convolve_test.cc 72.8 KB
Newer Older
John Koleszar's avatar
John Koleszar committed
1 2 3 4 5 6 7 8 9 10
/*
 *  Copyright (c) 2010 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.
 */

11
#include <string.h>
James Zern's avatar
James Zern committed
12
#include "test/acm_random.h"
13
#include "test/clear_system_state.h"
James Zern's avatar
James Zern committed
14 15 16
#include "test/register_state_check.h"
#include "test/util.h"
#include "third_party/googletest/src/include/gtest/gtest.h"
John Koleszar's avatar
John Koleszar committed
17 18 19

#include "./vpx_config.h"
#include "./vp9_rtcd.h"
20
#include "vp9/common/vp9_common.h"
21
#include "vp9/common/vp9_filter.h"
22
#include "vpx_mem/vpx_mem.h"
23
#include "vpx_ports/mem.h"
John Koleszar's avatar
John Koleszar committed
24 25

namespace {
26

hkuang's avatar
hkuang committed
27
static const unsigned int kMaxDimension = 64;
28

29 30 31 32 33
typedef void (*ConvolveFunc)(const uint8_t *src, ptrdiff_t src_stride,
                             uint8_t *dst, ptrdiff_t dst_stride,
                             const int16_t *filter_x, int filter_x_stride,
                             const int16_t *filter_y, int filter_y_stride,
                             int w, int h);
John Koleszar's avatar
John Koleszar committed
34 35

struct ConvolveFunctions {
36 37
  ConvolveFunctions(ConvolveFunc copy, ConvolveFunc avg,
                    ConvolveFunc h8, ConvolveFunc h8_avg,
38
                    ConvolveFunc v8, ConvolveFunc v8_avg,
39 40
                    ConvolveFunc hv8, ConvolveFunc hv8_avg,
                    int bd)
41 42
      : copy_(copy), avg_(avg), h8_(h8), v8_(v8), hv8_(hv8), h8_avg_(h8_avg),
        v8_avg_(v8_avg), hv8_avg_(hv8_avg), use_highbd_(bd) {}
John Koleszar's avatar
John Koleszar committed
43

44 45
  ConvolveFunc copy_;
  ConvolveFunc avg_;
46 47 48 49 50 51
  ConvolveFunc h8_;
  ConvolveFunc v8_;
  ConvolveFunc hv8_;
  ConvolveFunc h8_avg_;
  ConvolveFunc v8_avg_;
  ConvolveFunc hv8_avg_;
52
  int use_highbd_;  // 0 if high bitdepth not used, else the actual bit depth.
John Koleszar's avatar
John Koleszar committed
53 54
};

55
typedef std::tr1::tuple<int, int, const ConvolveFunctions *> ConvolveParam;
56

John Koleszar's avatar
John Koleszar committed
57 58 59
// Reference 8-tap subpixel filter, slightly modified to fit into this test.
#define VP9_FILTER_WEIGHT 128
#define VP9_FILTER_SHIFT 7
James Zern's avatar
James Zern committed
60
uint8_t clip_pixel(int x) {
John Koleszar's avatar
John Koleszar committed
61 62 63 64 65
  return x < 0 ? 0 :
         x > 255 ? 255 :
         x;
}

James Zern's avatar
James Zern committed
66 67 68 69 70 71 72 73
void filter_block2d_8_c(const uint8_t *src_ptr,
                        const unsigned int src_stride,
                        const int16_t *HFilter,
                        const int16_t *VFilter,
                        uint8_t *dst_ptr,
                        unsigned int dst_stride,
                        unsigned int output_width,
                        unsigned int output_height) {
John Koleszar's avatar
John Koleszar committed
74 75 76 77 78
  // Between passes, we use an intermediate buffer whose height is extended to
  // have enough horizontally filtered values as input for the vertical pass.
  // This buffer is allocated to be big enough for the largest block type we
  // support.
  const int kInterp_Extend = 4;
79 80 81 82 83 84 85 86 87 88 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
  const unsigned int intermediate_height =
      (kInterp_Extend - 1) + output_height + kInterp_Extend;
  unsigned int i, j;

  // Size of intermediate_buffer is max_intermediate_height * filter_max_width,
  // where max_intermediate_height = (kInterp_Extend - 1) + filter_max_height
  //                                 + kInterp_Extend
  //                               = 3 + 16 + 4
  //                               = 23
  // and filter_max_width          = 16
  //
  uint8_t intermediate_buffer[71 * kMaxDimension];
  const int intermediate_next_stride = 1 - intermediate_height * output_width;

  // Horizontal pass (src -> transposed intermediate).
  uint8_t *output_ptr = intermediate_buffer;
  const int src_next_row_stride = src_stride - output_width;
  src_ptr -= (kInterp_Extend - 1) * src_stride + (kInterp_Extend - 1);
  for (i = 0; i < intermediate_height; ++i) {
    for (j = 0; j < output_width; ++j) {
      // Apply filter...
      const int temp = (src_ptr[0] * HFilter[0]) +
          (src_ptr[1] * HFilter[1]) +
          (src_ptr[2] * HFilter[2]) +
          (src_ptr[3] * HFilter[3]) +
          (src_ptr[4] * HFilter[4]) +
          (src_ptr[5] * HFilter[5]) +
          (src_ptr[6] * HFilter[6]) +
          (src_ptr[7] * HFilter[7]) +
          (VP9_FILTER_WEIGHT >> 1);  // Rounding

      // Normalize back to 0-255...
      *output_ptr = clip_pixel(temp >> VP9_FILTER_SHIFT);
      ++src_ptr;
      output_ptr += intermediate_height;
    }
    src_ptr += src_next_row_stride;
    output_ptr += intermediate_next_stride;
  }

  // Vertical pass (transposed intermediate -> dst).
  src_ptr = intermediate_buffer;
  const int dst_next_row_stride = dst_stride - output_width;
  for (i = 0; i < output_height; ++i) {
    for (j = 0; j < output_width; ++j) {
      // Apply filter...
      const int temp = (src_ptr[0] * VFilter[0]) +
          (src_ptr[1] * VFilter[1]) +
          (src_ptr[2] * VFilter[2]) +
          (src_ptr[3] * VFilter[3]) +
          (src_ptr[4] * VFilter[4]) +
          (src_ptr[5] * VFilter[5]) +
          (src_ptr[6] * VFilter[6]) +
          (src_ptr[7] * VFilter[7]) +
          (VP9_FILTER_WEIGHT >> 1);  // Rounding

