partial_idct_test.cc 11.4 KB
Newer Older
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
/*
 *  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"
22
#include "./vpx_dsp_rtcd.h"
23
24
25
26
27
28
29
#include "vp9/common/vp9_blockd.h"
#include "vp9/common/vp9_scan.h"
#include "vpx/vpx_integer.h"

using libvpx_test::ACMRandom;

namespace {
30
31
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);
32
33
34
35
typedef std::tr1::tuple<FwdTxfmFunc,
                        InvTxfmFunc,
                        InvTxfmFunc,
                        TX_SIZE, int> PartialInvTxfmParam;
36
const int kMaxNumCoeffs = 1024;
37
class PartialIDctTest : public ::testing::TestWithParam<PartialInvTxfmParam> {
38
39
40
 public:
  virtual ~PartialIDctTest() {}
  virtual void SetUp() {
41
42
43
44
45
    ftxfm_ = GET_PARAM(0);
    full_itxfm_ = GET_PARAM(1);
    partial_itxfm_ = GET_PARAM(2);
    tx_size_  = GET_PARAM(3);
    last_nonzero_ = GET_PARAM(4);
46
47
48
49
50
51
52
  }

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

 protected:
  int last_nonzero_;
  TX_SIZE tx_size_;
53
54
55
  FwdTxfmFunc ftxfm_;
  InvTxfmFunc full_itxfm_;
  InvTxfmFunc partial_itxfm_;
56
57
};

58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
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;
  }
78
79
80
81
  DECLARE_ALIGNED(16, tran_low_t, test_coef_block1[kMaxNumCoeffs]);
  DECLARE_ALIGNED(16, tran_low_t, test_coef_block2[kMaxNumCoeffs]);
  DECLARE_ALIGNED(16, uint8_t, dst1[kMaxNumCoeffs]);
  DECLARE_ALIGNED(16, uint8_t, dst2[kMaxNumCoeffs]);
82
83
84
85

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

86
87
  DECLARE_ALIGNED(16, int16_t, input_extreme_block[kMaxNumCoeffs]);
  DECLARE_ALIGNED(16, tran_low_t, output_ref_block[kMaxNumCoeffs]);
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

  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;
    }

122
123
    ASM_REGISTER_STATE_CHECK(full_itxfm_(test_coef_block1, dst1, size));
    ASM_REGISTER_STATE_CHECK(partial_itxfm_(test_coef_block1, dst2, size));
124
125
126
127
128
129
130
131
132
133
134
135
136

    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";
}

137
138
139
140
TEST_P(PartialIDctTest, ResultsMatch) {
  ACMRandom rnd(ACMRandom::DeterministicSeed());
  int size;
  switch (tx_size_) {
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
    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;
156
  }
157
158
159
160
  DECLARE_ALIGNED(16, tran_low_t, test_coef_block1[kMaxNumCoeffs]);
  DECLARE_ALIGNED(16, tran_low_t, test_coef_block2[kMaxNumCoeffs]);
  DECLARE_ALIGNED(16, uint8_t, dst1[kMaxNumCoeffs]);
  DECLARE_ALIGNED(16, uint8_t, dst2[kMaxNumCoeffs]);
161
162
163
164
165
166
167
168
169
170
171
172
  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) {
173
      int16_t coef = static_cast<int16_t>(sqrt(1.0 * max_energy_leftover) *
174
175
176
177
178
179
180
181
182
183
184
185
                                          (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);

186
187
    ASM_REGISTER_STATE_CHECK(full_itxfm_(test_coef_block1, dst1, size));
    ASM_REGISTER_STATE_CHECK(partial_itxfm_(test_coef_block2, dst2, size));
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204

    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(
205
206
        make_tuple(&vp9_fdct32x32_c,
                   &vp9_idct32x32_1024_add_c,
207
                   &vp9_idct32x32_34_add_c,
208
                   TX_32X32, 34),
209
210
        make_tuple(&vp9_fdct32x32_c,
                   &vp9_idct32x32_1024_add_c,
211
                   &vp9_idct32x32_1_add_c,
212
                   TX_32X32, 1),
213
214
        make_tuple(&vp9_fdct16x16_c,
                   &vp9_idct16x16_256_add_c,
215
                   &vp9_idct16x16_10_add_c,
216
                   TX_16X16, 10),
217
218
        make_tuple(&vp9_fdct16x16_c,
                   &vp9_idct16x16_256_add_c,
219
                   &vp9_idct16x16_1_add_c,
220
                   TX_16X16, 1),
221
222
        make_tuple(&vp9_fdct8x8_c,
                   &vp9_idct8x8_64_add_c,
223
224
                   &vp9_idct8x8_12_add_c,
                   TX_8X8, 12),
225
226
        make_tuple(&vp9_fdct8x8_c,
                   &vp9_idct8x8_64_add_c,
227
                   &vp9_idct8x8_1_add_c,
228
                   TX_8X8, 1),
229
230
        make_tuple(&vp9_fdct4x4_c,
                   &vp9_idct4x4_16_add_c,
231
                   &vp9_idct4x4_1_add_c,
232
                   TX_4X4, 1)));
233

