1 /*
2 * Copyright (c) 2016, Alliance for Open Media. All rights reserved
3 *
4 * This source code is subject to the terms of the BSD 2 Clause License and
5 * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
6 * was not distributed with this source code in the LICENSE file, you can
7 * obtain it at www.aomedia.org/license/software. If the Alliance for Open
8 * Media Patent License 1.0 was not distributed with this source code in the
9 * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
10 */
11
12 #include <cmath>
13 #include <cstdlib>
14 #include <string>
15
16 #include "third_party/googletest/src/googletest/include/gtest/gtest.h"
17
18 #include "config/aom_config.h"
19 #include "config/aom_dsp_rtcd.h"
20
21 #include "test/acm_random.h"
22 #include "test/clear_system_state.h"
23 #include "test/register_state_check.h"
24 #include "test/util.h"
25 #include "av1/common/av1_loopfilter.h"
26 #include "av1/common/entropy.h"
27 #include "aom/aom_integer.h"
28
29 using libaom_test::ACMRandom;
30
31 namespace {
32 // Horizontally and Vertically need 32x32: 8 Coeffs preceeding filtered section
33 // 16 Coefs within filtered section
34 // 8 Coeffs following filtered section
35 const int kNumCoeffs = 1024;
36
37 const int number_of_iterations = 10000;
38
39 const int kSpeedTestNum = 500000;
40
41 #define LOOP_PARAM \
42 int p, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh
43 #define DUAL_LOOP_PARAM \
44 int p, const uint8_t *blimit0, const uint8_t *limit0, \
45 const uint8_t *thresh0, const uint8_t *blimit1, const uint8_t *limit1, \
46 const uint8_t *thresh1
47
48 typedef void (*loop_op_t)(uint8_t *s, LOOP_PARAM);
49 typedef void (*dual_loop_op_t)(uint8_t *s, DUAL_LOOP_PARAM);
50 typedef void (*hbdloop_op_t)(uint16_t *s, LOOP_PARAM, int bd);
51 typedef void (*hbddual_loop_op_t)(uint16_t *s, DUAL_LOOP_PARAM, int bd);
52
53 typedef ::testing::tuple<hbdloop_op_t, hbdloop_op_t, int> hbdloop_param_t;
54 typedef ::testing::tuple<hbddual_loop_op_t, hbddual_loop_op_t, int>
55 hbddual_loop_param_t;
56 typedef ::testing::tuple<loop_op_t, loop_op_t, int> loop_param_t;
57 typedef ::testing::tuple<dual_loop_op_t, dual_loop_op_t, int> dual_loop_param_t;
58
59 template <typename Pixel_t, int PIXEL_WIDTH_t>
InitInput(Pixel_t * s,Pixel_t * ref_s,ACMRandom * rnd,const uint8_t limit,const int mask,const int32_t p,const int i)60 void InitInput(Pixel_t *s, Pixel_t *ref_s, ACMRandom *rnd, const uint8_t limit,
61 const int mask, const int32_t p, const int i) {
62 uint16_t tmp_s[kNumCoeffs];
63
64 for (int j = 0; j < kNumCoeffs;) {
65 const uint8_t val = rnd->Rand8();
66 if (val & 0x80) { // 50% chance to choose a new value.
67 tmp_s[j] = rnd->Rand16();
68 j++;
69 } else { // 50% chance to repeat previous value in row X times.
70 int k = 0;
71 while (k++ < ((val & 0x1f) + 1) && j < kNumCoeffs) {
72 if (j < 1) {
73 tmp_s[j] = rnd->Rand16();
74 } else if (val & 0x20) { // Increment by a value within the limit.
75 tmp_s[j] = tmp_s[j - 1] + (limit - 1);
76 } else { // Decrement by a value within the limit.
77 tmp_s[j] = tmp_s[j - 1] - (limit - 1);
78 }
79 j++;
80 }
81 }
82 }
83
84 for (int j = 0; j < kNumCoeffs;) {
85 const uint8_t val = rnd->Rand8();
86 if (val & 0x80) {
87 j++;
88 } else { // 50% chance to repeat previous value in column X times.
89 int k = 0;
90 while (k++ < ((val & 0x1f) + 1) && j < kNumCoeffs) {
91 if (j < 1) {
92 tmp_s[j] = rnd->Rand16();
93 } else if (val & 0x20) { // Increment by a value within the limit.
94 tmp_s[(j % 32) * 32 + j / 32] =
95 tmp_s[((j - 1) % 32) * 32 + (j - 1) / 32] + (limit - 1);
96 } else { // Decrement by a value within the limit.