      // Normalize back to 0-255...
      *dst_ptr++ = clip_pixel(temp >> VP9_FILTER_SHIFT);
      src_ptr += intermediate_height;
    }
    src_ptr += intermediate_next_stride;
    dst_ptr += dst_next_row_stride;
  }
}

void block2d_average_c(uint8_t *src,
                       unsigned int src_stride,
                       uint8_t *output_ptr,
                       unsigned int output_stride,
                       unsigned int output_width,
                       unsigned int output_height) {
  unsigned int i, j;
  for (i = 0; i < output_height; ++i) {
    for (j = 0; j < output_width; ++j) {
      output_ptr[j] = (output_ptr[j] + src[i * src_stride + j] + 1) >> 1;
    }
    output_ptr += output_stride;
  }
}

void filter_average_block2d_8_c(const uint8_t *src_ptr,
                                const unsigned int src_stride,
                                const int16_t *HFilter,
                                const int16_t *VFilter,
                                uint8_t *dst_ptr,
                                unsigned int dst_stride,
                                unsigned int output_width,
                                unsigned int output_height) {
  uint8_t tmp[kMaxDimension * kMaxDimension];

  assert(output_width <= kMaxDimension);
  assert(output_height <= kMaxDimension);
  filter_block2d_8_c(src_ptr, src_stride, HFilter, VFilter, tmp, 64,
                     output_width, output_height);
  block2d_average_c(tmp, 64, dst_ptr, dst_stride,
                    output_width, output_height);
}

#if CONFIG_VP9_HIGHBITDEPTH
178 179 180 181 182 183 184 185 186
void highbd_filter_block2d_8_c(const uint16_t *src_ptr,
                               const unsigned int src_stride,
                               const int16_t *HFilter,
                               const int16_t *VFilter,
                               uint16_t *dst_ptr,
                               unsigned int dst_stride,
                               unsigned int output_width,
                               unsigned int output_height,
                               int bd) {
187 188 189 190 191
  // Between passes, we use an intermediate buffer whose height is extended to
  // have enough horizontally filtered values as input for the vertical pass.
  // This buffer is allocated to be big enough for the largest block type we
  // support.
  const int kInterp_Extend = 4;
John Koleszar's avatar
John Koleszar committed
192
  const unsigned int intermediate_height =
James Zern's avatar
James Zern committed
193
      (kInterp_Extend - 1) + output_height + kInterp_Extend;
John Koleszar's avatar
John Koleszar committed
194 195 196 197 198 199 200 201

  /* Size of intermediate_buffer is max_intermediate_height * filter_max_width,
   * where max_intermediate_height = (kInterp_Extend - 1) + filter_max_height
   *                                 + kInterp_Extend
   *                               = 3 + 16 + 4
   *                               = 23
   * and filter_max_width = 16
   */
202
  uint16_t intermediate_buffer[71 * kMaxDimension];
John Koleszar's avatar
John Koleszar committed
203 204 205 206
  const int intermediate_next_stride = 1 - intermediate_height * output_width;

  // Horizontal pass (src -> transposed intermediate).
  {
207
    uint16_t *output_ptr = intermediate_buffer;
John Koleszar's avatar
John Koleszar committed
208 209 210 211 212 213
    const int src_next_row_stride = src_stride - output_width;
    unsigned int i, j;
    src_ptr -= (kInterp_Extend - 1) * src_stride + (kInterp_Extend - 1);
    for (i = 0; i < intermediate_height; ++i) {
      for (j = 0; j < output_width; ++j) {
        // Apply filter...
James Zern's avatar
James Zern committed
214 215 216 217 218 219 220 221 222
        const int temp = (src_ptr[0] * HFilter[0]) +
                         (src_ptr[1] * HFilter[1]) +
                         (src_ptr[2] * HFilter[2]) +
                         (src_ptr[3] * HFilter[3]) +
                         (src_ptr[4] * HFilter[4]) +
                         (src_ptr[5] * HFilter[5]) +
                         (src_ptr[6] * HFilter[6]) +
                         (src_ptr[7] * HFilter[7]) +
                         (VP9_FILTER_WEIGHT >> 1);  // Rounding
John Koleszar's avatar
John Koleszar committed
223 224

        // Normalize back to 0-255...
225
        *output_ptr = clip_pixel_highbd(temp >> VP9_FILTER_SHIFT, bd);
John Koleszar's avatar
John Koleszar committed
226 227 228 229 230 231 232 233 234 235
        ++src_ptr;
        output_ptr += intermediate_height;
      }
      src_ptr += src_next_row_stride;
      output_ptr += intermediate_next_stride;
    }
  }

  // Vertical pass (transposed intermediate -> dst).
  {
236
    uint16_t *src_ptr = intermediate_buffer;
John Koleszar's avatar
John Koleszar committed
237 238 239 240 241
    const int dst_next_row_stride = dst_stride - output_width;
    unsigned int i, j;
    for (i = 0; i < output_height; ++i) {
      for (j = 0; j < output_width; ++j) {
        // Apply filter...
James Zern's avatar
James Zern committed
242 243 244 245 246 247 248 249 250
        const int temp = (src_ptr[0] * VFilter[0]) +
                         (src_ptr[1] * VFilter[1]) +
                         (src_ptr[2] * VFilter[2]) +
                         (src_ptr[3] * VFilter[3]) +
                         (src_ptr[4] * VFilter[4]) +
                         (src_ptr[5] * VFilter[5]) +
                         (src_ptr[6] * VFilter[6]) +
                         (src_ptr[7] * VFilter[7]) +
                         (VP9_FILTER_WEIGHT >> 1);  // Rounding
John Koleszar's avatar
John Koleszar committed
251 252

        // Normalize back to 0-255...
253
        *dst_ptr++ = clip_pixel_highbd(temp >> VP9_FILTER_SHIFT, bd);
John Koleszar's avatar
John Koleszar committed
254 255 256 257 258 259 260 261
        src_ptr += intermediate_height;
      }
      src_ptr += intermediate_next_stride;
      dst_ptr += dst_next_row_stride;
    }
  }
}

262 263 264 265 266 267 268
void highbd_block2d_average_c(uint16_t *src,
                              unsigned int src_stride,
                              uint16_t *output_ptr,
                              unsigned int output_stride,
                              unsigned int output_width,
                              unsigned int output_height,
                              int bd) {
John Koleszar's avatar
John Koleszar committed
269 270 271 272 273 274 275 276 277
  unsigned int i, j;
  for (i = 0; i < output_height; ++i) {
    for (j = 0; j < output_width; ++j) {
      output_ptr[j] = (output_ptr[j] + src[i * src_stride + j] + 1) >> 1;
    }
    output_ptr += output_stride;
  }
}

