1 /*
2 * Copyright (c) 2012 The WebM project authors. All Rights Reserved.
3 *
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
6 * tree. An additional intellectual property rights grant can be found
7 * in the file PATENTS. All contributing project authors may
8 * be found in the AUTHORS file in the root of the source tree.
9 */
10
11 #include <math.h>
12 #include <stdlib.h>
13 #include <string.h>
14
15 #include "third_party/googletest/src/include/gtest/gtest.h"
16 #include "test/acm_random.h"
17 #include "test/clear_system_state.h"
18 #include "test/register_state_check.h"
19 #include "test/util.h"
20
21 #include "./vp9_rtcd.h"
22 #include "vp9/common/vp9_entropy.h"
23 #include "vpx/vpx_integer.h"
24
25 extern "C" {
26 void vp9_idct16x16_256_add_c(const int16_t *input, uint8_t *output, int pitch);
27 }
28
29 using libvpx_test::ACMRandom;
30
31 namespace {
32
33 #ifdef _MSC_VER
round(double x)34 static int round(double x) {
35 if (x < 0)
36 return static_cast<int>(ceil(x - 0.5));
37 else
38 return static_cast<int>(floor(x + 0.5));
39 }
40 #endif
41
42 const int kNumCoeffs = 256;
43 const double PI = 3.1415926535898;
reference2_16x16_idct_2d(double * input,double * output)44 void reference2_16x16_idct_2d(double *input, double *output) {
45 double x;
46 for (int l = 0; l < 16; ++l) {
47 for (int k = 0; k < 16; ++k) {
48 double s = 0;
49 for (int i = 0; i < 16; ++i) {
50 for (int j = 0; j < 16; ++j) {
51 x = cos(PI * j * (l + 0.5) / 16.0) *
52 cos(PI * i * (k + 0.5) / 16.0) *
53 input[i * 16 + j] / 256;
54 if (i != 0)
55 x *= sqrt(2.0);
56 if (j != 0)
57 x *= sqrt(2.0);
58 s += x;
59 }
60 }
61 output[k*16+l] = s;
62 }
63 }
64 }
65
66
67 const double C1 = 0.995184726672197;
68 const double C2 = 0.98078528040323;
69 const double C3 = 0.956940335732209;
70 const double C4 = 0.923879532511287;
71 const double C5 = 0.881921264348355;
72 const double C6 = 0.831469612302545;
73 const double C7 = 0.773010453362737;
74 const double C8 = 0.707106781186548;
75 const double C9 = 0.634393284163646;
76 const double C10 = 0.555570233019602;
77 const double C11 = 0.471396736825998;
78 const double C12 = 0.38268343236509;
79 const double C13 = 0.290284677254462;
80 const double C14 = 0.195090322016128;
81 const double C15 = 0.098017140329561;
82
butterfly_16x16_dct_1d(double input[16],double output[16])83 void butterfly_16x16_dct_1d(double input[16], double output[16]) {
84 double step[16];
85 double intermediate[16];
86 double temp1, temp2;
87
88 // step 1
89 step[ 0] = input[0] + input[15];
90 step[ 1] = input[1] + input[14];
91 step[ 2] = input[2] + input[13];
92 step[ 3] = input[3] + input[12];
93 step[ 4] = input[4] + input[11];
94 step[ 5] = input[5] + input[10];
95 step[ 6] = input[6] + input[ 9];
96 step[ 7] = input[7] + input[ 8];
97 step[ 8] = input[7] - input[ 8];
98 step[ 9] = input[6] - input[ 9];
99 step[10] = input[5] - input[10];
100 step[11] = input[4] - input[11];
101 step[12] = input[3] - input[12];
102 step[13] = input[2] - input[13];
103 step[14] = input[1] - input[14];
104 step[15] = input[0] - input[15];
105
106 // step 2
107 output[0] = step[0] + step[7];
108 output[1] = step[1] + step[6];
109 output[2] = step[2] + step[5];
