, opt1, (int64_t));


FNS(sse2, sse);

#undef FNS
#undef FN

420
// The 2 unused parameters are place holders for PIC enabled build.
Johann's avatar
Johann committed
421 422 423 424 425 426 427 428 429
#define DECL(w, opt) \
int vpx_highbd_sub_pixel_avg_variance##w##xh_##opt(const uint16_t *src, \
                                                   ptrdiff_t src_stride, \
                                                   int x_offset, int y_offset, \
                                                   const uint16_t *dst, \
                                                   ptrdiff_t dst_stride, \
                                                   const uint16_t *sec, \
                                                   ptrdiff_t sec_stride, \
                                                   int height, \
430 431
                                                   unsigned int *sse, \
                                                   void *unused0, void *unused);
Johann's avatar
Johann committed
432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450
#define DECLS(opt1) \
DECL(16, opt1) \
DECL(8, opt1)

DECLS(sse2);
#undef DECL
#undef DECLS

#define FN(w, h, wf, wlog2, hlog2, opt, cast) \
uint32_t vpx_highbd_8_sub_pixel_avg_variance##w##x##h##_##opt( \
    const uint8_t *src8, int src_stride, int x_offset, int y_offset, \
    const uint8_t *dst8, int dst_stride, uint32_t *sse_ptr, \
    const uint8_t *sec8) { \
  uint32_t sse; \
  uint16_t *src = CONVERT_TO_SHORTPTR(src8); \
  uint16_t *dst = CONVERT_TO_SHORTPTR(dst8); \
  uint16_t *sec = CONVERT_TO_SHORTPTR(sec8); \
  int se = vpx_highbd_sub_pixel_avg_variance##wf##xh_##opt( \
               src, src_stride, x_offset, \
451
               y_offset, dst, dst_stride, sec, w, h, &sse, NULL, NULL); \
Johann's avatar
Johann committed
452 453 454 455
  if (w > wf) { \
    uint32_t sse2; \
    int se2 = vpx_highbd_sub_pixel_avg_variance##wf##xh_##opt( \
                  src + 16, src_stride, x_offset, y_offset, \
456
                  dst + 16, dst_stride, sec + 16, w, h, &sse2, NULL, NULL); \
Johann's avatar
Johann committed
457 458 459 460 461
    se += se2; \
    sse += sse2; \
    if (w > wf * 2) { \
      se2 = vpx_highbd_sub_pixel_avg_variance##wf##xh_##opt( \
                src + 32, src_stride, x_offset, y_offset, \
462
                dst + 32, dst_stride, sec + 32, w, h, &sse2, NULL, NULL); \
Johann's avatar
Johann committed
463 464 465 466
      se += se2; \
      sse += sse2; \
      se2 = vpx_highbd_sub_pixel_avg_variance##wf##xh_##opt( \
                src + 48, src_stride, x_offset, y_offset, \
467
                dst + 48, dst_stride, sec + 48, w, h, &sse2, NULL, NULL); \
Johann's avatar
Johann committed
468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486
      se += se2; \
      sse += sse2; \
    } \
  } \
  *sse_ptr = sse; \
  return sse - ((cast se * se) >> (wlog2 + hlog2)); \
} \
\
uint32_t vpx_highbd_10_sub_pixel_avg_variance##w##x##h##_##opt( \
    const uint8_t *src8, int src_stride, int x_offset, int y_offset, \
    const uint8_t *dst8, int dst_stride, uint32_t *sse_ptr, \
    const uint8_t *sec8) { \
  uint32_t sse; \
  uint16_t *src = CONVERT_TO_SHORTPTR(src8); \
  uint16_t *dst = CONVERT_TO_SHORTPTR(dst8); \
  uint16_t *sec = CONVERT_TO_SHORTPTR(sec8); \
  int se = vpx_highbd_sub_pixel_avg_variance##wf##xh_##opt( \
                                            src, src_stride, x_offset, \
                                            y_offset, dst, dst_stride, \
487
                                            sec, w, h, &sse, NULL, NULL); \
Johann's avatar
Johann committed
488 489 490 491 492 493
  if (w > wf) { \
    uint32_t sse2; \
    int se2 = vpx_highbd_sub_pixel_avg_variance##wf##xh_##opt( \
                                            src + 16, src_stride, \
                                            x_offset, y_offset, \
                                            dst + 16, dst_stride, \
494 495
                                            sec + 16, w, h, &sse2, \
                                            NULL, NULL); \
Johann's avatar
Johann committed
496 497 498 499 500 501 502
    se += se2; \
    sse += sse2; \
    if (w > wf * 2) { \
      se2 = vpx_highbd_sub_pixel_avg_variance##wf##xh_##opt( \
