OLD | NEW |
1 /* | 1 /* |
2 * Copyright (c) 2012 The WebM project authors. All Rights Reserved. | 2 * Copyright (c) 2012 The WebM project authors. All Rights Reserved. |
3 * | 3 * |
4 * Use of this source code is governed by a BSD-style license | 4 * Use of this source code is governed by a BSD-style license |
5 * that can be found in the LICENSE file in the root of the source | 5 * that can be found in the LICENSE file in the root of the source |
6 * tree. An additional intellectual property rights grant can be found | 6 * tree. An additional intellectual property rights grant can be found |
7 * in the file PATENTS. All contributing project authors may | 7 * in the file PATENTS. All contributing project authors may |
8 * be found in the AUTHORS file in the root of the source tree. | 8 * be found in the AUTHORS file in the root of the source tree. |
9 */ | 9 */ |
10 | 10 |
(...skipping 159 matching lines...) Expand 10 before | Expand all | Expand 10 after Loading... |
170 | 170 |
171 EXPECT_GE(count_test_block/5, total_error) | 171 EXPECT_GE(count_test_block/5, total_error) |
172 << "Error: 8x8 FDCT/IDCT or FHT/IHT has average roundtrip " | 172 << "Error: 8x8 FDCT/IDCT or FHT/IHT has average roundtrip " |
173 << "error > 1/5 per block"; | 173 << "error > 1/5 per block"; |
174 } | 174 } |
175 | 175 |
176 void RunExtremalCheck() { | 176 void RunExtremalCheck() { |
177 ACMRandom rnd(ACMRandom::DeterministicSeed()); | 177 ACMRandom rnd(ACMRandom::DeterministicSeed()); |
178 int max_error = 0; | 178 int max_error = 0; |
179 int total_error = 0; | 179 int total_error = 0; |
| 180 int total_coeff_error = 0; |
180 const int count_test_block = 100000; | 181 const int count_test_block = 100000; |
181 DECLARE_ALIGNED_ARRAY(16, int16_t, test_input_block, 64); | 182 DECLARE_ALIGNED_ARRAY(16, int16_t, test_input_block, 64); |
182 DECLARE_ALIGNED_ARRAY(16, int16_t, test_temp_block, 64); | 183 DECLARE_ALIGNED_ARRAY(16, int16_t, test_temp_block, 64); |
| 184 DECLARE_ALIGNED_ARRAY(16, int16_t, ref_temp_block, 64); |
183 DECLARE_ALIGNED_ARRAY(16, uint8_t, dst, 64); | 185 DECLARE_ALIGNED_ARRAY(16, uint8_t, dst, 64); |
184 DECLARE_ALIGNED_ARRAY(16, uint8_t, src, 64); | 186 DECLARE_ALIGNED_ARRAY(16, uint8_t, src, 64); |
185 | 187 |
186 for (int i = 0; i < count_test_block; ++i) { | 188 for (int i = 0; i < count_test_block; ++i) { |
187 // Initialize a test block with input range [-255, 255]. | 189 // Initialize a test block with input range [-255, 255]. |
188 for (int j = 0; j < 64; ++j) { | 190 for (int j = 0; j < 64; ++j) { |
189 src[j] = rnd.Rand8() % 2 ? 255 : 0; | 191 if (i == 0) { |
190 dst[j] = src[j] > 0 ? 0 : 255; | 192 src[j] = 255; |
| 193 dst[j] = 0; |
| 194 } else if (i == 1) { |
| 195 src[j] = 0; |
| 196 dst[j] = 255; |
| 197 } else { |
| 198 src[j] = rnd.Rand8() % 2 ? 255 : 0; |
| 199 dst[j] = rnd.Rand8() % 2 ? 255 : 0; |
| 200 } |
| 201 |
191 test_input_block[j] = src[j] - dst[j]; | 202 test_input_block[j] = src[j] - dst[j]; |
192 } | 203 } |
193 | 204 |
194 REGISTER_STATE_CHECK( | 205 REGISTER_STATE_CHECK( |
195 RunFwdTxfm(test_input_block, test_temp_block, pitch_)); | 206 RunFwdTxfm(test_input_block, test_temp_block, pitch_)); |
196 REGISTER_STATE_CHECK( | 207 REGISTER_STATE_CHECK( |