110 output[3] = step[3] + step[4];
111 output[4] = step[3] - step[4];
112 output[5] = step[2] - step[5];
113 output[6] = step[1] - step[6];
114 output[7] = step[0] - step[7];
115
116 temp1 = step[ 8] * C7;
117 temp2 = step[15] * C9;
118 output[ 8] = temp1 + temp2;
119
120 temp1 = step[ 9] * C11;
121 temp2 = step[14] * C5;
122 output[ 9] = temp1 - temp2;
123
124 temp1 = step[10] * C3;
125 temp2 = step[13] * C13;
126 output[10] = temp1 + temp2;
127
128 temp1 = step[11] * C15;
129 temp2 = step[12] * C1;
130 output[11] = temp1 - temp2;
131
132 temp1 = step[11] * C1;
133 temp2 = step[12] * C15;
134 output[12] = temp2 + temp1;
135
136 temp1 = step[10] * C13;
137 temp2 = step[13] * C3;
138 output[13] = temp2 - temp1;
139
140 temp1 = step[ 9] * C5;
141 temp2 = step[14] * C11;
142 output[14] = temp2 + temp1;
143
144 temp1 = step[ 8] * C9;
145 temp2 = step[15] * C7;
146 output[15] = temp2 - temp1;
147
148 // step 3
149 step[ 0] = output[0] + output[3];
150 step[ 1] = output[1] + output[2];
151 step[ 2] = output[1] - output[2];
152 step[ 3] = output[0] - output[3];
153
154 temp1 = output[4] * C14;
155 temp2 = output[7] * C2;
156 step[ 4] = temp1 + temp2;
157
158 temp1 = output[5] * C10;
159 temp2 = output[6] * C6;
160 step[ 5] = temp1 + temp2;
161
162 temp1 = output[5] * C6;
163 temp2 = output[6] * C10;
164 step[ 6] = temp2 - temp1;
165
166 temp1 = output[4] * C2;
167 temp2 = output[7] * C14;
168 step[ 7] = temp2 - temp1;
169
170 step[ 8] = output[ 8] + output[11];
171 step[ 9] = output[ 9] + output[10];
172 step[10] = output[ 9] - output[10];
173 step[11] = output[ 8] - output[11];
174
175 step[12] = output[12] + output[15];
176 step[13] = output[13] + output[14];
177 step[14] = output[13] - output[14];
178 step[15] = output[12] - output[15];
179
180 // step 4
181 output[ 0] = (step[ 0] + step[ 1]);
182 output[ 8] = (step[ 0] - step[ 1]);
183
184 temp1 = step[2] * C12;
185 temp2 = step[3] * C4;
186 temp1 = temp1 + temp2;
187 output[ 4] = 2*(temp1 * C8);
188
189 temp1 = step[2] * C4;
190 temp2 = step[3] * C12;
191 temp1 = temp2 - temp1;
192 output[12] = 2 * (temp1 * C8);
193
194 output[ 2] = 2 * ((step[4] + step[ 5]) * C8);
195 output[14] = 2 * ((step[7] - step[ 6]) * C8);
196
197 temp1 = step[4] - step[5];
198 temp2 = step[6] + step[7];
199 output[ 6] = (temp1 + temp2);
200 output[10] = (temp1 - temp2);
201
202 intermediate[8] = step[8] + step[14];
203 intermediate[9] = step[9] + step[15];
204
205 temp1 = intermediate[8] * C12;
206 temp2 = intermediate[9] * C4;
207 temp1 = temp1 - temp2;
208 output[3] = 2 * (temp1 * C8);
209
210 temp1 = intermediate[8] * C4;
211 temp2 = intermediate[9] * C12;
212 temp1 = temp2 + temp1;
213 output[13] = 2 * (temp1 * C8);
214
215 output[ 9] = 2 * ((step[10] + step[11]) * C8);
216
217 intermediate[11] = step[10] - step[11];
218 intermediate[12] = step[12] + step[13];
219 intermediate[13] = step[12] - step[13];
220 intermediate[14] = step[ 8] - step[14];
221 intermediate[15] = step[ 9] - step[15];
222
223 output[15] = (intermediate[11] + intermediate[12]);
224 output[ 1] = -(intermediate[11] - intermediate[12]);
225
226 output[ 7] = 2 * (intermediate[13] * C8);
227
228 temp1 = intermediate[14] * C12;
