partial_idct_test.cc 10.3 KB
Newer Older
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
/*
 *  Copyright (c) 2013 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 <math.h>
#include <stdlib.h>
#include <string.h>

#include "third_party/googletest/src/include/gtest/gtest.h"
#include "test/acm_random.h"
#include "test/clear_system_state.h"
#include "test/register_state_check.h"
#include "test/util.h"

#include "./vp9_rtcd.h"
#include "vp9/common/vp9_blockd.h"
#include "vp9/common/vp9_scan.h"
#include "vpx/vpx_integer.h"

using libvpx_test::ACMRandom;

namespace {
29 30
typedef void (*FwdTxfmFunc)(const int16_t *in, tran_low_t *out, int stride);
typedef void (*InvTxfmFunc)(const tran_low_t *in, uint8_t *out, int stride);
31 32 33 34
typedef std::tr1::tuple<FwdTxfmFunc,
                        InvTxfmFunc,
                        InvTxfmFunc,
                        TX_SIZE, int> PartialInvTxfmParam;
35
const int kMaxNumCoeffs = 1024;
36
class PartialIDctTest : public ::testing::TestWithParam<PartialInvTxfmParam> {
37 38 39
 public:
  virtual ~PartialIDctTest() {}
  virtual void SetUp() {
40 41 42 43 44
    ftxfm_ = GET_PARAM(0);
    full_itxfm_ = GET_PARAM(1);
    partial_itxfm_ = GET_PARAM(2);
    tx_size_  = GET_PARAM(3);
    last_nonzero_ = GET_PARAM(4);
45 46 47 48 49 50 51
  }

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

 protected:
  int last_nonzero_;
  TX_SIZE tx_size_;
52 53 54
  FwdTxfmFunc ftxfm_;
  InvTxfmFunc full_itxfm_;
  InvTxfmFunc partial_itxfm_;
55 56
};

57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76
TEST_P(PartialIDctTest, RunQuantCheck) {
  ACMRandom rnd(ACMRandom::DeterministicSeed());
  int size;
  switch (tx_size_) {
    case TX_4X4:
      size = 4;
      break;
    case TX_8X8:
      size = 8;
      break;
    case TX_16X16:
      size = 16;
      break;
    case TX_32X32:
      size = 32;
      break;
    default:
      FAIL() << "Wrong Size!";
      break;
  }
77 78
  DECLARE_ALIGNED_ARRAY(16, tran_low_t, test_coef_block1, kMaxNumCoeffs);
  DECLARE_ALIGNED_ARRAY(16, tran_low_t, test_coef_block2, kMaxNumCoeffs);
79 80 81 82 83 84 85
  DECLARE_ALIGNED_ARRAY(16, uint8_t, dst1, kMaxNumCoeffs);
  DECLARE_ALIGNED_ARRAY(16, uint8_t, dst2, kMaxNumCoeffs);

  const int count_test_block = 1000;
  const int block_size = size * size;

  DECLARE_ALIGNED_ARRAY(16, int16_t, input_extreme_block, kMaxNumCoeffs);
86
  DECLARE_ALIGNED_ARRAY(16, tran_low_t, output_ref_block, kMaxNumCoeffs);
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

  int max_error = 0;
  for (int i = 0; i < count_test_block; ++i) {
    // clear out destination buffer
    memset(dst1, 0, sizeof(*dst1) * block_size);
    memset(dst2, 0, sizeof(*dst2) * block_size);
    memset(test_coef_block1, 0, sizeof(*test_coef_block1) * block_size);
    memset(test_coef_block2, 0, sizeof(*test_coef_block2) * block_size);

    ACMRandom rnd(ACMRandom::DeterministicSeed());

    for (int i = 0; i < count_test_block; ++i) {
      // Initialize a test block with input range [-255, 255].
      if (i == 0) {
        for (int j = 0; j < block_size; ++j)
          input_extreme_block[j] = 255;
      } else if (i == 1) {
        for (int j = 0; j < block_size; ++j)
          input_extreme_block[j] = -255;
      } else {
        for (int j = 0; j < block_size; ++j) {
          input_extreme_block[j] = rnd.Rand8() % 2 ? 255 : -255;
        }
      }

      ftxfm_(input_extreme_block, output_ref_block, size);

      // quantization with maximum allowed step sizes
      test_coef_block1[0] = (output_ref_block[0] / 1336) * 1336;
      for (int j = 1; j < last_nonzero_; ++j)
        test_coef_block1[vp9_default_scan_orders[tx_size_].scan[j]]
                         = (output_ref_block[j] / 1828) * 1828;
    }

