2 * Copyright (c) 2012 The WebM project authors. All Rights Reserved.
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.
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"
21 #include "./vp9_rtcd.h"
22 #include "vp9/common/vp9_entropy.h"
23 #include "vpx/vpx_codec.h"
24 #include "vpx/vpx_integer.h"
26 using libvpx_test::ACMRandom;
31 static int round(double x) {
33 return static_cast<int>(ceil(x - 0.5));
35 return static_cast<int>(floor(x + 0.5));
39 const int kNumCoeffs = 256;
40 const double PI = 3.1415926535898;
41 void reference2_16x16_idct_2d(double *input, double *output) {
43 for (int l = 0; l < 16; ++l) {
44 for (int k = 0; k < 16; ++k) {
46 for (int i = 0; i < 16; ++i) {
47 for (int j = 0; j < 16; ++j) {
48 x = cos(PI * j * (l + 0.5) / 16.0) *
49 cos(PI * i * (k + 0.5) / 16.0) *
50 input[i * 16 + j] / 256;
64 const double C1 = 0.995184726672197;
65 const double C2 = 0.98078528040323;
66 const double C3 = 0.956940335732209;
67 const double C4 = 0.923879532511287;
68 const double C5 = 0.881921264348355;
69 const double C6 = 0.831469612302545;
70 const double C7 = 0.773010453362737;
71 const double C8 = 0.707106781186548;
72 const double C9 = 0.634393284163646;
73 const double C10 = 0.555570233019602;
74 const double C11 = 0.471396736825998;
75 const double C12 = 0.38268343236509;
76 const double C13 = 0.290284677254462;
77 const double C14 = 0.195090322016128;
78 const double C15 = 0.098017140329561;
80 void butterfly_16x16_dct_1d(double input[16], double output[16]) {
82 double intermediate[16];
86 step[ 0] = input[0] + input[15];
87 step[ 1] = input[1] + input[14];
88 step[ 2] = input[2] + input[13];
89 step[ 3] = input[3] + input[12];
90 step[ 4] = input[4] + input[11];
91 step[ 5] = input[5] + input[10];
92 step[ 6] = input[6] + input[ 9];
93 step[ 7] = input[7] + input[ 8];
94 step[ 8] = input[7] - input[ 8];
95 step[ 9] = input[6] - input[ 9];
96 step[10] = input[5] - input[10];
97 step[11] = input[4] - input[11];
98 step[12] = input[3] - input[12];
99 step[13] = input[2] - input[13];
100 step[14] = input[1] - input[14];
101 step[15] = input[0] - input[15];
104 output[0] = step[0] + step[7];
105 output[1] = step[1] + step[6];
106 output[2] = step[2] + step[5];
107 output[3] = step[3] + step[4];
108 output[4] = step[3] - step[4];
109 output[5] = step[2] - step[5];
110 output[6] = step[1] - step[6];
111 output[7] = step[0] - step[7];
113 temp1 = step[ 8] * C7;
114 temp2 = step[15] * C9;
115 output[ 8] = temp1 + temp2;
117 temp1 = step[ 9] * C11;
118 temp2 = step[14] * C5;
119 output[ 9] = temp1 - temp2;
121 temp1 = step[10] * C3;
122 temp2 = step[13] * C13;
123 output[10] = temp1 + temp2;
125 temp1 = step[11] * C15;
126 temp2 = step[12] * C1;
127 output[11] = temp1 - temp2;
129 temp1 = step[11] * C1;
130 temp2 = step[12] * C15;
131 output[12] = temp2 + temp1;
133 temp1 = step[10] * C13;
134 temp2 = step[13] * C3;
135 output[13] = temp2 - temp1;
137 temp1 = step[ 9] * C5;