97 tmp_s[(j % 32) * 32 + j / 32] =
98 tmp_s[((j - 1) % 32) * 32 + (j - 1) / 32] - (limit - 1);
99 }
100 j++;
101 }
102 }
103 }
104
105 for (int j = 0; j < kNumCoeffs; j++) {
106 if (i % 2) {
107 s[j] = tmp_s[j] & mask;
108 } else {
109 s[j] = tmp_s[p * (j % p) + j / p] & mask;
110 }
111 ref_s[j] = s[j];
112 }
113 }
114
GetOuterThresh(ACMRandom * rnd)115 uint8_t GetOuterThresh(ACMRandom *rnd) {
116 return static_cast<uint8_t>(rnd->PseudoUniform(3 * MAX_LOOP_FILTER + 5));
117 }
118
GetInnerThresh(ACMRandom * rnd)119 uint8_t GetInnerThresh(ACMRandom *rnd) {
120 return static_cast<uint8_t>(rnd->PseudoUniform(MAX_LOOP_FILTER + 1));
121 }
122
GetHevThresh(ACMRandom * rnd)123 uint8_t GetHevThresh(ACMRandom *rnd) {
124 return static_cast<uint8_t>(rnd->PseudoUniform(MAX_LOOP_FILTER + 1) >> 4);
125 }
126
127 template <typename func_type_t, typename params_t>
128 class LoopTestParam : public ::testing::TestWithParam<params_t> {
129 public:
~LoopTestParam()130 virtual ~LoopTestParam() {}
SetUp()131 virtual void SetUp() {
132 loopfilter_op_ = ::testing::get<0>(this->GetParam());
133 ref_loopfilter_op_ = ::testing::get<1>(this->GetParam());
134 bit_depth_ = ::testing::get<2>(this->GetParam());
135 mask_ = (1 << bit_depth_) - 1;
136 }
137
TearDown()138 virtual void TearDown() { libaom_test::ClearSystemState(); }
139
140 protected:
141 int bit_depth_;
142 int mask_;
143 func_type_t loopfilter_op_;
144 func_type_t ref_loopfilter_op_;
145 };
146
call_filter(uint16_t * s,LOOP_PARAM,int bd,hbdloop_op_t op)147 void call_filter(uint16_t *s, LOOP_PARAM, int bd, hbdloop_op_t op) {
148 op(s, p, blimit, limit, thresh, bd);
149 }
call_filter(uint8_t * s,LOOP_PARAM,int bd,loop_op_t op)150 void call_filter(uint8_t *s, LOOP_PARAM, int bd, loop_op_t op) {
151 (void)bd;
152 op(s, p, blimit, limit, thresh);
153 }
call_dualfilter(uint16_t * s,DUAL_LOOP_PARAM,int bd,hbddual_loop_op_t op)154 void call_dualfilter(uint16_t *s, DUAL_LOOP_PARAM, int bd,
155 hbddual_loop_op_t op) {
156 op(s, p, blimit0, limit0, thresh0, blimit1, limit1, thresh1, bd);
157 }
call_dualfilter(uint8_t * s,DUAL_LOOP_PARAM,int bd,dual_loop_op_t op)158 void call_dualfilter(uint8_t *s, DUAL_LOOP_PARAM, int bd, dual_loop_op_t op) {
159 (void)bd;
160 op(s, p, blimit0, limit0, thresh0, blimit1, limit1, thresh1);
161 };
162
163 typedef LoopTestParam<hbdloop_op_t, hbdloop_param_t> Loop8Test6Param_hbd;
164 typedef LoopTestParam<loop_op_t, loop_param_t> Loop8Test6Param_lbd;
165 typedef LoopTestParam<hbddual_loop_op_t, hbddual_loop_param_t>
166 Loop8Test9Param_hbd;
167 typedef LoopTestParam<dual_loop_op_t, dual_loop_param_t> Loop8Test9Param_lbd;
168
169 #define OPCHECK(a, b) \
170 ACMRandom rnd(ACMRandom::DeterministicSeed()); \
171 const int count_test_block = number_of_iterations; \
172 const int32_t p = kNumCoeffs / 32; \
173 DECLARE_ALIGNED(b, a, s[kNumCoeffs]); \
174 DECLARE_ALIGNED(b, a, ref_s[kNumCoeffs]); \
175 int err_count_total = 0; \
176 int first_failure = -1; \
177 for (int i = 0; i < count_test_block; ++i) { \
178 int err_count = 0; \
179 uint8_t tmp = GetOuterThresh(&rnd); \
180 DECLARE_ALIGNED(16, const uint8_t, \
181 blimit[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp, \
182 tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp }; \
183 tmp = GetInnerThresh(&rnd); \
184 DECLARE_ALIGNED(16, const uint8_t, \
185 limit[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp, \