121 122
    ASM_REGISTER_STATE_CHECK(full_itxfm_(test_coef_block1, dst1, size));
    ASM_REGISTER_STATE_CHECK(partial_itxfm_(test_coef_block1, dst2, size));
123 124 125 126 127 128 129 130 131 132 133 134 135

    for (int j = 0; j < block_size; ++j) {
      const int diff = dst1[j] - dst2[j];
      const int error = diff * diff;
      if (max_error < error)
        max_error = error;
    }
  }

  EXPECT_EQ(0, max_error)
      << "Error: partial inverse transform produces different results";
}

136 137 138 139
TEST_P(PartialIDctTest, ResultsMatch) {
  ACMRandom rnd(ACMRandom::DeterministicSeed());
  int size;
  switch (tx_size_) {
140 141 142 143 144 145 146 147 148 149 150 151 152 153 154
    case TX_4X4:
      size = 4;
      break;
    case TX_8X8:
      size = 8;
      break;
    case TX_16X16:
      size = 16;
      break;
    case TX_32X32:
      size = 32;
      break;
    default:
      FAIL() << "Wrong Size!";
      break;
155
  }
156 157
  DECLARE_ALIGNED_ARRAY(16, tran_low_t, test_coef_block1, kMaxNumCoeffs);
  DECLARE_ALIGNED_ARRAY(16, tran_low_t, test_coef_block2, kMaxNumCoeffs);
158 159 160 161 162 163 164 165 166 167 168 169 170 171
  DECLARE_ALIGNED_ARRAY(16, uint8_t, dst1, kMaxNumCoeffs);
  DECLARE_ALIGNED_ARRAY(16, uint8_t, dst2, kMaxNumCoeffs);
  const int count_test_block = 1000;
  const int max_coeff = 32766 / 4;
  const int block_size = size * size;
  int max_error = 0;
  for (int i = 0; i < count_test_block; ++i) {
    // clear out destination buffer
    memset(dst1, 0, sizeof(*dst1) * block_size);
    memset(dst2, 0, sizeof(*dst2) * block_size);
    memset(test_coef_block1, 0, sizeof(*test_coef_block1) * block_size);
    memset(test_coef_block2, 0, sizeof(*test_coef_block2) * block_size);
    int max_energy_leftover = max_coeff * max_coeff;
    for (int j = 0; j < last_nonzero_; ++j) {
172
      int16_t coef = static_cast<int16_t>(sqrt(1.0 * max_energy_leftover) *
173 174 175 176 177 178 179 180 181 182 183 184
                                          (rnd.Rand16() - 32768) / 65536);
      max_energy_leftover -= coef * coef;
      if (max_energy_leftover < 0) {
        max_energy_leftover = 0;
        coef = 0;
      }
      test_coef_block1[vp9_default_scan_orders[tx_size_].scan[j]] = coef;
    }

    memcpy(test_coef_block2, test_coef_block1,
           sizeof(*test_coef_block2) * block_size);

185 186
    ASM_REGISTER_STATE_CHECK(full_itxfm_(test_coef_block1, dst1, size));
    ASM_REGISTER_STATE_CHECK(partial_itxfm_(test_coef_block2, dst2, size));
187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203

    for (int j = 0; j < block_size; ++j) {
      const int diff = dst1[j] - dst2[j];
      const int error = diff * diff;
      if (max_error < error)
        max_error = error;
    }
  }

  EXPECT_EQ(0, max_error)
      << "Error: partial inverse transform produces different results";
}
using std::tr1::make_tuple;

INSTANTIATE_TEST_CASE_P(
    C, PartialIDctTest,
    ::testing::Values(
204 205
        make_tuple(&vp9_fdct32x32_c,
                   &vp9_idct32x32_1024_add_c,
206
                   &vp9_idct32x32_34_add_c,
207
                   TX_32X32, 34),
208 209
        make_tuple(&vp9_fdct32x32_c,
                   &vp9_idct32x32_1024_add_c,
210
                   &vp9_idct32x32_1_add_c,
211
                   TX_32X32, 1),
212 213
        make_tuple(&vp9_fdct16x16_c,
                   &vp9_idct16x16_256_add_c,
214
                   &vp9_idct16x16_10_add_c,
215
                   TX_16X16, 10),
216 217
        make_tuple(&vp9_fdct16x16_c,
                   &vp9_idct16x16_256_add_c,
218
                   &vp9_idct16x16_1_add_c,
219
                   TX_16X16, 1),
220 221
        make_tuple(&vp9_fdct8x8_c,
                   &vp9_idct8x8_64_add_c,
222 223
                   &vp9_idct8x8_12_add_c,
                   TX_8X8, 12),
224 225
        make_tuple(&vp9_fdct8x8_c,
                   &vp9_idct8x8_64_add_c,
226
                   &vp9_idct8x8_1_add_c,
227
                   TX_8X8, 1),
228 229
        make_tuple(&vp9_fdct4x4_c,
                   &vp9_idct4x4_16_add_c,
230
                   &vp9_idct4x4_1_add_c,
231
                   TX_4X4, 1)));
232