138 temp2 = step[14] * C11;
139 output[14] = temp2 + temp1;
141 temp1 = step[ 8] * C9;
142 temp2 = step[15] * C7;
143 output[15] = temp2 - temp1;
146 step[ 0] = output[0] + output[3];
147 step[ 1] = output[1] + output[2];
148 step[ 2] = output[1] - output[2];
149 step[ 3] = output[0] - output[3];
151 temp1 = output[4] * C14;
152 temp2 = output[7] * C2;
153 step[ 4] = temp1 + temp2;
155 temp1 = output[5] * C10;
156 temp2 = output[6] * C6;
157 step[ 5] = temp1 + temp2;
159 temp1 = output[5] * C6;
160 temp2 = output[6] * C10;
161 step[ 6] = temp2 - temp1;
163 temp1 = output[4] * C2;
164 temp2 = output[7] * C14;
165 step[ 7] = temp2 - temp1;
167 step[ 8] = output[ 8] + output[11];
168 step[ 9] = output[ 9] + output[10];
169 step[10] = output[ 9] - output[10];
170 step[11] = output[ 8] - output[11];
172 step[12] = output[12] + output[15];
173 step[13] = output[13] + output[14];
174 step[14] = output[13] - output[14];
175 step[15] = output[12] - output[15];
178 output[ 0] = (step[ 0] + step[ 1]);
179 output[ 8] = (step[ 0] - step[ 1]);
181 temp1 = step[2] * C12;
182 temp2 = step[3] * C4;
183 temp1 = temp1 + temp2;
184 output[ 4] = 2*(temp1 * C8);
186 temp1 = step[2] * C4;
187 temp2 = step[3] * C12;
188 temp1 = temp2 - temp1;
189 output[12] = 2 * (temp1 * C8);
191 output[ 2] = 2 * ((step[4] + step[ 5]) * C8);
192 output[14] = 2 * ((step[7] - step[ 6]) * C8);
194 temp1 = step[4] - step[5];
195 temp2 = step[6] + step[7];
196 output[ 6] = (temp1 + temp2);
197 output[10] = (temp1 - temp2);
199 intermediate[8] = step[8] + step[14];
200 intermediate[9] = step[9] + step[15];
202 temp1 = intermediate[8] * C12;
203 temp2 = intermediate[9] * C4;
204 temp1 = temp1 - temp2;
205 output[3] = 2 * (temp1 * C8);
207 temp1 = intermediate[8] * C4;
208 temp2 = intermediate[9] * C12;
209 temp1 = temp2 + temp1;
210 output[13] = 2 * (temp1 * C8);
212 output[ 9] = 2 * ((step[10] + step[11]) * C8);
214 intermediate[11] = step[10] - step[11];
215 intermediate[12] = step[12] + step[13];
216 intermediate[13] = step[12] - step[13];
217 intermediate[14] = step[ 8] - step[14];
218 intermediate[15] = step[ 9] - step[15];
220 output[15] = (intermediate[11] + intermediate[12]);
221 output[ 1] = -(intermediate[11] - intermediate[12]);
223 output[ 7] = 2 * (intermediate[13] * C8);
225 temp1 = intermediate[14] * C12;
226 temp2 = intermediate[15] * C4;
227 temp1 = temp1 - temp2;
228 output[11] = -2 * (temp1 * C8);
230 temp1 = intermediate[14] * C4;
231 temp2 = intermediate[15] * C12;
232 temp1 = temp2 + temp1;
233 output[ 5] = 2 * (temp1 * C8);
236 void reference_16x16_dct_2d(int16_t input[256], double output[256]) {
237 // First transform columns
238 for (int i = 0; i < 16; ++i) {
239 double temp_in[16], temp_out[16];
240 for (int j = 0; j < 16; ++j)
241 temp_in[j] = input[j * 16 + i];
242 butterfly_16x16_dct_1d(temp_in, temp_out);
243 for (int j = 0; j < 16; ++j)
244 output[j * 16 + i] = temp_out[j];
246 // Then transform rows