186 tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp }; \
187 tmp = GetHevThresh(&rnd); \
188 DECLARE_ALIGNED(16, const uint8_t, \
189 thresh[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp, \
190 tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp }; \
191 InitInput<a, b>(s, ref_s, &rnd, *limit, mask_, p, i); \
192 call_filter(ref_s + 8 + p * 8, p, blimit, limit, thresh, bit_depth_, \
193 ref_loopfilter_op_); \
194 ASM_REGISTER_STATE_CHECK(call_filter(s + 8 + p * 8, p, blimit, limit, \
195 thresh, bit_depth_, loopfilter_op_)); \
196 for (int j = 0; j < kNumCoeffs; ++j) { \
197 err_count += ref_s[j] != s[j]; \
198 } \
199 if (err_count && !err_count_total) { \
200 first_failure = i; \
201 } \
202 err_count_total += err_count; \
203 } \
204 EXPECT_EQ(0, err_count_total) \
205 << "Error: Loop8Test6Param, C output doesn't match SIMD " \
206 "loopfilter output. " \
207 << "First failed at test case " << first_failure;
208
TEST_P(Loop8Test6Param_hbd,OperationCheck)209 TEST_P(Loop8Test6Param_hbd, OperationCheck) { OPCHECK(uint16_t, 16); }
TEST_P(Loop8Test6Param_lbd,OperationCheck)210 TEST_P(Loop8Test6Param_lbd, OperationCheck) { OPCHECK(uint8_t, 8); }
211
212 #define VALCHECK(a, b) \
213 ACMRandom rnd(ACMRandom::DeterministicSeed()); \
214 const int count_test_block = number_of_iterations; \
215 DECLARE_ALIGNED(b, a, s[kNumCoeffs]); \
216 DECLARE_ALIGNED(b, a, ref_s[kNumCoeffs]); \
217 int err_count_total = 0; \
218 int first_failure = -1; \
219 for (int i = 0; i < count_test_block; ++i) { \
220 int err_count = 0; \
221 uint8_t tmp = GetOuterThresh(&rnd); \
222 DECLARE_ALIGNED(16, const uint8_t, \
223 blimit[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp, \
224 tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp }; \
225 tmp = GetInnerThresh(&rnd); \
226 DECLARE_ALIGNED(16, const uint8_t, \
227 limit[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp, \
228 tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp }; \
229 tmp = GetHevThresh(&rnd); \
230 DECLARE_ALIGNED(16, const uint8_t, \
231 thresh[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp, \
232 tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp }; \
233 int32_t p = kNumCoeffs / 32; \
234 for (int j = 0; j < kNumCoeffs; ++j) { \
235 s[j] = rnd.Rand16() & mask_; \
236 ref_s[j] = s[j]; \
237 } \
238 call_filter(ref_s + 8 + p * 8, p, blimit, limit, thresh, bit_depth_, \
239 ref_loopfilter_op_); \
240 ASM_REGISTER_STATE_CHECK(call_filter(s + 8 + p * 8, p, blimit, limit, \
241 thresh, bit_depth_, loopfilter_op_)); \
242 for (int j = 0; j < kNumCoeffs; ++j) { \
243 err_count += ref_s[j] != s[j]; \
244 } \
245 if (err_count && !err_count_total) { \
246 first_failure = i; \
247 } \
248 err_count_total += err_count; \
249 } \
250 EXPECT_EQ(0, err_count_total) \
251 << "Error: Loop8Test6Param, C output doesn't match SIMD " \
252 "loopfilter output. " \
253 << "First failed at test case " << first_failure;
254
TEST_P(Loop8Test6Param_hbd,ValueCheck)255 TEST_P(Loop8Test6Param_hbd, ValueCheck) { VALCHECK(uint16_t, 16); }
TEST_P(Loop8Test6Param_lbd,ValueCheck)256 TEST_P(Loop8Test6Param_lbd, ValueCheck) { VALCHECK(uint8_t, 8); }
257
258 #define SPEEDCHECK(a, b) \
259 ACMRandom rnd(ACMRandom::DeterministicSeed()); \
260 const int count_test_block = kSpeedTestNum; \
261 const int32_t bd = bit_depth_; \
262 DECLARE_ALIGNED(b, a, s[kNumCoeffs]); \
263 uint8_t tmp = GetOuterThresh(&rnd); \