233
#if HAVE_NEON
234 235 236
INSTANTIATE_TEST_CASE_P(
    NEON, PartialIDctTest,
    ::testing::Values(
237 238
        make_tuple(&vp9_fdct32x32_c,
                   &vp9_idct32x32_1024_add_c,
239 240
                   &vp9_idct32x32_1_add_neon,
                   TX_32X32, 1),
241 242
        make_tuple(&vp9_fdct16x16_c,
                   &vp9_idct16x16_256_add_c,
243 244
                   &vp9_idct16x16_10_add_neon,
                   TX_16X16, 10),
245 246
        make_tuple(&vp9_fdct16x16_c,
                   &vp9_idct16x16_256_add_c,
247 248
                   &vp9_idct16x16_1_add_neon,
                   TX_16X16, 1),
249 250
        make_tuple(&vp9_fdct8x8_c,
                   &vp9_idct8x8_64_add_c,
251 252
                   &vp9_idct8x8_12_add_neon,
                   TX_8X8, 12),
253 254
        make_tuple(&vp9_fdct8x8_c,
                   &vp9_idct8x8_64_add_c,
255 256
                   &vp9_idct8x8_1_add_neon,
                   TX_8X8, 1),
257 258
        make_tuple(&vp9_fdct4x4_c,
                   &vp9_idct4x4_16_add_c,
259 260
                   &vp9_idct4x4_1_add_neon,
                   TX_4X4, 1)));
261
#endif  // HAVE_NEON
262

263
#if HAVE_SSE2 && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
264 265 266
INSTANTIATE_TEST_CASE_P(
    SSE2, PartialIDctTest,
    ::testing::Values(
267 268
        make_tuple(&vp9_fdct32x32_c,
                   &vp9_idct32x32_1024_add_c,
269
                   &vp9_idct32x32_34_add_sse2,
270
                   TX_32X32, 34),
271 272
        make_tuple(&vp9_fdct32x32_c,
                   &vp9_idct32x32_1024_add_c,
273
                   &vp9_idct32x32_1_add_sse2,
274
                   TX_32X32, 1),
275 276
        make_tuple(&vp9_fdct16x16_c,
                   &vp9_idct16x16_256_add_c,
277
                   &vp9_idct16x16_10_add_sse2,
278
                   TX_16X16, 10),
279 280
        make_tuple(&vp9_fdct16x16_c,
                   &vp9_idct16x16_256_add_c,
281
                   &vp9_idct16x16_1_add_sse2,
282
                   TX_16X16, 1),
283 284
        make_tuple(&vp9_fdct8x8_c,
                   &vp9_idct8x8_64_add_c,
285 286
                   &vp9_idct8x8_12_add_sse2,
                   TX_8X8, 12),
287 288
        make_tuple(&vp9_fdct8x8_c,
                   &vp9_idct8x8_64_add_c,
289
                   &vp9_idct8x8_1_add_sse2,
290
                   TX_8X8, 1),
291 292
        make_tuple(&vp9_fdct4x4_c,
                   &vp9_idct4x4_16_add_c,
293
                   &vp9_idct4x4_1_add_sse2,
294 295
                   TX_4X4, 1)));
#endif
296

297 298 299 300 301 302 303 304 305 306 307
#if HAVE_SSSE3 && ARCH_X86_64 && !CONFIG_VP9_HIGHBITDEPTH && \
    !CONFIG_EMULATE_HARDWARE
INSTANTIATE_TEST_CASE_P(
    SSSE3_64, PartialIDctTest,
    ::testing::Values(
        make_tuple(&vp9_fdct8x8_c,
                   &vp9_idct8x8_64_add_c,
                   &vp9_idct8x8_12_add_ssse3,
                   TX_8X8, 12)));
#endif

308
#if HAVE_SSSE3 && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
309 310 311 312 313 314
INSTANTIATE_TEST_CASE_P(
    SSSE3, PartialIDctTest,
    ::testing::Values(
        make_tuple(&vp9_fdct16x16_c,
                   &vp9_idct16x16_256_add_c,
                   &vp9_idct16x16_10_add_ssse3,
315 316
                   TX_16X16, 10)));
#endif
317
}  // namespace