247 for (int i = 0; i < 16; ++i) {
248 double temp_in[16], temp_out[16];
249 for (int j = 0; j < 16; ++j)
250 temp_in[j] = output[j + i * 16];
251 butterfly_16x16_dct_1d(temp_in, temp_out);
252 // Scale by some magic number
253 for (int j = 0; j < 16; ++j)
254 output[j + i * 16] = temp_out[j]/2;
258 typedef void (*FdctFunc)(const int16_t *in, tran_low_t *out, int stride);
259 typedef void (*IdctFunc)(const tran_low_t *in, uint8_t *out, int stride);
260 typedef void (*FhtFunc)(const int16_t *in, tran_low_t *out, int stride,
262 typedef void (*IhtFunc)(const tran_low_t *in, uint8_t *out, int stride,
265 typedef std::tr1::tuple<FdctFunc, IdctFunc, int, vpx_bit_depth_t> Dct16x16Param;
266 typedef std::tr1::tuple<FhtFunc, IhtFunc, int, vpx_bit_depth_t> Ht16x16Param;
268 void fdct16x16_ref(const int16_t *in, tran_low_t *out, int stride,
270 vp9_fdct16x16_c(in, out, stride);
273 void idct16x16_ref(const tran_low_t *in, uint8_t *dest, int stride,
275 vp9_idct16x16_256_add_c(in, dest, stride);
278 void fht16x16_ref(const int16_t *in, tran_low_t *out, int stride,
280 vp9_fht16x16_c(in, out, stride, tx_type);
283 void iht16x16_ref(const tran_low_t *in, uint8_t *dest, int stride,
285 vp9_iht16x16_256_add_c(in, dest, stride, tx_type);
288 #if CONFIG_VP9_HIGHBITDEPTH
289 void idct16x16_10(const tran_low_t *in, uint8_t *out, int stride) {
290 vp9_high_idct16x16_256_add_c(in, out, stride, 10);
293 void idct16x16_12(const tran_low_t *in, uint8_t *out, int stride) {
294 vp9_high_idct16x16_256_add_c(in, out, stride, 12);
297 void idct16x16_10_ref(const tran_low_t *in, uint8_t *out, int stride,
299 idct16x16_10(in, out, stride);
302 void idct16x16_12_ref(const tran_low_t *in, uint8_t *out, int stride,
304 idct16x16_12(in, out, stride);
307 void iht16x16_10(const tran_low_t *in, uint8_t *out, int stride, int tx_type) {
308 vp9_high_iht16x16_256_add_c(in, out, stride, tx_type, 10);
311 void iht16x16_12(const tran_low_t *in, uint8_t *out, int stride, int tx_type) {
312 vp9_high_iht16x16_256_add_c(in, out, stride, tx_type, 12);
316 class Trans16x16TestBase {
318 virtual ~Trans16x16TestBase() {}
321 virtual void RunFwdTxfm(int16_t *in, tran_low_t *out, int stride) = 0;
323 virtual void RunInvTxfm(tran_low_t *out, uint8_t *dst, int stride) = 0;
325 void RunAccuracyCheck() {
326 ACMRandom rnd(ACMRandom::DeterministicSeed());
327 uint32_t max_error = 0;
328 int64_t total_error = 0;
329 const int count_test_block = 10000;
330 for (int i = 0; i < count_test_block; ++i) {
331 DECLARE_ALIGNED_ARRAY(16, int16_t, test_input_block, kNumCoeffs);
332 DECLARE_ALIGNED_ARRAY(16, tran_low_t, test_temp_block, kNumCoeffs);
333 DECLARE_ALIGNED_ARRAY(16, uint8_t, dst, kNumCoeffs);
334 DECLARE_ALIGNED_ARRAY(16, uint8_t, src, kNumCoeffs);
335 #if CONFIG_VP9_HIGHBITDEPTH
336 DECLARE_ALIGNED_ARRAY(16, uint16_t, dst16, kNumCoeffs);
337 DECLARE_ALIGNED_ARRAY(16, uint16_t, src16, kNumCoeffs);
340 // Initialize a test block with input range [-mask_, mask_].