264 DECLARE_ALIGNED(16, const uint8_t, \
265 blimit[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp, \
266 tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp }; \
267 tmp = GetInnerThresh(&rnd); \
268 DECLARE_ALIGNED(16, const uint8_t, \
269 limit[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp, \
270 tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp }; \
271 tmp = GetHevThresh(&rnd); \
272 DECLARE_ALIGNED(16, const uint8_t, \
273 thresh[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp, \
274 tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp }; \
275 int32_t p = kNumCoeffs / 32; \
276 for (int j = 0; j < kNumCoeffs; ++j) { \
277 s[j] = rnd.Rand16() & mask_; \
278 } \
279 for (int i = 0; i < count_test_block; ++i) { \
280 call_filter(s + 8 + p * 8, p, blimit, limit, thresh, bd, loopfilter_op_); \
281 }
282
TEST_P(Loop8Test6Param_hbd,DISABLED_Speed)283 TEST_P(Loop8Test6Param_hbd, DISABLED_Speed) { SPEEDCHECK(uint16_t, 16); }
TEST_P(Loop8Test6Param_lbd,DISABLED_Speed)284 TEST_P(Loop8Test6Param_lbd, DISABLED_Speed) { SPEEDCHECK(uint8_t, 8); }
285
286 #define OPCHECKd(a, b) \
287 ACMRandom rnd(ACMRandom::DeterministicSeed()); \
288 const int count_test_block = number_of_iterations; \
289 DECLARE_ALIGNED(b, a, s[kNumCoeffs]); \
290 DECLARE_ALIGNED(b, a, ref_s[kNumCoeffs]); \
291 int err_count_total = 0; \
292 int first_failure = -1; \
293 for (int i = 0; i < count_test_block; ++i) { \
294 int err_count = 0; \
295 uint8_t tmp = GetOuterThresh(&rnd); \
296 DECLARE_ALIGNED(16, const uint8_t, \
297 blimit0[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp, \
298 tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp }; \
299 tmp = GetInnerThresh(&rnd); \
300 DECLARE_ALIGNED(16, const uint8_t, \
301 limit0[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp, \
302 tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp }; \
303 tmp = GetHevThresh(&rnd); \
304 DECLARE_ALIGNED(16, const uint8_t, \
305 thresh0[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp, \
306 tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp }; \
307 tmp = GetOuterThresh(&rnd); \
308 DECLARE_ALIGNED(16, const uint8_t, \
309 blimit1[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp, \
310 tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp }; \
311 tmp = GetInnerThresh(&rnd); \
312 DECLARE_ALIGNED(16, const uint8_t, \
313 limit1[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp, \
314 tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp }; \
315 tmp = GetHevThresh(&rnd); \
316 DECLARE_ALIGNED(16, const uint8_t, \
317 thresh1[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp, \
318 tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp }; \
319 int32_t p = kNumCoeffs / 32; \
320 const uint8_t limit = *limit0 < *limit1 ? *limit0 : *limit1; \
321 InitInput<a, b>(s, ref_s, &rnd, limit, mask_, p, i); \
322 call_dualfilter(ref_s + 8 + p * 8, p, blimit0, limit0, thresh0, blimit1, \
323 limit1, thresh1, bit_depth_, ref_loopfilter_op_); \
324 ASM_REGISTER_STATE_CHECK( \
325 call_dualfilter(s + 8 + p * 8, p, blimit0, limit0, thresh0, blimit1, \
326 limit1, thresh1, bit_depth_, loopfilter_op_)); \
327 for (int j = 0; j < kNumCoeffs; ++j) { \
328 err_count += ref_s[j] != s[j]; \
329 } \
330 if (err_count && !err_count_total) { \
331 first_failure = i; \
332 } \
333 err_count_total += err_count; \
334 } \
335 EXPECT_EQ(0, err_count_total) \