278 279 280 281 282 283 284 285 286
void highbd_filter_average_block2d_8_c(const uint16_t *src_ptr,
                                       const unsigned int src_stride,
                                       const int16_t *HFilter,
                                       const int16_t *VFilter,
                                       uint16_t *dst_ptr,
                                       unsigned int dst_stride,
                                       unsigned int output_width,
                                       unsigned int output_height,
                                       int bd) {
287 288 289 290
  uint16_t tmp[kMaxDimension * kMaxDimension];

  assert(output_width <= kMaxDimension);
  assert(output_height <= kMaxDimension);
291 292 293 294
  highbd_filter_block2d_8_c(src_ptr, src_stride, HFilter, VFilter, tmp, 64,
                            output_width, output_height, bd);
  highbd_block2d_average_c(tmp, 64, dst_ptr, dst_stride,
                           output_width, output_height, bd);
John Koleszar's avatar
John Koleszar committed
295
}
296
#endif  // CONFIG_VP9_HIGHBITDEPTH
John Koleszar's avatar
John Koleszar committed
297

298
class ConvolveTest : public ::testing::TestWithParam<ConvolveParam> {
299 300 301 302
 public:
  static void SetUpTestCase() {
    // Force input_ to be unaligned, output to be 16 byte aligned.
    input_ = reinterpret_cast<uint8_t*>(
303
        vpx_memalign(kDataAlignment, kInputBufferSize + 1)) + 1;
304
    output_ = reinterpret_cast<uint8_t*>(
305
        vpx_memalign(kDataAlignment, kOutputBufferSize));
306 307
    output_ref_ = reinterpret_cast<uint8_t*>(
        vpx_memalign(kDataAlignment, kOutputBufferSize));
308 309 310 311 312 313
#if CONFIG_VP9_HIGHBITDEPTH
    input16_ = reinterpret_cast<uint16_t*>(
        vpx_memalign(kDataAlignment,
                     (kInputBufferSize + 1) * sizeof(uint16_t))) + 1;
    output16_ = reinterpret_cast<uint16_t*>(
        vpx_memalign(kDataAlignment, (kOutputBufferSize) * sizeof(uint16_t)));
314 315
    output16_ref_ = reinterpret_cast<uint16_t*>(
        vpx_memalign(kDataAlignment, (kOutputBufferSize) * sizeof(uint16_t)));
316
#endif
317 318
  }

319 320
  virtual void TearDown() { libvpx_test::ClearSystemState(); }

321 322 323 324 325
  static void TearDownTestCase() {
    vpx_free(input_ - 1);
    input_ = NULL;
    vpx_free(output_);
    output_ = NULL;
326 327
    vpx_free(output_ref_);
    output_ref_ = NULL;
328 329 330 331 332
#if CONFIG_VP9_HIGHBITDEPTH
    vpx_free(input16_ - 1);
    input16_ = NULL;
    vpx_free(output16_);
    output16_ = NULL;
333 334
    vpx_free(output16_ref_);
    output16_ref_ = NULL;
335
#endif
336 337
  }

James Zern's avatar
James Zern committed
338 339
 protected:
  static const int kDataAlignment = 16;
340
  static const int kOuterBlockSize = 256;
James Zern's avatar
James Zern committed
341 342 343 344 345 346 347 348 349 350 351 352
  static const int kInputStride = kOuterBlockSize;
  static const int kOutputStride = kOuterBlockSize;
  static const int kInputBufferSize = kOuterBlockSize * kOuterBlockSize;
  static const int kOutputBufferSize = kOuterBlockSize * kOuterBlockSize;

  int Width() const { return GET_PARAM(0); }
  int Height() const { return GET_PARAM(1); }
  int BorderLeft() const {
    const int center = (kOuterBlockSize - Width()) / 2;
    return (center + (kDataAlignment - 1)) & ~(kDataAlignment - 1);
  }
  int BorderTop() const { return (kOuterBlockSize - Height()) / 2; }
John Koleszar's avatar
John Koleszar committed
353

James Zern's avatar
James Zern committed
354 355 356 357 358 359
  bool IsIndexInBorder(int i) {
    return (i < BorderTop() * kOuterBlockSize ||
            i >= (BorderTop() + Height()) * kOuterBlockSize ||
            i % kOuterBlockSize < BorderLeft() ||
            i % kOuterBlockSize >= (BorderLeft() + Width()));
  }
John Koleszar's avatar
John Koleszar committed
360

James Zern's avatar
James Zern committed
361 362
  virtual void SetUp() {
    UUT_ = GET_PARAM(2);
363
#if CONFIG_VP9_HIGHBITDEPTH
364 365
    if (UUT_->use_highbd_ != 0)
      mask_ = (1 << UUT_->use_highbd_) - 1;
366 367 368
    else
      mask_ = 255;
#endif
Johann's avatar
Johann committed
369
    /* Set up guard blocks for an inner block centered in the outer block */
James Zern's avatar
James Zern committed
370 371 372 373 374
    for (int i = 0; i < kOutputBufferSize; ++i) {
      if (IsIndexInBorder(i))
        output_[i] = 255;
      else
        output_[i] = 0;
John Koleszar's avatar
John Koleszar committed
375 376
    }

James Zern's avatar
James Zern committed
377
    ::libvpx_test::ACMRandom prng;
378
    for (int i = 0; i < kInputBufferSize; ++i) {
379
      if (i & 1) {
380
        input_[i] = 255;
381 382 383 384
#if CONFIG_VP9_HIGHBITDEPTH
        input16_[i] = mask_;
#endif
      } else {
385
        input_[i] = prng.Rand8Extremes();
386 387 388 389
#if CONFIG_VP9_HIGHBITDEPTH
        input16_[i] = prng.Rand16() & mask_;
#endif
      }
390
    }
James Zern's avatar
James Zern committed
391
  }
John Koleszar's avatar
John Koleszar committed
392

393 394
  void SetConstantInput(int value) {
    memset(input_, value, kInputBufferSize);
395 396 397
#if CONFIG_VP9_HIGHBITDEPTH
    vpx_memset16(input16_, value, kInputBufferSize);
#endif
398 399
  }