341 for (int j = 0; j < kNumCoeffs; ++j) {
342 if (bit_depth_ == VPX_BITS_8) {
343 src[j] = rnd.Rand8();
344 dst[j] = rnd.Rand8();
345 test_input_block[j] = src[j] - dst[j];
346 #if CONFIG_VP9_HIGHBITDEPTH
348 src16[j] = rnd.Rand16() & mask_;
349 dst16[j] = rnd.Rand16() & mask_;
350 test_input_block[j] = src16[j] - dst16[j];
355 ASM_REGISTER_STATE_CHECK(RunFwdTxfm(test_input_block,
356 test_temp_block, pitch_));
357 if (bit_depth_ == VPX_BITS_8) {
358 ASM_REGISTER_STATE_CHECK(
359 RunInvTxfm(test_temp_block, dst, pitch_));
360 #if CONFIG_VP9_HIGHBITDEPTH
362 ASM_REGISTER_STATE_CHECK(
363 RunInvTxfm(test_temp_block, CONVERT_TO_BYTEPTR(dst16), pitch_));
367 for (int j = 0; j < kNumCoeffs; ++j) {
368 #if CONFIG_VP9_HIGHBITDEPTH
369 const uint32_t diff =
370 bit_depth_ == VPX_BITS_8 ? dst[j] - src[j] : dst16[j] - src16[j];
372 const uint32_t diff = dst[j] - src[j];
374 const uint32_t error = diff * diff;
375 if (max_error < error)
377 total_error += error;
381 EXPECT_GE(1u << 2 * (bit_depth_ - 8), max_error)
382 << "Error: 16x16 FHT/IHT has an individual round trip error > 1";
384 EXPECT_GE(count_test_block << 2 * (bit_depth_ - 8), total_error)
385 << "Error: 16x16 FHT/IHT has average round trip error > 1 per block";
388 void RunCoeffCheck() {
389 ACMRandom rnd(ACMRandom::DeterministicSeed());
390 const int count_test_block = 1000;
391 DECLARE_ALIGNED_ARRAY(16, int16_t, input_block, kNumCoeffs);
392 DECLARE_ALIGNED_ARRAY(16, tran_low_t, output_ref_block, kNumCoeffs);
393 DECLARE_ALIGNED_ARRAY(16, tran_low_t, output_block, kNumCoeffs);
395 for (int i = 0; i < count_test_block; ++i) {
396 // Initialize a test block with input range [-mask_, mask_].
397 for (int j = 0; j < kNumCoeffs; ++j)
398 input_block[j] = (rnd.Rand16() & mask_) - (rnd.Rand16() & mask_);
400 fwd_txfm_ref(input_block, output_ref_block, pitch_, tx_type_);
401 ASM_REGISTER_STATE_CHECK(RunFwdTxfm(input_block, output_block, pitch_));
403 // The minimum quant value is 4.
404 for (int j = 0; j < kNumCoeffs; ++j)
405 EXPECT_EQ(output_block[j], output_ref_block[j]);
410 ACMRandom rnd(ACMRandom::DeterministicSeed());
411 const int count_test_block = 1000;
412 DECLARE_ALIGNED_ARRAY(16, int16_t, input_block, kNumCoeffs);
413 DECLARE_ALIGNED_ARRAY(16, int16_t, input_extreme_block, kNumCoeffs);
414 DECLARE_ALIGNED_ARRAY(16, tran_low_t, output_ref_block, kNumCoeffs);
415 DECLARE_ALIGNED_ARRAY(16, tran_low_t, output_block, kNumCoeffs);
417 for (int i = 0; i < count_test_block; ++i) {
418 // Initialize a test block with input range [-mask_, mask_].
419 for (int j = 0; j < kNumCoeffs; ++j) {
420 input_block[j] = (rnd.Rand16() & mask_) - (rnd.Rand16() & mask_);
421 input_extreme_block[j] = rnd.Rand8() % 2 ? mask_ : -mask_;
424 for (int j = 0; j < kNumCoeffs; ++j)
425 input_extreme_block[j] = mask_;
427 for (int j = 0; j < kNumCoeffs; ++j)
428 input_extreme_block[j] = -mask_;
431 fwd_txfm_ref(input_extreme_block, output_ref_block, pitch_, tx_type_);
432 ASM_REGISTER_STATE_CHECK(RunFwdTxfm(input_extreme_block,
433 output_block, pitch_));
435 // The minimum quant value is 4.
436 for (int j = 0; j < kNumCoeffs; ++j) {
437 EXPECT_EQ(output_block[j], output_ref_block[j]);
438 EXPECT_GE(4 * DCT_MAX_VALUE << (bit_depth_ - 8), abs(output_block[j]))
439 << "Error: 16x16 FDCT has coefficient larger than 4*DCT_MAX_VALUE";
444 void RunQuantCheck(int dc_thred, int ac_thred) {
445 ACMRandom rnd(ACMRandom::DeterministicSeed());
446 const int count_test_block = 1000;
447 DECLARE_ALIGNED_ARRAY(16, int16_t, input_block, kNumCoeffs);
448 DECLARE_ALIGNED_ARRAY(16, int16_t, input_extreme_block, kNumCoeffs);
449 DECLARE_ALIGNED_ARRAY(16, tran_low_t, output_ref_block, kNumCoeffs);
451 DECLARE_ALIGNED_ARRAY(16, uint8_t, dst, kNumCoeffs);
452 DECLARE_ALIGNED_ARRAY(16, uint8_t, ref, kNumCoeffs);
453 #if CONFIG_VP9_HIGHBITDEPTH
454 DECLARE_ALIGNED_ARRAY(16, uint16_t, dst16, kNumCoeffs);
455 DECLARE_ALIGNED_ARRAY(16, uint16_t, ref16, kNumCoeffs);
458 for (int i = 0; i < count_test_block; ++i) {
459 // Initialize a test block with input range [-mask_, mask_].