336 << "Error: Loop8Test9Param, C output doesn't match SIMD " \
337 "loopfilter output. " \
338 << "First failed at test case " << first_failure;
339
TEST_P(Loop8Test9Param_hbd,OperationCheck)340 TEST_P(Loop8Test9Param_hbd, OperationCheck) { OPCHECKd(uint16_t, 16); }
TEST_P(Loop8Test9Param_lbd,OperationCheck)341 TEST_P(Loop8Test9Param_lbd, OperationCheck) { OPCHECKd(uint8_t, 8); }
342
343 #define VALCHECKd(a, b) \
344 ACMRandom rnd(ACMRandom::DeterministicSeed()); \
345 const int count_test_block = number_of_iterations; \
346 DECLARE_ALIGNED(b, a, s[kNumCoeffs]); \
347 DECLARE_ALIGNED(b, a, ref_s[kNumCoeffs]); \
348 int err_count_total = 0; \
349 int first_failure = -1; \
350 for (int i = 0; i < count_test_block; ++i) { \
351 int err_count = 0; \
352 uint8_t tmp = GetOuterThresh(&rnd); \
353 DECLARE_ALIGNED(16, const uint8_t, \
354 blimit0[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp, \
355 tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp }; \
356 tmp = GetInnerThresh(&rnd); \
357 DECLARE_ALIGNED(16, const uint8_t, \
358 limit0[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp, \
359 tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp }; \
360 tmp = GetHevThresh(&rnd); \
361 DECLARE_ALIGNED(16, const uint8_t, \
362 thresh0[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp, \
363 tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp }; \
364 tmp = GetOuterThresh(&rnd); \
365 DECLARE_ALIGNED(16, const uint8_t, \
366 blimit1[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp, \
367 tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp }; \
368 tmp = GetInnerThresh(&rnd); \
369 DECLARE_ALIGNED(16, const uint8_t, \
370 limit1[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp, \
371 tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp }; \
372 tmp = GetHevThresh(&rnd); \
373 DECLARE_ALIGNED(16, const uint8_t, \
374 thresh1[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp, \
375 tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp }; \
376 int32_t p = kNumCoeffs / 32; \
377 for (int j = 0; j < kNumCoeffs; ++j) { \
378 s[j] = rnd.Rand16() & mask_; \
379 ref_s[j] = s[j]; \
380 } \
381 call_dualfilter(ref_s + 8 + p * 8, p, blimit0, limit0, thresh0, blimit1, \
382 limit1, thresh1, bit_depth_, ref_loopfilter_op_); \
383 ASM_REGISTER_STATE_CHECK( \
384 call_dualfilter(s + 8 + p * 8, p, blimit0, limit0, thresh0, blimit1, \
385 limit1, thresh1, bit_depth_, loopfilter_op_)); \
386 for (int j = 0; j < kNumCoeffs; ++j) { \
387 err_count += ref_s[j] != s[j]; \
388 } \
389 if (err_count && !err_count_total) { \
390 first_failure = i; \
391 } \
392 err_count_total += err_count; \
393 } \
394 EXPECT_EQ(0, err_count_total) \
395 << "Error: Loop8Test9Param, C output doesn't match SIMD " \
396 "loopfilter output. " \
397 << "First failed at test case " << first_failure;
398
TEST_P(Loop8Test9Param_hbd,ValueCheck)399 TEST_P(Loop8Test9Param_hbd, ValueCheck) { VALCHECKd(uint16_t, 16); }
TEST_P(Loop8Test9Param_lbd,ValueCheck)400 TEST_P(Loop8Test9Param_lbd, ValueCheck) { VALCHECKd(uint8_t, 8); }
401
402 #define SPEEDCHECKd(a, b) \
403 ACMRandom rnd(ACMRandom::DeterministicSeed()); \
404 const int count_test_block = kSpeedTestNum; \
405 DECLARE_ALIGNED(b, a, s[kNumCoeffs]); \
406 uint8_t tmp = GetOuterThresh(&rnd); \
407 DECLARE_ALIGNED(16, const uint8_t, \
408 blimit0[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp, \