229 temp2 = intermediate[15] * C4;
230 temp1 = temp1 - temp2;
231 output[11] = -2 * (temp1 * C8);
232
233 temp1 = intermediate[14] * C4;
234 temp2 = intermediate[15] * C12;
235 temp1 = temp2 + temp1;
236 output[ 5] = 2 * (temp1 * C8);
237 }
238
reference_16x16_dct_2d(int16_t input[256],double output[256])239 void reference_16x16_dct_2d(int16_t input[256], double output[256]) {
240 // First transform columns
241 for (int i = 0; i < 16; ++i) {
242 double temp_in[16], temp_out[16];
243 for (int j = 0; j < 16; ++j)
244 temp_in[j] = input[j * 16 + i];
245 butterfly_16x16_dct_1d(temp_in, temp_out);
246 for (int j = 0; j < 16; ++j)
247 output[j * 16 + i] = temp_out[j];
248 }
249 // Then transform rows
250 for (int i = 0; i < 16; ++i) {
251 double temp_in[16], temp_out[16];
252 for (int j = 0; j < 16; ++j)
253 temp_in[j] = output[j + i * 16];
254 butterfly_16x16_dct_1d(temp_in, temp_out);
255 // Scale by some magic number
256 for (int j = 0; j < 16; ++j)
257 output[j + i * 16] = temp_out[j]/2;
258 }
259 }
260
261 typedef void (*FdctFunc)(const int16_t *in, int16_t *out, int stride);
262 typedef void (*IdctFunc)(const int16_t *in, uint8_t *out, int stride);
263 typedef void (*FhtFunc)(const int16_t *in, int16_t *out, int stride,
264 int tx_type);
265 typedef void (*IhtFunc)(const int16_t *in, uint8_t *out, int stride,
266 int tx_type);
267
268 typedef std::tr1::tuple<FdctFunc, IdctFunc, int> Dct16x16Param;
269 typedef std::tr1::tuple<FhtFunc, IhtFunc, int> Ht16x16Param;
270
fdct16x16_ref(const int16_t * in,int16_t * out,int stride,int tx_type)271 void fdct16x16_ref(const int16_t *in, int16_t *out, int stride, int tx_type) {
272 vp9_fdct16x16_c(in, out, stride);
273 }
274
idct16x16_ref(const int16_t * in,uint8_t * dest,int stride,int tx_type)275 void idct16x16_ref(const int16_t *in, uint8_t *dest, int stride, int tx_type) {
276 vp9_idct16x16_256_add_c(in, dest, stride);
277 }
278
fht16x16_ref(const int16_t * in,int16_t * out,int stride,int tx_type)279 void fht16x16_ref(const int16_t *in, int16_t *out, int stride, int tx_type) {
280 vp9_fht16x16_c(in, out, stride, tx_type);
281 }
282
iht16x16_ref(const int16_t * in,uint8_t * dest,int stride,int tx_type)283 void iht16x16_ref(const int16_t *in, uint8_t *dest, int stride, int tx_type) {
284 vp9_iht16x16_256_add_c(in, dest, stride, tx_type);
285 }
286
287 class Trans16x16TestBase {
288 public:
~Trans16x16TestBase()289 virtual ~Trans16x16TestBase() {}
290
291 protected:
292 virtual void RunFwdTxfm(int16_t *in, int16_t *out, int stride) = 0;
293
294 virtual void RunInvTxfm(int16_t *out, uint8_t *dst, int stride) = 0;
295
RunAccuracyCheck()296 void RunAccuracyCheck() {
297 ACMRandom rnd(ACMRandom::DeterministicSeed());
298 uint32_t max_error = 0;
299 int64_t total_error = 0;
300 const int count_test_block = 10000;
301 for (int i = 0; i < count_test_block; ++i) {
302 DECLARE_ALIGNED_ARRAY(16, int16_t, test_input_block, kNumCoeffs);
303 DECLARE_ALIGNED_ARRAY(16, int16_t, test_temp_block, kNumCoeffs);
304 DECLARE_ALIGNED_ARRAY(16, uint8_t, dst, kNumCoeffs);
305 DECLARE_ALIGNED_ARRAY(16, uint8_t, src, kNumCoeffs);
306
307 // Initialize a test block with input range [-255, 255].