460 for (int j = 0; j < kNumCoeffs; ++j) {
461 if (bit_depth_ == VPX_BITS_8)
462 input_block[j] = rnd.Rand8() - rnd.Rand8();
464 input_block[j] = (rnd.Rand16() & mask_) - (rnd.Rand16() & mask_);
465 input_extreme_block[j] = rnd.Rand8() % 2 ? mask_ : -mask_;
468 for (int j = 0; j < kNumCoeffs; ++j)
469 input_extreme_block[j] = mask_;
471 for (int j = 0; j < kNumCoeffs; ++j)
472 input_extreme_block[j] = -mask_;
474 fwd_txfm_ref(input_extreme_block, output_ref_block, pitch_, tx_type_);
476 // clear reconstructed pixel buffers
477 vpx_memset(dst, 0, kNumCoeffs * sizeof(uint8_t));
478 vpx_memset(ref, 0, kNumCoeffs * sizeof(uint8_t));
479 #if CONFIG_VP9_HIGHBITDEPTH
480 vpx_memset(dst16, 0, kNumCoeffs * sizeof(uint16_t));
481 vpx_memset(ref16, 0, kNumCoeffs * sizeof(uint16_t));
484 // quantization with maximum allowed step sizes
485 output_ref_block[0] = (output_ref_block[0] / dc_thred) * dc_thred;
486 for (int j = 1; j < kNumCoeffs; ++j)
487 output_ref_block[j] = (output_ref_block[j] / ac_thred) * ac_thred;
488 if (bit_depth_ == VPX_BITS_8) {
489 inv_txfm_ref(output_ref_block, ref, pitch_, tx_type_);
490 ASM_REGISTER_STATE_CHECK(RunInvTxfm(output_ref_block, dst, pitch_));
491 #if CONFIG_VP9_HIGHBITDEPTH
493 inv_txfm_ref(output_ref_block, CONVERT_TO_BYTEPTR(ref16), pitch_,
495 ASM_REGISTER_STATE_CHECK(RunInvTxfm(output_ref_block,
496 CONVERT_TO_BYTEPTR(dst16), pitch_));
499 if (bit_depth_ == VPX_BITS_8) {
500 for (int j = 0; j < kNumCoeffs; ++j)
501 EXPECT_EQ(ref[j], dst[j]);
502 #if CONFIG_VP9_HIGHBITDEPTH
504 for (int j = 0; j < kNumCoeffs; ++j)
505 EXPECT_EQ(ref16[j], dst16[j]);
511 void RunInvAccuracyCheck() {
512 ACMRandom rnd(ACMRandom::DeterministicSeed());
513 const int count_test_block = 1000;
514 DECLARE_ALIGNED_ARRAY(16, int16_t, in, kNumCoeffs);
515 DECLARE_ALIGNED_ARRAY(16, tran_low_t, coeff, kNumCoeffs);
516 DECLARE_ALIGNED_ARRAY(16, uint8_t, dst, kNumCoeffs);
517 DECLARE_ALIGNED_ARRAY(16, uint8_t, src, kNumCoeffs);
518 #if CONFIG_VP9_HIGHBITDEPTH
519 DECLARE_ALIGNED_ARRAY(16, uint16_t, dst16, kNumCoeffs);
520 DECLARE_ALIGNED_ARRAY(16, uint16_t, src16, kNumCoeffs);
523 for (int i = 0; i < count_test_block; ++i) {
524 double out_r[kNumCoeffs];
526 // Initialize a test block with input range [-255, 255].