James Zern's avatar
James Zern committed
234
#if HAVE_NEON && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
235
236
237
INSTANTIATE_TEST_CASE_P(
    NEON, PartialIDctTest,
    ::testing::Values(
238
239
        make_tuple(&vp9_fdct32x32_c,
                   &vp9_idct32x32_1024_add_c,
240
241
                   &vp9_idct32x32_1_add_neon,
                   TX_32X32, 1),
242
243
        make_tuple(&vp9_fdct16x16_c,
                   &vp9_idct16x16_256_add_c,
244
245
                   &vp9_idct16x16_10_add_neon,
                   TX_16X16, 10),
246
247
        make_tuple(&vp9_fdct16x16_c,
                   &vp9_idct16x16_256_add_c,
248
249
                   &vp9_idct16x16_1_add_neon,
                   TX_16X16, 1),
250
251
        make_tuple(&vp9_fdct8x8_c,
                   &vp9_idct8x8_64_add_c,
252
253
                   &vp9_idct8x8_12_add_neon,
                   TX_8X8, 12),
254
255
        make_tuple(&vp9_fdct8x8_c,
                   &vp9_idct8x8_64_add_c,
256
257
                   &vp9_idct8x8_1_add_neon,
                   TX_8X8, 1),
258
259
        make_tuple(&vp9_fdct4x4_c,
                   &vp9_idct4x4_16_add_c,
260
261
                   &vp9_idct4x4_1_add_neon,
                   TX_4X4, 1)));
James Zern's avatar
James Zern committed
262
#endif  // HAVE_NEON && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
263

264
#if HAVE_SSE2 && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
265
266
267
INSTANTIATE_TEST_CASE_P(
    SSE2, PartialIDctTest,
    ::testing::Values(
268
269
        make_tuple(&vp9_fdct32x32_c,
                   &vp9_idct32x32_1024_add_c,
270
                   &vp9_idct32x32_34_add_sse2,
271
                   TX_32X32, 34),
272
273
        make_tuple(&vp9_fdct32x32_c,
                   &vp9_idct32x32_1024_add_c,
274
                   &vp9_idct32x32_1_add_sse2,
275
                   TX_32X32, 1),
276
277
        make_tuple(&vp9_fdct16x16_c,
                   &vp9_idct16x16_256_add_c,
278
                   &vp9_idct16x16_10_add_sse2,
279
                   TX_16X16, 10),
280
281
        make_tuple(&vp9_fdct16x16_c,
                   &vp9_idct16x16_256_add_c,
282
                   &vp9_idct16x16_1_add_sse2,
283
                   TX_16X16, 1),
284
285
        make_tuple(&vp9_fdct8x8_c,
                   &vp9_idct8x8_64_add_c,
286
287
                   &vp9_idct8x8_12_add_sse2,
                   TX_8X8, 12),
288
289
        make_tuple(&vp9_fdct8x8_c,
                   &vp9_idct8x8_64_add_c,
290
                   &vp9_idct8x8_1_add_sse2,
291
                   TX_8X8, 1),
292
293
        make_tuple(&vp9_fdct4x4_c,
                   &vp9_idct4x4_16_add_c,
294
                   &vp9_idct4x4_1_add_sse2,
295
296
                   TX_4X4, 1)));
#endif
297

Johann's avatar
Johann committed
298
299
#if HAVE_SSSE3 && CONFIG_USE_X86INC && ARCH_X86_64 && \
    !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
300
301
302
303
304
305
306
307
308
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

309
#if HAVE_MSA && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
310
311
312
INSTANTIATE_TEST_CASE_P(
    MSA, PartialIDctTest,
    ::testing::Values(
313
314
315
316
317
318
319
320
        make_tuple(&vp9_fdct32x32_c,
                   &vp9_idct32x32_1024_add_c,
                   &vp9_idct32x32_34_add_msa,
                   TX_32X32, 34),
        make_tuple(&vp9_fdct32x32_c,
                   &vp9_idct32x32_1024_add_c,
                   &vp9_idct32x32_1_add_msa,
                   TX_32X32, 1),
321
322
323
324
325
326
327
        make_tuple(&vp9_fdct16x16_c,
                   &vp9_idct16x16_256_add_c,
                   &vp9_idct16x16_10_add_msa,
                   TX_16X16, 10),
        make_tuple(&vp9_fdct16x16_c,
                   &vp9_idct16x16_256_add_c,
                   &vp9_idct16x16_1_add_msa,
328
329
330
331
332
333
334
335
                   TX_16X16, 1),
        make_tuple(&vp9_fdct8x8_c,
                   &vp9_idct8x8_64_add_c,
                   &vp9_idct8x8_12_add_msa,
                   TX_8X8, 10),
        make_tuple(&vp9_fdct8x8_c,
                   &vp9_idct8x8_64_add_c,
                   &vp9_idct8x8_1_add_msa,
336
337
338
339
340
                   TX_8X8, 1),
        make_tuple(&vp9_fdct4x4_c,
                   &vp9_idct4x4_16_add_c,
                   &vp9_idct4x4_1_add_msa,
                   TX_4X4, 1)));
341
342
#endif  // HAVE_MSA && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE

343
}  // namespace