308 for (int j = 0; j < kNumCoeffs; ++j) {
309 src[j] = rnd.Rand8();
310 dst[j] = rnd.Rand8();
311 test_input_block[j] = src[j] - dst[j];
312 }
313
314 ASM_REGISTER_STATE_CHECK(RunFwdTxfm(test_input_block,
315 test_temp_block, pitch_));
316 ASM_REGISTER_STATE_CHECK(RunInvTxfm(test_temp_block, dst, pitch_));
317
318 for (int j = 0; j < kNumCoeffs; ++j) {
319 const uint32_t diff = dst[j] - src[j];
320 const uint32_t error = diff * diff;
321 if (max_error < error)
322 max_error = error;
323 total_error += error;
324 }
325 }
326
327 EXPECT_GE(1u, max_error)
328 << "Error: 16x16 FHT/IHT has an individual round trip error > 1";
329
330 EXPECT_GE(count_test_block , total_error)
331 << "Error: 16x16 FHT/IHT has average round trip error > 1 per block";
332 }
333
RunCoeffCheck()334 void RunCoeffCheck() {
335 ACMRandom rnd(ACMRandom::DeterministicSeed());
336 const int count_test_block = 1000;
337 DECLARE_ALIGNED_ARRAY(16, int16_t, input_block, kNumCoeffs);
338 DECLARE_ALIGNED_ARRAY(16, int16_t, output_ref_block, kNumCoeffs);
339 DECLARE_ALIGNED_ARRAY(16, int16_t, output_block, kNumCoeffs);
340
341 for (int i = 0; i < count_test_block; ++i) {
342 // Initialize a test block with input range [-255, 255].
343 for (int j = 0; j < kNumCoeffs; ++j)
344 input_block[j] = rnd.Rand8() - rnd.Rand8();
345
346 fwd_txfm_ref(input_block, output_ref_block, pitch_, tx_type_);
347 ASM_REGISTER_STATE_CHECK(RunFwdTxfm(input_block, output_block, pitch_));
348
349 // The minimum quant value is 4.
350 for (int j = 0; j < kNumCoeffs; ++j)
351 EXPECT_EQ(output_block[j], output_ref_block[j]);
352 }
353 }
354
RunMemCheck()355 void RunMemCheck() {
356 ACMRandom rnd(ACMRandom::DeterministicSeed());
357 const int count_test_block = 1000;
358 DECLARE_ALIGNED_ARRAY(16, int16_t, input_block, kNumCoeffs);
359 DECLARE_ALIGNED_ARRAY(16, int16_t, input_extreme_block, kNumCoeffs);
360 DECLARE_ALIGNED_ARRAY(16, int16_t, output_ref_block, kNumCoeffs);
361 DECLARE_ALIGNED_ARRAY(16, int16_t, output_block, kNumCoeffs);
362
363 for (int i = 0; i < count_test_block; ++i) {
364 // Initialize a test block with input range [-255, 255].