400
  void CopyOutputToRef() {
James Zern's avatar
James Zern committed
401
    memcpy(output_ref_, output_, kOutputBufferSize);
402
#if CONFIG_VP9_HIGHBITDEPTH
James Zern's avatar
James Zern committed
403
    memcpy(output16_ref_, output16_, kOutputBufferSize);
404 405 406
#endif
  }

James Zern's avatar
James Zern committed
407 408 409 410
  void CheckGuardBlocks() {
    for (int i = 0; i < kOutputBufferSize; ++i) {
      if (IsIndexInBorder(i))
        EXPECT_EQ(255, output_[i]);
John Koleszar's avatar
John Koleszar committed
411
    }
James Zern's avatar
James Zern committed
412
  }
John Koleszar's avatar
John Koleszar committed
413

414 415
  uint8_t *input() const {
#if CONFIG_VP9_HIGHBITDEPTH
416
    if (UUT_->use_highbd_ == 0) {
417 418 419 420 421 422
      return input_ + BorderTop() * kOuterBlockSize + BorderLeft();
    } else {
      return CONVERT_TO_BYTEPTR(input16_ + BorderTop() * kOuterBlockSize +
                                BorderLeft());
    }
#else
James Zern's avatar
James Zern committed
423
    return input_ + BorderTop() * kOuterBlockSize + BorderLeft();
424
#endif
James Zern's avatar
James Zern committed
425
  }
John Koleszar's avatar
John Koleszar committed
426

427 428
  uint8_t *output() const {
#if CONFIG_VP9_HIGHBITDEPTH
429
    if (UUT_->use_highbd_ == 0) {
430 431 432 433 434 435
      return output_ + BorderTop() * kOuterBlockSize + BorderLeft();
    } else {
      return CONVERT_TO_BYTEPTR(output16_ + BorderTop() * kOuterBlockSize +
                                BorderLeft());
    }
#else
James Zern's avatar
James Zern committed
436
    return output_ + BorderTop() * kOuterBlockSize + BorderLeft();
437 438 439
#endif
  }

440 441 442 443 444 445 446 447 448 449 450 451 452
  uint8_t *output_ref() const {
#if CONFIG_VP9_HIGHBITDEPTH
    if (UUT_->use_highbd_ == 0) {
      return output_ref_ + BorderTop() * kOuterBlockSize + BorderLeft();
    } else {
      return CONVERT_TO_BYTEPTR(output16_ref_ + BorderTop() * kOuterBlockSize +
                                BorderLeft());
    }
#else
    return output_ref_ + BorderTop() * kOuterBlockSize + BorderLeft();
#endif
  }

453 454
  uint16_t lookup(uint8_t *list, int index) const {
#if CONFIG_VP9_HIGHBITDEPTH
455
    if (UUT_->use_highbd_ == 0) {
456 457 458 459 460 461 462 463 464 465 466
      return list[index];
    } else {
      return CONVERT_TO_SHORTPTR(list)[index];
    }
#else
    return list[index];
#endif
  }

  void assign_val(uint8_t *list, int index, uint16_t val) const {
#if CONFIG_VP9_HIGHBITDEPTH
467
    if (UUT_->use_highbd_ == 0) {
468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485
      list[index] = (uint8_t) val;
    } else {
      CONVERT_TO_SHORTPTR(list)[index] = val;
    }
#else
    list[index] = (uint8_t) val;
#endif
  }

  void wrapper_filter_average_block2d_8_c(const uint8_t *src_ptr,
                                          const unsigned int src_stride,
                                          const int16_t *HFilter,
                                          const int16_t *VFilter,
                                          uint8_t *dst_ptr,
                                          unsigned int dst_stride,
                                          unsigned int output_width,
                                          unsigned int output_height) {
#if CONFIG_VP9_HIGHBITDEPTH
486
    if (UUT_->use_highbd_ == 0) {
487 488 489 490
      filter_average_block2d_8_c(src_ptr, src_stride, HFilter, VFilter,
                                 dst_ptr, dst_stride, output_width,
                                 output_height);
    } else {
491 492 493 494 495
      highbd_filter_average_block2d_8_c(CONVERT_TO_SHORTPTR(src_ptr),
                                        src_stride, HFilter, VFilter,
                                        CONVERT_TO_SHORTPTR(dst_ptr),
                                        dst_stride, output_width, output_height,
                                        UUT_->use_highbd_);
496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512
    }
#else
    filter_average_block2d_8_c(src_ptr, src_stride, HFilter, VFilter,
                               dst_ptr, dst_stride, output_width,
                               output_height);
#endif
  }

  void wrapper_filter_block2d_8_c(const uint8_t *src_ptr,
                                  const unsigned int src_stride,
                                  const int16_t *HFilter,
                                  const int16_t *VFilter,
                                  uint8_t *dst_ptr,
                                  unsigned int dst_stride,
                                  unsigned int output_width,
                                  unsigned int output_height) {
#if CONFIG_VP9_HIGHBITDEPTH
513
    if (UUT_->use_highbd_ == 0) {
514 515 516
      filter_block2d_8_c(src_ptr, src_stride, HFilter, VFilter,
                         dst_ptr, dst_stride, output_width, output_height);
    } else {
517 518 519 520
      highbd_filter_block2d_8_c(CONVERT_TO_SHORTPTR(src_ptr), src_stride,
                                HFilter, VFilter,
                                CONVERT_TO_SHORTPTR(dst_ptr), dst_stride,
                                output_width, output_height, UUT_->use_highbd_);
521 522 523 524 525
    }
#else
    filter_block2d_8_c(src_ptr, src_stride, HFilter, VFilter,
                       dst_ptr, dst_stride, output_width, output_height);
#endif
James Zern's avatar
James Zern committed
526 527 528 529 530
  }

  const ConvolveFunctions* UUT_;
  static uint8_t* input_;
  static uint8_t* output_;
531
  static uint8_t* output_ref_;
532 533 534
#if CONFIG_VP9_HIGHBITDEPTH
  static uint16_t* input16_;
  static uint16_t* output16_;
535
  static uint16_t* output16_ref_;
536 537
  int mask_;
#endif
John Koleszar's avatar
John Koleszar committed
538
};
539

540 541
uint8_t* ConvolveTest::input_ = NULL;
uint8_t* ConvolveTest::output_ = NULL;
542
uint8_t* ConvolveTest::output_ref_ = NULL;
543 544 545
#if CONFIG_VP9_HIGHBITDEPTH
uint16_t* ConvolveTest::input16_ = NULL;
uint16_t* ConvolveTest::output16_ = NULL;
546
uint16_t* ConvolveTest::output16_ref_ = NULL;
547
#endif
John Koleszar's avatar
John Koleszar committed
548 549 550 551 552