527 for (int j = 0; j < kNumCoeffs; ++j) {
528 if (bit_depth_ == VPX_BITS_8) {
529 src[j] = rnd.Rand8();
530 dst[j] = rnd.Rand8();
531 in[j] = src[j] - dst[j];
532 #if CONFIG_VP9_HIGHBITDEPTH
534 src16[j] = rnd.Rand16() & mask_;
535 dst16[j] = rnd.Rand16() & mask_;
536 in[j] = src16[j] - dst16[j];
541 reference_16x16_dct_2d(in, out_r);
542 for (int j = 0; j < kNumCoeffs; ++j)
543 coeff[j] = round(out_r[j]);
545 if (bit_depth_ == VPX_BITS_8) {
546 ASM_REGISTER_STATE_CHECK(RunInvTxfm(coeff, dst, 16));
547 #if CONFIG_VP9_HIGHBITDEPTH
549 ASM_REGISTER_STATE_CHECK(RunInvTxfm(coeff, CONVERT_TO_BYTEPTR(dst16),
554 for (int j = 0; j < kNumCoeffs; ++j) {
555 #if CONFIG_VP9_HIGHBITDEPTH
556 const uint32_t diff =
557 bit_depth_ == VPX_BITS_8 ? dst[j] - src[j] : dst16[j] - src16[j];
559 const uint32_t diff = dst[j] - src[j];
561 const uint32_t error = diff * diff;
563 << "Error: 16x16 IDCT has error " << error
564 << " at index " << j;
570 vpx_bit_depth_t bit_depth_;
572 FhtFunc fwd_txfm_ref;
573 IhtFunc inv_txfm_ref;
577 : public Trans16x16TestBase,
578 public ::testing::TestWithParam<Dct16x16Param> {
580 virtual ~Trans16x16DCT() {}
582 virtual void SetUp() {
583 fwd_txfm_ = GET_PARAM(0);
584 inv_txfm_ = GET_PARAM(1);
585 tx_type_ = GET_PARAM(2);
586 bit_depth_ = GET_PARAM(3);
588 fwd_txfm_ref = fdct16x16_ref;
589 inv_txfm_ref = idct16x16_ref;
590 mask_ = (1 << bit_depth_) - 1;
591 #if CONFIG_VP9_HIGHBITDEPTH
592 switch (bit_depth_) {
594 inv_txfm_ref = idct16x16_10_ref;
597 inv_txfm_ref = idct16x16_12_ref;
600 inv_txfm_ref = idct16x16_ref;
604 inv_txfm_ref = idct16x16_ref;
607 virtual void TearDown() { libvpx_test::ClearSystemState(); }
610 void RunFwdTxfm(int16_t *in, tran_low_t *out, int stride) {
611 fwd_txfm_(in, out, stride);
613 void RunInvTxfm(tran_low_t *out, uint8_t *dst, int stride) {
614 inv_txfm_(out, dst, stride);
621 TEST_P(Trans16x16DCT, AccuracyCheck) {
625 TEST_P(Trans16x16DCT, CoeffCheck) {
629 TEST_P(Trans16x16DCT, MemCheck) {
633 TEST_P(Trans16x16DCT, QuantCheck) {
634 // Use maximally allowed quantization step sizes for DC and AC
635 // coefficients respectively.
636 RunQuantCheck(1336, 1828);
639 TEST_P(Trans16x16DCT, InvAccuracyCheck) {
640 RunInvAccuracyCheck();
644 : public Trans16x16TestBase,
645 public ::testing::TestWithParam<Ht16x16Param> {
647 virtual ~Trans16x16HT() {}
649 virtual void SetUp() {
650 fwd_txfm_ = GET_PARAM(0);
651 inv_txfm_ = GET_PARAM(1);
652 tx_type_ = GET_PARAM(2);
653 bit_depth_ = GET_PARAM(3);
655 fwd_txfm_ref = fht16x16_ref;
656 inv_txfm_ref = iht16x16_ref;
657 mask_ = (1 << bit_depth_) - 1;
658 #if CONFIG_VP9_HIGHBITDEPTH
659 switch (bit_depth_) {
661 inv_txfm_ref = iht16x16_10;
664 inv_txfm_ref = iht16x16_12;
667 inv_txfm_ref = iht16x16_ref;
671 inv_txfm_ref = iht16x16_ref;
674 virtual void TearDown() { libvpx_test::ClearSystemState(); }
677 void RunFwdTxfm(int16_t *in, tran_low_t *out, int stride) {
678 fwd_txfm_(in, out, stride, tx_type_);
680 void RunInvTxfm(tran_low_t *out, uint8_t *dst, int stride) {
681 inv_txfm_(out, dst, stride, tx_type_);
688 TEST_P(Trans16x16HT, AccuracyCheck) {
692 TEST_P(Trans16x16HT, CoeffCheck) {
696 TEST_P(Trans16x16HT, MemCheck) {
700 TEST_P(Trans16x16HT, QuantCheck) {
701 // The encoder skips any non-DC intra prediction modes,
702 // when the quantization step size goes beyond 988.