365 for (int j = 0; j < kNumCoeffs; ++j) {
366 input_block[j] = rnd.Rand8() - rnd.Rand8();
367 input_extreme_block[j] = rnd.Rand8() % 2 ? 255 : -255;
368 }
369 if (i == 0) {
370 for (int j = 0; j < kNumCoeffs; ++j)
371 input_extreme_block[j] = 255;
372 } else if (i == 1) {
373 for (int j = 0; j < kNumCoeffs; ++j)
374 input_extreme_block[j] = -255;
375 }
376
377 fwd_txfm_ref(input_extreme_block, output_ref_block, pitch_, tx_type_);
378 ASM_REGISTER_STATE_CHECK(RunFwdTxfm(input_extreme_block,
379 output_block, pitch_));
380
381 // The minimum quant value is 4.
382 for (int j = 0; j < kNumCoeffs; ++j) {
383 EXPECT_EQ(output_block[j], output_ref_block[j]);
384 EXPECT_GE(4 * DCT_MAX_VALUE, abs(output_block[j]))
385 << "Error: 16x16 FDCT has coefficient larger than 4*DCT_MAX_VALUE";
386 }
387 }
388 }
389
RunQuantCheck(int dc_thred,int ac_thred)390 void RunQuantCheck(int dc_thred, int ac_thred) {
391 ACMRandom rnd(ACMRandom::DeterministicSeed());
392 const int count_test_block = 1000;
393 DECLARE_ALIGNED_ARRAY(16, int16_t, input_block, kNumCoeffs);
394 DECLARE_ALIGNED_ARRAY(16, int16_t, input_extreme_block, kNumCoeffs);
395 DECLARE_ALIGNED_ARRAY(16, int16_t, output_ref_block, kNumCoeffs);
396
397 DECLARE_ALIGNED_ARRAY(16, uint8_t, dst, kNumCoeffs);
398 DECLARE_ALIGNED_ARRAY(16, uint8_t, ref, kNumCoeffs);
399
400 for (int i = 0; i < count_test_block; ++i) {
401 // Initialize a test block with input range [-255, 255].
402 for (int j = 0; j < kNumCoeffs; ++j) {
403 input_block[j] = rnd.Rand8() - rnd.Rand8();
404 input_extreme_block[j] = rnd.Rand8() % 2 ? 255 : -255;
405 }
406 if (i == 0)
407 for (int j = 0; j < kNumCoeffs; ++j)
408 input_extreme_block[j] = 255;
409 if (i == 1)
410 for (int j = 0; j < kNumCoeffs; ++j)
411 input_extreme_block[j] = -255;
412
413 fwd_txfm_ref(input_extreme_block, output_ref_block, pitch_, tx_type_);
414
415 // clear reconstructed pixel buffers
416 vpx_memset(dst, 0, kNumCoeffs * sizeof(uint8_t));
417 vpx_memset(ref, 0, kNumCoeffs * sizeof(uint8_t));
418
419 // quantization with maximum allowed step sizes
420 output_ref_block[0] = (output_ref_block[0] / dc_thred) * dc_thred;
421 for (int j = 1; j < kNumCoeffs; ++j)
422 output_ref_block[j] = (output_ref_block[j] / ac_thred) * ac_thred;
423 inv_txfm_ref(output_ref_block, ref, pitch_, tx_type_);
424 ASM_REGISTER_STATE_CHECK(RunInvTxfm(output_ref_block, dst, pitch_));
425
426 for (int j = 0; j < kNumCoeffs; ++j)
427 EXPECT_EQ(ref[j], dst[j]);
428 }
429 }
430
RunInvAccuracyCheck()431 void RunInvAccuracyCheck() {
432 ACMRandom rnd(ACMRandom::DeterministicSeed());
433 const int count_test_block = 1000;
434 DECLARE_ALIGNED_ARRAY(16, int16_t, in, kNumCoeffs);
435 DECLARE_ALIGNED_ARRAY(16, int16_t, coeff, kNumCoeffs);
436 DECLARE_ALIGNED_ARRAY(16, uint8_t, dst, kNumCoeffs);
437 DECLARE_ALIGNED_ARRAY(16, uint8_t, src, kNumCoeffs);
438
439 for (int i = 0; i < count_test_block; ++i) {
440 double out_r[kNumCoeffs];
441
442 // Initialize a test block with input range [-255, 255].