TEST_P(ConvolveTest, GuardBlocks) {
  CheckGuardBlocks();
}

553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589
TEST_P(ConvolveTest, Copy) {
  uint8_t* const in = input();
  uint8_t* const out = output();

  ASM_REGISTER_STATE_CHECK(
      UUT_->copy_(in, kInputStride, out, kOutputStride, NULL, 0, NULL, 0,
                  Width(), Height()));

  CheckGuardBlocks();

  for (int y = 0; y < Height(); ++y)
    for (int x = 0; x < Width(); ++x)
      ASSERT_EQ(lookup(out, y * kOutputStride + x),
                lookup(in, y * kInputStride + x))
          << "(" << x << "," << y << ")";
}

TEST_P(ConvolveTest, Avg) {
  uint8_t* const in = input();
  uint8_t* const out = output();
  uint8_t* const out_ref = output_ref();
  CopyOutputToRef();

  ASM_REGISTER_STATE_CHECK(
      UUT_->avg_(in, kInputStride, out, kOutputStride, NULL, 0, NULL, 0,
                Width(), Height()));

  CheckGuardBlocks();

  for (int y = 0; y < Height(); ++y)
    for (int x = 0; x < Width(); ++x)
      ASSERT_EQ(lookup(out, y * kOutputStride + x),
                ROUND_POWER_OF_TWO(lookup(in, y * kInputStride + x) +
                                   lookup(out_ref, y * kOutputStride + x), 1))
          << "(" << x << "," << y << ")";
}

John Koleszar's avatar
John Koleszar committed
590 591 592
TEST_P(ConvolveTest, CopyHoriz) {
  uint8_t* const in = input();
  uint8_t* const out = output();
593
  DECLARE_ALIGNED(256, const int16_t, filter8[8]) = {0, 0, 0, 128, 0, 0, 0, 0};
John Koleszar's avatar
John Koleszar committed
594

595
  ASM_REGISTER_STATE_CHECK(
John Koleszar's avatar
John Koleszar committed
596 597 598 599 600 601 602
      UUT_->h8_(in, kInputStride, out, kOutputStride, filter8, 16, filter8, 16,
                Width(), Height()));

  CheckGuardBlocks();

  for (int y = 0; y < Height(); ++y)
    for (int x = 0; x < Width(); ++x)
603 604
      ASSERT_EQ(lookup(out, y * kOutputStride + x),
                lookup(in, y * kInputStride + x))
John Koleszar's avatar
John Koleszar committed
605 606 607 608 609 610
          << "(" << x << "," << y << ")";
}

TEST_P(ConvolveTest, CopyVert) {
  uint8_t* const in = input();
  uint8_t* const out = output();
611
  DECLARE_ALIGNED(256, const int16_t, filter8[8]) = {0, 0, 0, 128, 0, 0, 0, 0};
John Koleszar's avatar
John Koleszar committed
612

613
  ASM_REGISTER_STATE_CHECK(
John Koleszar's avatar
John Koleszar committed
614 615 616 617 618 619 620
      UUT_->v8_(in, kInputStride, out, kOutputStride, filter8, 16, filter8, 16,
                Width(), Height()));

  CheckGuardBlocks();

  for (int y = 0; y < Height(); ++y)
    for (int x = 0; x < Width(); ++x)
621 622
      ASSERT_EQ(lookup(out, y * kOutputStride + x),
                lookup(in, y * kInputStride + x))
John Koleszar's avatar
John Koleszar committed
623 624 625 626 627 628
          << "(" << x << "," << y << ")";
}

TEST_P(ConvolveTest, Copy2D) {
  uint8_t* const in = input();
  uint8_t* const out = output();
629
  DECLARE_ALIGNED(256, const int16_t, filter8[8]) = {0, 0, 0, 128, 0, 0, 0, 0};
John Koleszar's avatar
John Koleszar committed
630

631
  ASM_REGISTER_STATE_CHECK(
John Koleszar's avatar
John Koleszar committed
632 633 634 635 636 637 638
      UUT_->hv8_(in, kInputStride, out, kOutputStride, filter8, 16, filter8, 16,
                 Width(), Height()));

  CheckGuardBlocks();

  for (int y = 0; y < Height(); ++y)
    for (int x = 0; x < Width(); ++x)
639 640
      ASSERT_EQ(lookup(out, y * kOutputStride + x),
                lookup(in, y * kInputStride + x))
John Koleszar's avatar
John Koleszar committed
641 642 643
          << "(" << x << "," << y << ")";
}

644
const int kNumFilterBanks = 4;
645 646 647 648
const int kNumFilters = 16;

TEST(ConvolveTest, FiltersWontSaturateWhenAddedPairwise) {
  for (int filter_bank = 0; filter_bank < kNumFilterBanks; ++filter_bank) {
649 650
    const InterpKernel *filters =
        vp9_get_interp_kernel(static_cast<INTERP_FILTER>(filter_bank));
651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666
    for (int i = 0; i < kNumFilters; i++) {
      const int p0 = filters[i][0] + filters[i][1];
      const int p1 = filters[i][2] + filters[i][3];
      const int p2 = filters[i][4] + filters[i][5];
      const int p3 = filters[i][6] + filters[i][7];
      EXPECT_LE(p0, 128);
      EXPECT_LE(p1, 128);
      EXPECT_LE(p2, 128);
      EXPECT_LE(p3, 128);
      EXPECT_LE(p0 + p3, 128);
      EXPECT_LE(p0 + p3 + p1, 128);
      EXPECT_LE(p0 + p3 + p1 + p2, 128);
      EXPECT_EQ(p0 + p1 + p2 + p3, 128);
    }
  }
}
667

668 669
const int16_t kInvalidFilter[8] = { 0 };

John Koleszar's avatar
John Koleszar committed
670 671 672
TEST_P(ConvolveTest, MatchesReferenceSubpixelFilter) {
  uint8_t* const in = input();
  uint8_t* const out = output();
673 674 675 676
#if CONFIG_VP9_HIGHBITDEPTH
  uint8_t ref8[kOutputStride * kMaxDimension];
  uint16_t ref16[kOutputStride * kMaxDimension];
  uint8_t* ref;
677
  if (UUT_->use_highbd_ == 0) {
678 679 680 681 682
    ref = ref8;
  } else {
    ref = CONVERT_TO_BYTEPTR(ref16);
  }
#else
John Koleszar's avatar
John Koleszar committed
683
  uint8_t ref[kOutputStride * kMaxDimension];
684
#endif
685 686

  for (int filter_bank = 0; filter_bank < kNumFilterBanks; ++filter_bank) {
687 688
    const InterpKernel *filters =
        vp9_get_interp_kernel(static_cast<INTERP_FILTER>(filter_bank));
689 690 691
    const InterpKernel *const eighttap_smooth =
        vp9_get_interp_kernel(EIGHTTAP_SMOOTH);