                                            src + 32, src_stride, \
                                            x_offset, y_offset, \
                                            dst + 32, dst_stride, \
503 504
                                            sec + 32, w, h, &sse2, \
                                            NULL, NULL); \
Johann's avatar
Johann committed
505 506 507 508 509 510
      se += se2; \
      sse += sse2; \
      se2 = vpx_highbd_sub_pixel_avg_variance##wf##xh_##opt( \
                                            src + 48, src_stride, \
                                            x_offset, y_offset, \
                                            dst + 48, dst_stride, \
511 512
                                            sec + 48, w, h, &sse2, \
                                            NULL, NULL); \
Johann's avatar
Johann committed
513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539
      se += se2; \
      sse += sse2; \
    } \
  } \
  se = ROUND_POWER_OF_TWO(se, 2); \
  sse = ROUND_POWER_OF_TWO(sse, 4); \
  *sse_ptr = sse; \
  return sse - ((cast se * se) >> (wlog2 + hlog2)); \
} \
\
uint32_t vpx_highbd_12_sub_pixel_avg_variance##w##x##h##_##opt( \
    const uint8_t *src8, int src_stride, int x_offset, int y_offset, \
    const uint8_t *dst8, int dst_stride, uint32_t *sse_ptr, \
    const uint8_t *sec8) { \
  int start_row; \
  uint32_t sse; \
  int se = 0; \
  uint64_t long_sse = 0; \
  uint16_t *src = CONVERT_TO_SHORTPTR(src8); \
  uint16_t *dst = CONVERT_TO_SHORTPTR(dst8); \
  uint16_t *sec = CONVERT_TO_SHORTPTR(sec8); \
  for (start_row = 0; start_row < h; start_row +=16) { \
    uint32_t sse2; \
    int height = h - start_row < 16 ? h - start_row : 16; \
    int se2 = vpx_highbd_sub_pixel_avg_variance##wf##xh_##opt( \
                src + (start_row * src_stride), src_stride, x_offset, \
                y_offset, dst + (start_row * dst_stride), dst_stride, \
540
                sec + (start_row * w), w, height, &sse2, NULL, NULL); \
Johann's avatar
Johann committed
541 542 543 544 545 546 547
    se += se2; \
    long_sse += sse2; \
    if (w > wf) { \
      se2 = vpx_highbd_sub_pixel_avg_variance##wf##xh_##opt( \
                src + 16 + (start_row * src_stride), src_stride, \
                x_offset, y_offset, \
                dst + 16 + (start_row * dst_stride), dst_stride, \
548
                sec + 16 + (start_row * w), w, height, &sse2, NULL, NULL); \
Johann's avatar
Johann committed
549 550 551 552 553 554 555
      se += se2; \
      long_sse += sse2; \
      if (w > wf * 2) { \
        se2 = vpx_highbd_sub_pixel_avg_variance##wf##xh_##opt( \
                src + 32 + (start_row * src_stride), src_stride, \
                x_offset, y_offset, \
                dst + 32 + (start_row * dst_stride), dst_stride, \
556
                sec + 32 + (start_row * w), w, height, &sse2, NULL, NULL); \
Johann's avatar
Johann committed
557 558 559 560 561 562
        se += se2; \
        long_sse += sse2; \
        se2 = vpx_highbd_sub_pixel_avg_variance##wf##xh_##opt( \
                src + 48 + (start_row * src_stride), src_stride, \
                x_offset, y_offset, \
                dst + 48 + (start_row * dst_stride), dst_stride, \
563
                sec + 48 + (start_row * w), w, height, &sse2, NULL, NULL); \
Johann's avatar
Johann committed
564 565 566 567 568 569
        se += se2; \
        long_sse += sse2; \
      } \
    } \
  } \
  se = ROUND_POWER_OF_TWO(se, 4); \
Yaowu Xu's avatar
Yaowu Xu committed
570
  sse = (uint32_t)ROUND_POWER_OF_TWO(long_sse, 8); \
Johann's avatar
Johann committed
571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593
  *sse_ptr = sse; \
  return sse - ((cast se * se) >> (wlog2 + hlog2)); \
}


#define FNS(opt1) \
FN(64, 64, 16, 6, 6, opt1, (int64_t)); \
FN(64, 32, 16, 6, 5, opt1, (int64_t)); \
FN(32, 64, 16, 5, 6, opt1, (int64_t)); \
FN(32, 32, 16, 5, 5, opt1, (int64_t)); \
FN(32, 16, 16, 5, 4, opt1, (int64_t)); \
FN(16, 32, 16, 4, 5, opt1, (int64_t)); \
FN(16, 16, 16, 4, 4, opt1, (int64_t)); \
FN(16, 8, 16, 4, 3, opt1, (int64_t)); \
FN(8, 16, 8, 4, 3, opt1, (int64_t)); \
FN(8, 8, 8, 3, 3, opt1, (int64_t)); \
FN(8, 4, 8, 3, 2, opt1, (int64_t));

FNS(sse2);

#undef FNS
#undef FN
#endif  // CONFIG_USE_X86INC