443 for (int j = 0; j < kNumCoeffs; ++j) {
444 src[j] = rnd.Rand8();
445 dst[j] = rnd.Rand8();
446 in[j] = src[j] - dst[j];
447 }
448
449 reference_16x16_dct_2d(in, out_r);
450 for (int j = 0; j < kNumCoeffs; ++j)
451 coeff[j] = round(out_r[j]);
452
453 ASM_REGISTER_STATE_CHECK(RunInvTxfm(coeff, dst, 16));
454
455 for (int j = 0; j < kNumCoeffs; ++j) {
456 const uint32_t diff = dst[j] - src[j];
457 const uint32_t error = diff * diff;
458 EXPECT_GE(1u, error)
459 << "Error: 16x16 IDCT has error " << error
460 << " at index " << j;
461 }
462 }
463 }
464 int pitch_;
465 int tx_type_;
466 FhtFunc fwd_txfm_ref;
467 IhtFunc inv_txfm_ref;
468 };
469
470 class Trans16x16DCT
471 : public Trans16x16TestBase,
472 public ::testing::TestWithParam<Dct16x16Param> {
473 public:
~Trans16x16DCT()474 virtual ~Trans16x16DCT() {}
475
SetUp()476 virtual void SetUp() {
477 fwd_txfm_ = GET_PARAM(0);
478 inv_txfm_ = GET_PARAM(1);
479 tx_type_ = GET_PARAM(2);
480 pitch_ = 16;
481 fwd_txfm_ref = fdct16x16_ref;
482 inv_txfm_ref = idct16x16_ref;
483 }
TearDown()484 virtual void TearDown() { libvpx_test::ClearSystemState(); }
485
486 protected:
RunFwdTxfm(int16_t * in,int16_t * out,int stride)487 void RunFwdTxfm(int16_t *in, int16_t *out, int stride) {
488 fwd_txfm_(in, out, stride);
489 }
RunInvTxfm(int16_t * out,uint8_t * dst,int stride)490 void RunInvTxfm(int16_t *out, uint8_t *dst, int stride) {
491 inv_txfm_(out, dst, stride);
492 }
493
494 FdctFunc fwd_txfm_;
495 IdctFunc inv_txfm_;
496 };
497
TEST_P(Trans16x16DCT,AccuracyCheck)498 TEST_P(Trans16x16DCT, AccuracyCheck) {
499 RunAccuracyCheck();
500 }
501
TEST_P(Trans16x16DCT,CoeffCheck)502 TEST_P(Trans16x16DCT, CoeffCheck) {
503 RunCoeffCheck();
504 }
505
TEST_P(Trans16x16DCT,MemCheck)506 TEST_P(Trans16x16DCT, MemCheck) {
507 RunMemCheck();
508 }
509
TEST_P(Trans16x16DCT,QuantCheck)510 TEST_P(Trans16x16DCT, QuantCheck) {
511 // Use maximally allowed quantization step sizes for DC and AC
512 // coefficients respectively.