| 208 fwd_txfm_ref(test_input_block, ref_temp_block, pitch_, tx_type_)); |
| 209 REGISTER_STATE_CHECK( |
197 RunInvTxfm(test_temp_block, dst, pitch_)); | 210 RunInvTxfm(test_temp_block, dst, pitch_)); |
198 | 211 |
199 for (int j = 0; j < 64; ++j) { | 212 for (int j = 0; j < 64; ++j) { |
200 const int diff = dst[j] - src[j]; | 213 const int diff = dst[j] - src[j]; |
201 const int error = diff * diff; | 214 const int error = diff * diff; |
202 if (max_error < error) | 215 if (max_error < error) |
203 max_error = error; | 216 max_error = error; |
204 total_error += error; | 217 total_error += error; |
| 218 |
| 219 const int coeff_diff = test_temp_block[j] - ref_temp_block[j]; |
| 220 total_coeff_error += abs(coeff_diff); |
205 } | 221 } |
206 | 222 |
207 EXPECT_GE(1, max_error) | 223 EXPECT_GE(1, max_error) |
208 << "Error: Extremal 8x8 FDCT/IDCT or FHT/IHT has" | 224 << "Error: Extremal 8x8 FDCT/IDCT or FHT/IHT has" |
209 << "an individual roundtrip error > 1"; | 225 << "an individual roundtrip error > 1"; |
210 | 226 |
211 EXPECT_GE(count_test_block/5, total_error) | 227 EXPECT_GE(count_test_block/5, total_error) |
212 << "Error: Extremal 8x8 FDCT/IDCT or FHT/IHT has average" | 228 << "Error: Extremal 8x8 FDCT/IDCT or FHT/IHT has average" |
213 << " roundtrip error > 1/5 per block"; | 229 << " roundtrip error > 1/5 per block"; |
| 230 |
| 231 EXPECT_EQ(0, total_coeff_error) |
| 232 << "Error: Extremal 8x8 FDCT/FHT has" |
| 233 << "overflow issues in the intermediate steps > 1"; |
214 } | 234 } |
215 } | 235 } |
216 | 236 |
217 int pitch_; | 237 int pitch_; |
218 int tx_type_; | 238 int tx_type_; |
219 fht_t fwd_txfm_ref; | 239 fht_t fwd_txfm_ref; |
220 }; | 240 }; |
221 | 241 |
222 class FwdTrans8x8DCT | 242 class FwdTrans8x8DCT |
223 : public FwdTrans8x8TestBase, | 243 : public FwdTrans8x8TestBase, |
(...skipping 116 matching lines...) Expand 10 before | Expand all | Expand 10 after Loading... |
340 make_tuple(&vp9_fht8x8_sse2, &vp9_iht8x8_64_add_sse2, 2), | 360 make_tuple(&vp9_fht8x8_sse2, &vp9_iht8x8_64_add_sse2, 2), |
341 make_tuple(&vp9_fht8x8_sse2, &vp9_iht8x8_64_add_sse2, 3))); | 361 make_tuple(&vp9_fht8x8_sse2, &vp9_iht8x8_64_add_sse2, 3))); |
342 #endif | 362 #endif |
343 | 363 |
344 #if HAVE_SSSE3 && ARCH_X86_64 | 364 #if HAVE_SSSE3 && ARCH_X86_64 |
345 INSTANTIATE_TEST_CASE_P( | 365 INSTANTIATE_TEST_CASE_P( |
346 SSSE3, FwdTrans8x8DCT, | 366 SSSE3, FwdTrans8x8DCT, |
347 ::testing::Values( | 367 ::testing::Values( |
348 make_tuple(&vp9_fdct8x8_ssse3, &vp9_idct8x8_64_add_ssse3, 0))); | 368 make_tuple(&vp9_fdct8x8_ssse3, &vp9_idct8x8_64_add_ssse3, 0))); |
349 #endif | 369 #endif |
| 370 |
| 371 #if HAVE_AVX2 |
| 372 INSTANTIATE_TEST_CASE_P( |
| 373 AVX2, FwdTrans8x8DCT, |
| 374 ::testing::Values( |
| 375 make_tuple(&vp9_fdct8x8_avx2, &vp9_idct8x8_64_add_c, 0))); |
| 376 INSTANTIATE_TEST_CASE_P( |
| 377 AVX2, FwdTrans8x8HT, |
| 378 ::testing::Values( |
| 379 make_tuple(&vp9_fht8x8_avx2, &vp9_iht8x8_64_add_c, 0), |
| 380 make_tuple(&vp9_fht8x8_avx2, &vp9_iht8x8_64_add_c, 1), |
| 381 make_tuple(&vp9_fht8x8_avx2, &vp9_iht8x8_64_add_c, 2), |
| 382 make_tuple(&vp9_fht8x8_avx2, &vp9_iht8x8_64_add_c, 3))); |
| 383 #endif |
350 } // namespace | 384 } // namespace |
OLD | NEW |