692 693
    for (int filter_x = 0; filter_x < kNumFilters; ++filter_x) {
      for (int filter_y = 0; filter_y < kNumFilters; ++filter_y) {
694 695 696 697
        wrapper_filter_block2d_8_c(in, kInputStride,
                                   filters[filter_x], filters[filter_y],
                                   ref, kOutputStride,
                                   Width(), Height());
698

699
        if (filters == eighttap_smooth || (filter_x && filter_y))
700
          ASM_REGISTER_STATE_CHECK(
701 702 703 704
              UUT_->hv8_(in, kInputStride, out, kOutputStride,
                         filters[filter_x], 16, filters[filter_y], 16,
                         Width(), Height()));
        else if (filter_y)
705
          ASM_REGISTER_STATE_CHECK(
706
              UUT_->v8_(in, kInputStride, out, kOutputStride,
707
                        kInvalidFilter, 16, filters[filter_y], 16,
708 709
                        Width(), Height()));
        else
710
          ASM_REGISTER_STATE_CHECK(
711
              UUT_->h8_(in, kInputStride, out, kOutputStride,
712
                        filters[filter_x], 16, kInvalidFilter, 16,
713 714 715 716 717 718
                        Width(), Height()));

        CheckGuardBlocks();

        for (int y = 0; y < Height(); ++y)
          for (int x = 0; x < Width(); ++x)
719 720
            ASSERT_EQ(lookup(ref, y * kOutputStride + x),
                      lookup(out, y * kOutputStride + x))
721 722 723 724
                << "mismatch at (" << x << "," << y << "), "
                << "filters (" << filter_bank << ","
                << filter_x << "," << filter_y << ")";
      }
John Koleszar's avatar
John Koleszar committed
725 726 727 728 729 730 731
    }
  }
}

TEST_P(ConvolveTest, MatchesReferenceAveragingSubpixelFilter) {
  uint8_t* const in = input();
  uint8_t* const out = output();
732 733 734 735
#if CONFIG_VP9_HIGHBITDEPTH
  uint8_t ref8[kOutputStride * kMaxDimension];
  uint16_t ref16[kOutputStride * kMaxDimension];
  uint8_t* ref;
736
  if (UUT_->use_highbd_ == 0) {
737 738 739 740 741
    ref = ref8;
  } else {
    ref = CONVERT_TO_BYTEPTR(ref16);
  }
#else
John Koleszar's avatar
John Koleszar committed
742
  uint8_t ref[kOutputStride * kMaxDimension];
743
#endif
John Koleszar's avatar
John Koleszar committed
744 745 746 747 748

  // Populate ref and out with some random data
  ::libvpx_test::ACMRandom prng;
  for (int y = 0; y < Height(); ++y) {
    for (int x = 0; x < Width(); ++x) {
749 750
      uint16_t r;
#if CONFIG_VP9_HIGHBITDEPTH
751
      if (UUT_->use_highbd_ == 0 || UUT_->use_highbd_ == 8) {
752 753 754 755 756 757 758
        r = prng.Rand8Extremes();
      } else {
        r = prng.Rand16() & mask_;
      }
#else
      r = prng.Rand8Extremes();
#endif
John Koleszar's avatar
John Koleszar committed
759

760 761
      assign_val(out, y * kOutputStride + x, r);
      assign_val(ref, y * kOutputStride + x, r);
John Koleszar's avatar
John Koleszar committed
762 763 764
    }
  }

765
  for (int filter_bank = 0; filter_bank < kNumFilterBanks; ++filter_bank) {
766 767
    const InterpKernel *filters =
        vp9_get_interp_kernel(static_cast<INTERP_FILTER>(filter_bank));
768 769
    const InterpKernel *const eighttap_smooth =
        vp9_get_interp_kernel(EIGHTTAP_SMOOTH);
770 771 772

    for (int filter_x = 0; filter_x < kNumFilters; ++filter_x) {
      for (int filter_y = 0; filter_y < kNumFilters; ++filter_y) {
773 774 775 776
        wrapper_filter_average_block2d_8_c(in, kInputStride,
                                           filters[filter_x], filters[filter_y],
                                           ref, kOutputStride,
                                           Width(), Height());
777

778
        if (filters == eighttap_smooth || (filter_x && filter_y))
779
          ASM_REGISTER_STATE_CHECK(
780 781 782 783
              UUT_->hv8_avg_(in, kInputStride, out, kOutputStride,
                             filters[filter_x], 16, filters[filter_y], 16,
                             Width(), Height()));
        else if (filter_y)
784
          ASM_REGISTER_STATE_CHECK(
785 786 787 788
              UUT_->v8_avg_(in, kInputStride, out, kOutputStride,
                            filters[filter_x], 16, filters[filter_y], 16,
                            Width(), Height()));
        else
789
          ASM_REGISTER_STATE_CHECK(
790 791 792 793 794 795 796 797
              UUT_->h8_avg_(in, kInputStride, out, kOutputStride,
                            filters[filter_x], 16, filters[filter_y], 16,
                            Width(), Height()));

        CheckGuardBlocks();

        for (int y = 0; y < Height(); ++y)
          for (int x = 0; x < Width(); ++x)
798 799
            ASSERT_EQ(lookup(ref, y * kOutputStride + x),
                      lookup(out, y * kOutputStride + x))
800 801 802 803
                << "mismatch at (" << x << "," << y << "), "
                << "filters (" << filter_bank << ","
                << filter_x << "," << filter_y << ")";
      }
John Koleszar's avatar
John Koleszar committed
804 805 806 807
    }
  }
}

808 809 810 811 812 813 814
TEST_P(ConvolveTest, FilterExtremes) {
  uint8_t *const in = input();
  uint8_t *const out = output();
#if CONFIG_VP9_HIGHBITDEPTH
  uint8_t ref8[kOutputStride * kMaxDimension];
  uint16_t ref16[kOutputStride * kMaxDimension];
  uint8_t *ref;
815
  if (UUT_->use_highbd_ == 0) {
816 817 818 819 820 821 822 823 824 825 826 827 828 829
    ref = ref8;
  } else {
    ref = CONVERT_TO_BYTEPTR(ref16);
  }
#else
  uint8_t ref[kOutputStride * kMaxDimension];
#endif