409 tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp }; \
410 tmp = GetInnerThresh(&rnd); \
411 DECLARE_ALIGNED(16, const uint8_t, \
412 limit0[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp, \
413 tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp }; \
414 tmp = GetHevThresh(&rnd); \
415 DECLARE_ALIGNED(16, const uint8_t, \
416 thresh0[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp, \
417 tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp }; \
418 tmp = GetOuterThresh(&rnd); \
419 DECLARE_ALIGNED(16, const uint8_t, \
420 blimit1[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp, \
421 tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp }; \
422 tmp = GetInnerThresh(&rnd); \
423 DECLARE_ALIGNED(16, const uint8_t, \
424 limit1[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp, \
425 tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp }; \
426 tmp = GetHevThresh(&rnd); \
427 DECLARE_ALIGNED(16, const uint8_t, \
428 thresh1[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp, \
429 tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp }; \
430 int32_t p = kNumCoeffs / 32; \
431 for (int j = 0; j < kNumCoeffs; ++j) { \
432 s[j] = rnd.Rand16() & mask_; \
433 } \
434 for (int i = 0; i < count_test_block; ++i) { \
435 call_dualfilter(s + 8 + p * 8, p, blimit0, limit0, thresh0, blimit1, \
436 limit1, thresh1, bit_depth_, loopfilter_op_); \
437 }
438
TEST_P(Loop8Test9Param_hbd,DISABLED_Speed)439 TEST_P(Loop8Test9Param_hbd, DISABLED_Speed) { SPEEDCHECKd(uint16_t, 16); }
TEST_P(Loop8Test9Param_lbd,DISABLED_Speed)440 TEST_P(Loop8Test9Param_lbd, DISABLED_Speed) { SPEEDCHECKd(uint8_t, 8); }
441
442 using ::testing::make_tuple;
443
444 #if HAVE_SSE2
445
446 const hbdloop_param_t kHbdLoop8Test6[] = {
447 make_tuple(&aom_highbd_lpf_horizontal_4_sse2, &aom_highbd_lpf_horizontal_4_c,
448 8),
449 make_tuple(&aom_highbd_lpf_vertical_4_sse2, &aom_highbd_lpf_vertical_4_c, 8),
450 make_tuple(&aom_highbd_lpf_horizontal_6_sse2, &aom_highbd_lpf_horizontal_6_c,
451 8),
452 make_tuple(&aom_highbd_lpf_horizontal_8_sse2, &aom_highbd_lpf_horizontal_8_c,
453 8),
454 make_tuple(&aom_highbd_lpf_horizontal_14_sse2,
455 &aom_highbd_lpf_horizontal_14_c, 8),
456 make_tuple(&aom_highbd_lpf_vertical_6_sse2, &aom_highbd_lpf_vertical_6_c, 8),
457 make_tuple(&aom_highbd_lpf_vertical_8_sse2, &aom_highbd_lpf_vertical_8_c, 8),
458
459 make_tuple(&aom_highbd_lpf_vertical_14_sse2, &aom_highbd_lpf_vertical_14_c,
460 8),
461 make_tuple(&aom_highbd_lpf_horizontal_4_sse2, &aom_highbd_lpf_horizontal_4_c,
462 10),
463 make_tuple(&aom_highbd_lpf_vertical_4_sse2, &aom_highbd_lpf_vertical_4_c, 10),
464 make_tuple(&aom_highbd_lpf_horizontal_6_sse2, &aom_highbd_lpf_horizontal_6_c,
465 10),
466 make_tuple(&aom_highbd_lpf_horizontal_8_sse2, &aom_highbd_lpf_horizontal_8_c,
467 10),
468 make_tuple(&aom_highbd_lpf_horizontal_14_sse2,
469 &aom_highbd_lpf_horizontal_14_c, 10),
470 make_tuple(&aom_highbd_lpf_vertical_6_sse2, &aom_highbd_lpf_vertical_6_c, 10),
471 make_tuple(&aom_highbd_lpf_vertical_8_sse2, &aom_highbd_lpf_vertical_8_c, 10),
472 make_tuple(&aom_highbd_lpf_vertical_14_sse2, &aom_highbd_lpf_vertical_14_c,
473 10),
474 make_tuple(&aom_highbd_lpf_horizontal_4_sse2, &aom_highbd_lpf_horizontal_4_c,
475 12),
476 make_tuple(&aom_highbd_lpf_vertical_4_sse2, &aom_highbd_lpf_vertical_4_c, 12),
477 make_tuple(&aom_highbd_lpf_horizontal_6_sse2, &aom_highbd_lpf_horizontal_6_c,
478 12),