513 RunQuantCheck(1336, 1828);
514 }
515
TEST_P(Trans16x16DCT,InvAccuracyCheck)516 TEST_P(Trans16x16DCT, InvAccuracyCheck) {
517 RunInvAccuracyCheck();
518 }
519
520 class Trans16x16HT
521 : public Trans16x16TestBase,
522 public ::testing::TestWithParam<Ht16x16Param> {
523 public:
~Trans16x16HT()524 virtual ~Trans16x16HT() {}
525
SetUp()526 virtual void SetUp() {
527 fwd_txfm_ = GET_PARAM(0);
528 inv_txfm_ = GET_PARAM(1);
529 tx_type_ = GET_PARAM(2);
530 pitch_ = 16;
531 fwd_txfm_ref = fht16x16_ref;
532 inv_txfm_ref = iht16x16_ref;
533 }
TearDown()534 virtual void TearDown() { libvpx_test::ClearSystemState(); }
535
536 protected:
RunFwdTxfm(int16_t * in,int16_t * out,int stride)537 void RunFwdTxfm(int16_t *in, int16_t *out, int stride) {
538 fwd_txfm_(in, out, stride, tx_type_);
539 }
RunInvTxfm(int16_t * out,uint8_t * dst,int stride)540 void RunInvTxfm(int16_t *out, uint8_t *dst, int stride) {
541 inv_txfm_(out, dst, stride, tx_type_);
542 }
543
544 FhtFunc fwd_txfm_;
545 IhtFunc inv_txfm_;
546 };
547
TEST_P(Trans16x16HT,AccuracyCheck)548 TEST_P(Trans16x16HT, AccuracyCheck) {
549 RunAccuracyCheck();
550 }
551
TEST_P(Trans16x16HT,CoeffCheck)552 TEST_P(Trans16x16HT, CoeffCheck) {
553 RunCoeffCheck();
554 }
555
TEST_P(Trans16x16HT,MemCheck)556 TEST_P(Trans16x16HT, MemCheck) {
557 RunMemCheck();
558 }
559
TEST_P(Trans16x16HT,QuantCheck)560 TEST_P(Trans16x16HT, QuantCheck) {
561 // The encoder skips any non-DC intra prediction modes,
562 // when the quantization step size goes beyond 988.
563 RunQuantCheck(549, 988);
564 }
565
566 using std::tr1::make_tuple;
567
568 INSTANTIATE_TEST_CASE_P(
569 C, Trans16x16DCT,
570 ::testing::Values(
571 make_tuple(&vp9_fdct16x16_c, &vp9_idct16x16_256_add_c, 0)));
572 INSTANTIATE_TEST_CASE_P(
573 C, Trans16x16HT,
574 ::testing::Values(
575 make_tuple(&vp9_fht16x16_c, &vp9_iht16x16_256_add_c, 0),
576 make_tuple(&vp9_fht16x16_c, &vp9_iht16x16_256_add_c, 1),
577 make_tuple(&vp9_fht16x16_c, &vp9_iht16x16_256_add_c, 2),
578 make_tuple(&vp9_fht16x16_c, &vp9_iht16x16_256_add_c, 3)));
579
580 #if HAVE_NEON_ASM
581 INSTANTIATE_TEST_CASE_P(
582 NEON, Trans16x16DCT,
583 ::testing::Values(
584 make_tuple(&vp9_fdct16x16_c,
585 &vp9_idct16x16_256_add_neon, 0)));
586 #endif
587
588 #if HAVE_SSE2
589 INSTANTIATE_TEST_CASE_P(
590 SSE2, Trans16x16DCT,
591 ::testing::Values(
592 make_tuple(&vp9_fdct16x16_sse2,
593 &vp9_idct16x16_256_add_sse2, 0)));
594 INSTANTIATE_TEST_CASE_P(
595 SSE2, Trans16x16HT,
596 ::testing::Values(
597 make_tuple(&vp9_fht16x16_sse2, &vp9_iht16x16_256_add_sse2, 0),
598 make_tuple(&vp9_fht16x16_sse2, &vp9_iht16x16_256_add_sse2, 1),
599 make_tuple(&vp9_fht16x16_sse2, &vp9_iht16x16_256_add_sse2, 2),
600 make_tuple(&vp9_fht16x16_sse2, &vp9_iht16x16_256_add_sse2, 3)));
601 #endif
602
603 #if HAVE_SSSE3
604 INSTANTIATE_TEST_CASE_P(
605 SSSE3, Trans16x16DCT,
606 ::testing::Values(
607 make_tuple(&vp9_fdct16x16_c, &vp9_idct16x16_256_add_ssse3, 0)));
608 #endif
609 } // namespace
610