  // Populate ref and out with some random data
  ::libvpx_test::ACMRandom prng;
  for (int y = 0; y < Height(); ++y) {
    for (int x = 0; x < Width(); ++x) {
      uint16_t r;
#if CONFIG_VP9_HIGHBITDEPTH
830
      if (UUT_->use_highbd_ == 0 || UUT_->use_highbd_ == 8) {
831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904
        r = prng.Rand8Extremes();
      } else {
        r = prng.Rand16() & mask_;
      }
#else
      r = prng.Rand8Extremes();
#endif
      assign_val(out, y * kOutputStride + x, r);
      assign_val(ref, y * kOutputStride + x, r);
    }
  }

  for (int axis = 0; axis < 2; axis++) {
    int seed_val = 0;
    while (seed_val < 256) {
      for (int y = 0; y < 8; ++y) {
        for (int x = 0; x < 8; ++x) {
#if CONFIG_VP9_HIGHBITDEPTH
            assign_val(in, y * kOutputStride + x - SUBPEL_TAPS / 2 + 1,
                       ((seed_val >> (axis ? y : x)) & 1) * mask_);
#else
            assign_val(in, y * kOutputStride + x - SUBPEL_TAPS / 2 + 1,
                       ((seed_val >> (axis ? y : x)) & 1) * 255);
#endif
          if (axis) seed_val++;
        }
        if (axis)
          seed_val-= 8;
        else
          seed_val++;
      }
      if (axis) seed_val += 8;

      for (int filter_bank = 0; filter_bank < kNumFilterBanks; ++filter_bank) {
        const InterpKernel *filters =
            vp9_get_interp_kernel(static_cast<INTERP_FILTER>(filter_bank));
        const InterpKernel *const eighttap_smooth =
            vp9_get_interp_kernel(EIGHTTAP_SMOOTH);
        for (int filter_x = 0; filter_x < kNumFilters; ++filter_x) {
          for (int filter_y = 0; filter_y < kNumFilters; ++filter_y) {
            wrapper_filter_block2d_8_c(in, kInputStride,
                                       filters[filter_x], filters[filter_y],
                                       ref, kOutputStride,
                                       Width(), Height());
            if (filters == eighttap_smooth || (filter_x && filter_y))
              ASM_REGISTER_STATE_CHECK(
                  UUT_->hv8_(in, kInputStride, out, kOutputStride,
                             filters[filter_x], 16, filters[filter_y], 16,
                             Width(), Height()));
            else if (filter_y)
              ASM_REGISTER_STATE_CHECK(
                  UUT_->v8_(in, kInputStride, out, kOutputStride,
                            kInvalidFilter, 16, filters[filter_y], 16,
                            Width(), Height()));
            else
              ASM_REGISTER_STATE_CHECK(
                  UUT_->h8_(in, kInputStride, out, kOutputStride,
                            filters[filter_x], 16, kInvalidFilter, 16,
                            Width(), Height()));

            for (int y = 0; y < Height(); ++y)
              for (int x = 0; x < Width(); ++x)
                ASSERT_EQ(lookup(ref, y * kOutputStride + x),
                          lookup(out, y * kOutputStride + x))
                    << "mismatch at (" << x << "," << y << "), "
                    << "filters (" << filter_bank << ","
                    << filter_x << "," << filter_y << ")";
          }
        }
      }
    }
  }
}

905
DECLARE_ALIGNED(256, const int16_t, kChangeFilters[16][8]) = {
John Koleszar's avatar
John Koleszar committed
906 907 908 909 910 911 912 913 914 915 916 917 918 919 920
    { 0,   0,   0,   0,   0,   0,   0, 128},
    { 0,   0,   0,   0,   0,   0, 128},
    { 0,   0,   0,   0,   0, 128},
    { 0,   0,   0,   0, 128},
    { 0,   0,   0, 128},
    { 0,   0, 128},
    { 0, 128},
    { 128},
    { 0,   0,   0,   0,   0,   0,   0, 128},
    { 0,   0,   0,   0,   0,   0, 128},
    { 0,   0,   0,   0,   0, 128},
    { 0,   0,   0,   0, 128},
    { 0,   0,   0, 128},
    { 0,   0, 128},
    { 0, 128},
921 922 923
    { 128}
};

924
/* This test exercises the horizontal and vertical filter functions. */
925 926 927
TEST_P(ConvolveTest, ChangeFilterWorks) {
  uint8_t* const in = input();
  uint8_t* const out = output();
928 929 930 931 932 933 934 935 936 937 938

  /* Assume that the first input sample is at the 8/16th position. */
  const int kInitialSubPelOffset = 8;

  /* Filters are 8-tap, so the first filter tap will be applied to the pixel
   * at position -3 with respect to the current filtering position. Since
   * kInitialSubPelOffset is set to 8, we first select sub-pixel filter 8,
   * which is non-zero only in the last tap. So, applying the filter at the
   * current input position will result in an output equal to the pixel at
   * offset +4 (-3 + 7) with respect to the current filtering position.
   */
939
  const int kPixelSelected = 4;
John Koleszar's avatar
John Koleszar committed
940

941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956
  /* Assume that each output pixel requires us to step on by 17/16th pixels in
   * the input.
   */
  const int kInputPixelStep = 17;

  /* The filters are setup in such a way that the expected output produces
   * sets of 8 identical output samples. As the filter position moves to the
   * next 1/16th pixel position the only active (=128) filter tap moves one
   * position to the left, resulting in the same input pixel being replicated
   * in to the output for 8 consecutive samples. After each set of 8 positions
   * the filters select a different input pixel. kFilterPeriodAdjust below
   * computes which input pixel is written to the output for a specified
   * x or y position.
   */

  /* Test the horizontal filter. */
957 958 959 960
  ASM_REGISTER_STATE_CHECK(
      UUT_->h8_(in, kInputStride, out, kOutputStride,
                kChangeFilters[kInitialSubPelOffset],
                kInputPixelStep, NULL, 0, Width(), Height()));
John Koleszar's avatar
John Koleszar committed
961

962
  for (int x = 0; x < Width(); ++x) {
963
    const int kFilterPeriodAdjust = (x >> 3) << 3;
964 965 966 967
    const int ref_x =
        kPixelSelected + ((kInitialSubPelOffset
            + kFilterPeriodAdjust * kInputPixelStep)
                          >> SUBPEL_BITS);
968 969
    ASSERT_EQ(lookup(in, ref_x), lookup(out, x))
        << "x == " << x << "width = " << Width();
John Koleszar's avatar
John Koleszar committed
970 971
  }