479 make_tuple(&aom_highbd_lpf_horizontal_8_sse2, &aom_highbd_lpf_horizontal_8_c,
480 12),
481 make_tuple(&aom_highbd_lpf_horizontal_14_sse2,
482 &aom_highbd_lpf_horizontal_14_c, 12),
483 make_tuple(&aom_highbd_lpf_vertical_14_sse2, &aom_highbd_lpf_vertical_14_c,
484 12),
485 make_tuple(&aom_highbd_lpf_vertical_6_sse2, &aom_highbd_lpf_vertical_6_c, 12),
486 make_tuple(&aom_highbd_lpf_vertical_8_sse2, &aom_highbd_lpf_vertical_8_c, 12)
487 };
488
489 INSTANTIATE_TEST_CASE_P(SSE2, Loop8Test6Param_hbd,
490 ::testing::ValuesIn(kHbdLoop8Test6));
491
492 const loop_param_t kLoop8Test6[] = {
493 make_tuple(&aom_lpf_horizontal_4_sse2, &aom_lpf_horizontal_4_c, 8),
494 make_tuple(&aom_lpf_horizontal_8_sse2, &aom_lpf_horizontal_8_c, 8),
495 make_tuple(&aom_lpf_horizontal_6_sse2, &aom_lpf_horizontal_6_c, 8),
496 make_tuple(&aom_lpf_vertical_6_sse2, &aom_lpf_vertical_6_c, 8),
497 make_tuple(&aom_lpf_horizontal_14_sse2, &aom_lpf_horizontal_14_c, 8),
498 make_tuple(&aom_lpf_vertical_4_sse2, &aom_lpf_vertical_4_c, 8),
499 make_tuple(&aom_lpf_vertical_8_sse2, &aom_lpf_vertical_8_c, 8),
500 make_tuple(&aom_lpf_vertical_14_sse2, &aom_lpf_vertical_14_c, 8),
501 };
502
503 INSTANTIATE_TEST_CASE_P(SSE2, Loop8Test6Param_lbd,
504 ::testing::ValuesIn(kLoop8Test6));
505
506 const dual_loop_param_t kLoop8Test9[] = {
507 make_tuple(&aom_lpf_horizontal_4_dual_sse2, &aom_lpf_horizontal_4_dual_c, 8),
508 make_tuple(&aom_lpf_vertical_4_dual_sse2, &aom_lpf_vertical_4_dual_c, 8),
509 make_tuple(&aom_lpf_horizontal_6_dual_sse2, &aom_lpf_horizontal_6_dual_c, 8),
510 make_tuple(&aom_lpf_vertical_6_dual_sse2, &aom_lpf_vertical_6_dual_c, 8),
511 make_tuple(&aom_lpf_horizontal_8_dual_sse2, &aom_lpf_horizontal_8_dual_c, 8),
512 make_tuple(&aom_lpf_vertical_8_dual_sse2, &aom_lpf_vertical_8_dual_c, 8),
513 make_tuple(&aom_lpf_horizontal_14_dual_sse2, &aom_lpf_horizontal_14_dual_c,
514 8),
515 make_tuple(&aom_lpf_vertical_14_dual_sse2, &aom_lpf_vertical_14_dual_c, 8)
516 };
517
518 INSTANTIATE_TEST_CASE_P(SSE2, Loop8Test9Param_lbd,
519 ::testing::ValuesIn(kLoop8Test9));
520
521 #endif // HAVE_SSE2
522
523 #if HAVE_SSE2
524 const hbddual_loop_param_t kHbdLoop8Test9[] = {
525 make_tuple(&aom_highbd_lpf_horizontal_4_dual_sse2,
526 &aom_highbd_lpf_horizontal_4_dual_c, 8),
527 make_tuple(&aom_highbd_lpf_horizontal_6_dual_sse2,
528 &aom_highbd_lpf_horizontal_6_dual_c, 8),
529 make_tuple(&aom_highbd_lpf_horizontal_8_dual_sse2,
530 &aom_highbd_lpf_horizontal_8_dual_c, 8),
531 make_tuple(&aom_highbd_lpf_horizontal_14_dual_sse2,
532 &aom_highbd_lpf_horizontal_14_dual_c, 8),
533 make_tuple(&aom_highbd_lpf_vertical_4_dual_sse2,
534 &aom_highbd_lpf_vertical_4_dual_c, 8),
535 make_tuple(&aom_highbd_lpf_vertical_6_dual_sse2,
536 &aom_highbd_lpf_vertical_6_dual_c, 8),
537 make_tuple(&aom_highbd_lpf_vertical_8_dual_sse2,
538 &aom_highbd_lpf_vertical_8_dual_c, 8),
539 make_tuple(&aom_highbd_lpf_vertical_14_dual_sse2,
540 &aom_highbd_lpf_vertical_14_dual_c, 8),
541 make_tuple(&aom_highbd_lpf_horizontal_4_dual_sse2,
542 &aom_highbd_lpf_horizontal_4_dual_c, 10),
543 make_tuple(&aom_highbd_lpf_horizontal_6_dual_sse2,
544 &aom_highbd_lpf_horizontal_6_dual_c, 10),
545 make_tuple(&aom_highbd_lpf_horizontal_8_dual_sse2,
546 &aom_highbd_lpf_horizontal_8_dual_c, 10),
547 make_tuple(&aom_highbd_lpf_horizontal_14_dual_sse2,
548 &aom_highbd_lpf_horizontal_14_dual_c, 10),
549 make_tuple(&aom_highbd_lpf_vertical_4_dual_sse2,