972
  /* Test the vertical filter. */
973 974 975 976
  ASM_REGISTER_STATE_CHECK(
      UUT_->v8_(in, kInputStride, out, kOutputStride,
                NULL, 0, kChangeFilters[kInitialSubPelOffset],
                kInputPixelStep, Width(), Height()));
John Koleszar's avatar
John Koleszar committed
977

978
  for (int y = 0; y < Height(); ++y) {
979
    const int kFilterPeriodAdjust = (y >> 3) << 3;
980 981 982 983
    const int ref_y =
        kPixelSelected + ((kInitialSubPelOffset
            + kFilterPeriodAdjust * kInputPixelStep)
                          >> SUBPEL_BITS);
984 985
    ASSERT_EQ(lookup(in, ref_y * kInputStride), lookup(out, y * kInputStride))
        << "y == " << y;
John Koleszar's avatar
John Koleszar committed
986 987
  }

988
  /* Test the horizontal and vertical filters in combination. */
989 990 991 992 993
  ASM_REGISTER_STATE_CHECK(
      UUT_->hv8_(in, kInputStride, out, kOutputStride,
                 kChangeFilters[kInitialSubPelOffset], kInputPixelStep,
                 kChangeFilters[kInitialSubPelOffset], kInputPixelStep,
                 Width(), Height()));
John Koleszar's avatar
John Koleszar committed
994

995
  for (int y = 0; y < Height(); ++y) {
996
    const int kFilterPeriodAdjustY = (y >> 3) << 3;
997 998 999 1000
    const int ref_y =
        kPixelSelected + ((kInitialSubPelOffset
            + kFilterPeriodAdjustY * kInputPixelStep)
                          >> SUBPEL_BITS);
1001
    for (int x = 0; x < Width(); ++x) {
1002
      const int kFilterPeriodAdjustX = (x >> 3) << 3;
1003 1004 1005 1006
      const int ref_x =
          kPixelSelected + ((kInitialSubPelOffset
              + kFilterPeriodAdjustX * kInputPixelStep)
                            >> SUBPEL_BITS);
1007

1008 1009
      ASSERT_EQ(lookup(in, ref_y * kInputStride + ref_x),
                lookup(out, y * kOutputStride + x))
John Koleszar's avatar
John Koleszar committed
1010 1011 1012 1013 1014
          << "x == " << x << ", y == " << y;
    }
  }
}

1015 1016 1017 1018 1019
/* This test exercises that enough rows and columns are filtered with every
   possible initial fractional positions and scaling steps. */
TEST_P(ConvolveTest, CheckScalingFiltering) {
  uint8_t* const in = input();
  uint8_t* const out = output();
1020
  const InterpKernel *const eighttap = vp9_get_interp_kernel(EIGHTTAP);
1021 1022 1023 1024 1025 1026

  SetConstantInput(127);

  for (int frac = 0; frac < 16; ++frac) {
    for (int step = 1; step <= 32; ++step) {
      /* Test the horizontal and vertical filters in combination. */
1027 1028 1029 1030
      ASM_REGISTER_STATE_CHECK(UUT_->hv8_(in, kInputStride, out, kOutputStride,
                                          eighttap[frac], step,
                                          eighttap[frac], step,
                                          Width(), Height()));
1031 1032 1033 1034 1035

      CheckGuardBlocks();

      for (int y = 0; y < Height(); ++y) {
        for (int x = 0; x < Width(); ++x) {
1036 1037
          ASSERT_EQ(lookup(in, y * kInputStride + x),
                    lookup(out, y * kOutputStride + x))
1038 1039 1040 1041 1042 1043 1044 1045
              << "x == " << x << ", y == " << y
              << ", frac == " << frac << ", step == " << step;
        }
      }
    }
  }
}

John Koleszar's avatar
John Koleszar committed
1046 1047
using std::tr1::make_tuple;

1048
#if CONFIG_VP9_HIGHBITDEPTH
1049 1050 1051 1052 1053 1054 1055 1056
#if HAVE_SSE2 && ARCH_X86_64
void wrap_convolve8_horiz_sse2_8(const uint8_t *src, ptrdiff_t src_stride,
                                 uint8_t *dst, ptrdiff_t dst_stride,
                                 const int16_t *filter_x,
                                 int filter_x_stride,
                                 const int16_t *filter_y,
                                 int filter_y_stride,
                                 int w, int h) {
1057 1058 1059
  vp9_highbd_convolve8_horiz_sse2(src, src_stride, dst, dst_stride, filter_x,
                                  filter_x_stride, filter_y, filter_y_stride,
                                  w, h, 8);
1060 1061 1062 1063 1064 1065 1066 1067 1068
}

void wrap_convolve8_avg_horiz_sse2_8(const uint8_t *src, ptrdiff_t src_stride,
                                     uint8_t *dst, ptrdiff_t dst_stride,
                                     const int16_t *filter_x,
                                     int filter_x_stride,
                                     const int16_t *filter_y,
                                     int filter_y_stride,
                                     int w, int h) {
1069 1070 1071
  vp9_highbd_convolve8_avg_horiz_sse2(src, src_stride, dst, dst_stride,
                                      filter_x, filter_x_stride,
                                      filter_y, filter_y_stride, w, h, 8);
1072 1073 1074 1075 1076 1077 1078 1079 1080
}

void wrap_convolve8_vert_sse2_8(const uint8_t *src, ptrdiff_t src_stride,
                                uint8_t *dst, ptrdiff_t dst_stride,
                                const int16_t *filter_x,
                                int filter_x_stride,
                                const int16_t *filter_y,
                                int filter_y_stride,
                                int w, int h) {
1081 1082 1083
  vp9_highbd_convolve8_vert_sse2(src, src_stride, dst, dst_stride,
                                 filter_x, filter_x_stride,
                                 filter_y, filter_y_stride, w, h, 8);
1084 1085 1086 1087 1088 1089 1090 1091 1092
}

void wrap_convolve8_avg_vert_sse2_8(const uint8_t *src, ptrdiff_t src_stride,
                                    uint8_t *dst, ptrdiff_t dst_stride,
                                    const int16_t *filter_x,
                                    int filter_x_stride,
                                    const int16_t *filter_y,
                                    int filter_y_stride,
                                    int w, int h) {