550 &aom_highbd_lpf_vertical_4_dual_c, 10),
551 make_tuple(&aom_highbd_lpf_vertical_6_dual_sse2,
552 &aom_highbd_lpf_vertical_6_dual_c, 10),
553 make_tuple(&aom_highbd_lpf_vertical_8_dual_sse2,
554 &aom_highbd_lpf_vertical_8_dual_c, 10),
555 make_tuple(&aom_highbd_lpf_vertical_14_dual_sse2,
556 &aom_highbd_lpf_vertical_14_dual_c, 10),
557 make_tuple(&aom_highbd_lpf_horizontal_4_dual_sse2,
558 &aom_highbd_lpf_horizontal_4_dual_c, 12),
559 make_tuple(&aom_highbd_lpf_horizontal_6_dual_sse2,
560 &aom_highbd_lpf_horizontal_6_dual_c, 12),
561 make_tuple(&aom_highbd_lpf_horizontal_8_dual_sse2,
562 &aom_highbd_lpf_horizontal_8_dual_c, 12),
563 make_tuple(&aom_highbd_lpf_horizontal_14_dual_sse2,
564 &aom_highbd_lpf_horizontal_14_dual_c, 12),
565 make_tuple(&aom_highbd_lpf_vertical_4_dual_sse2,
566 &aom_highbd_lpf_vertical_4_dual_c, 12),
567 make_tuple(&aom_highbd_lpf_vertical_6_dual_sse2,
568 &aom_highbd_lpf_vertical_6_dual_c, 12),
569 make_tuple(&aom_highbd_lpf_vertical_8_dual_sse2,
570 &aom_highbd_lpf_vertical_8_dual_c, 12),
571 make_tuple(&aom_highbd_lpf_vertical_14_dual_sse2,
572 &aom_highbd_lpf_vertical_14_dual_c, 12),
573 };
574
575 INSTANTIATE_TEST_CASE_P(SSE2, Loop8Test9Param_hbd,
576 ::testing::ValuesIn(kHbdLoop8Test9));
577
578 #endif // HAVE_SSE2
579
580 #if HAVE_NEON
581 const loop_param_t kLoop8Test6[] = {
582 make_tuple(&aom_lpf_vertical_14_neon, &aom_lpf_vertical_14_c, 8),
583 make_tuple(&aom_lpf_vertical_8_neon, &aom_lpf_vertical_8_c, 8),
584 make_tuple(&aom_lpf_vertical_6_neon, &aom_lpf_vertical_6_c, 8),
585 make_tuple(&aom_lpf_vertical_4_neon, &aom_lpf_vertical_4_c, 8),
586 make_tuple(&aom_lpf_horizontal_14_neon, &aom_lpf_horizontal_14_c, 8),
587 make_tuple(&aom_lpf_horizontal_8_neon, &aom_lpf_horizontal_8_c, 8),
588 make_tuple(&aom_lpf_horizontal_6_neon, &aom_lpf_horizontal_6_c, 8),
589 make_tuple(&aom_lpf_horizontal_4_neon, &aom_lpf_horizontal_4_c, 8)
590 };
591
592 INSTANTIATE_TEST_CASE_P(NEON, Loop8Test6Param_lbd,
593 ::testing::ValuesIn(kLoop8Test6));
594 #endif // HAVE_NEON
595
596 #if HAVE_AVX2
597 const hbddual_loop_param_t kHbdLoop8Test9Avx2[] = {
598 make_tuple(&aom_highbd_lpf_horizontal_4_dual_avx2,
599 &aom_highbd_lpf_horizontal_4_dual_c, 8),
600 make_tuple(&aom_highbd_lpf_horizontal_4_dual_avx2,
601 &aom_highbd_lpf_horizontal_4_dual_c, 10),
602 make_tuple(&aom_highbd_lpf_horizontal_4_dual_avx2,
603 &aom_highbd_lpf_horizontal_4_dual_c, 12),
604 make_tuple(&aom_highbd_lpf_horizontal_8_dual_avx2,
605 &aom_highbd_lpf_horizontal_8_dual_c, 8),
606 make_tuple(&aom_highbd_lpf_horizontal_8_dual_avx2,
607 &aom_highbd_lpf_horizontal_8_dual_c, 10),
608 make_tuple(&aom_highbd_lpf_horizontal_8_dual_avx2,
609 &aom_highbd_lpf_horizontal_8_dual_c, 12),
610 make_tuple(&aom_highbd_lpf_vertical_4_dual_avx2,
611 &aom_highbd_lpf_vertical_4_dual_c, 8),
612 make_tuple(&aom_highbd_lpf_vertical_4_dual_avx2,
613 &aom_highbd_lpf_vertical_4_dual_c, 10),
614 make_tuple(&aom_highbd_lpf_vertical_4_dual_avx2,
615 &aom_highbd_lpf_vertical_4_dual_c, 12),
616 make_tuple(&aom_highbd_lpf_vertical_8_dual_avx2,
617 &aom_highbd_lpf_vertical_8_dual_c, 8),
618 make_tuple(&aom_highbd_lpf_vertical_8_dual_avx2,
619 &aom_highbd_lpf_vertical_8_dual_c, 10),
620 make_tuple(&aom_highbd_lpf_vertical_8_dual_avx2,
621 &aom_highbd_lpf_vertical_8_dual_c, 12),
622 };
623
624 INSTANTIATE_TEST_CASE_P(AVX2, Loop8Test9Param_hbd,
625 ::testing::ValuesIn(kHbdLoop8Test9Avx2));
626 #endif
627 } // namespace
628