703 RunQuantCheck(549, 988);
706 using std::tr1::make_tuple;
708 #if CONFIG_VP9_HIGHBITDEPTH
709 INSTANTIATE_TEST_CASE_P(
712 make_tuple(&vp9_high_fdct16x16_c, &idct16x16_10, 0, VPX_BITS_10),
713 make_tuple(&vp9_high_fdct16x16_c, &idct16x16_12, 0, VPX_BITS_12),
714 make_tuple(&vp9_fdct16x16_c, &vp9_idct16x16_256_add_c, 0, VPX_BITS_8)));
716 INSTANTIATE_TEST_CASE_P(
719 make_tuple(&vp9_fdct16x16_c, &vp9_idct16x16_256_add_c, 0, VPX_BITS_8)));
722 #if CONFIG_VP9_HIGHBITDEPTH
723 INSTANTIATE_TEST_CASE_P(
726 make_tuple(&vp9_high_fht16x16_c, &iht16x16_10, 0, VPX_BITS_10),
727 make_tuple(&vp9_high_fht16x16_c, &iht16x16_10, 1, VPX_BITS_10),
728 make_tuple(&vp9_high_fht16x16_c, &iht16x16_10, 2, VPX_BITS_10),
729 make_tuple(&vp9_high_fht16x16_c, &iht16x16_10, 3, VPX_BITS_10),
730 make_tuple(&vp9_high_fht16x16_c, &iht16x16_12, 0, VPX_BITS_12),
731 make_tuple(&vp9_high_fht16x16_c, &iht16x16_12, 1, VPX_BITS_12),
732 make_tuple(&vp9_high_fht16x16_c, &iht16x16_12, 2, VPX_BITS_12),
733 make_tuple(&vp9_high_fht16x16_c, &iht16x16_12, 3, VPX_BITS_12),
734 make_tuple(&vp9_fht16x16_c, &vp9_iht16x16_256_add_c, 0, VPX_BITS_8),
735 make_tuple(&vp9_fht16x16_c, &vp9_iht16x16_256_add_c, 1, VPX_BITS_8),
736 make_tuple(&vp9_fht16x16_c, &vp9_iht16x16_256_add_c, 2, VPX_BITS_8),
737 make_tuple(&vp9_fht16x16_c, &vp9_iht16x16_256_add_c, 3, VPX_BITS_8)));
739 INSTANTIATE_TEST_CASE_P(
742 make_tuple(&vp9_fht16x16_c, &vp9_iht16x16_256_add_c, 0, VPX_BITS_8),
743 make_tuple(&vp9_fht16x16_c, &vp9_iht16x16_256_add_c, 1, VPX_BITS_8),
744 make_tuple(&vp9_fht16x16_c, &vp9_iht16x16_256_add_c, 2, VPX_BITS_8),
745 make_tuple(&vp9_fht16x16_c, &vp9_iht16x16_256_add_c, 3, VPX_BITS_8)));
748 #if HAVE_NEON_ASM && !CONFIG_VP9_HIGHBITDEPTH
749 INSTANTIATE_TEST_CASE_P(
752 make_tuple(&vp9_fdct16x16_c,
753 &vp9_idct16x16_256_add_neon, 0, VPX_BITS_8)));
756 #if HAVE_SSE2 && !CONFIG_VP9_HIGHBITDEPTH
757 INSTANTIATE_TEST_CASE_P(
760 make_tuple(&vp9_fdct16x16_sse2,
761 &vp9_idct16x16_256_add_sse2, 0, VPX_BITS_8)));
762 INSTANTIATE_TEST_CASE_P(
765 make_tuple(&vp9_fht16x16_sse2, &vp9_iht16x16_256_add_sse2, 0,
767 make_tuple(&vp9_fht16x16_sse2, &vp9_iht16x16_256_add_sse2, 1,
769 make_tuple(&vp9_fht16x16_sse2, &vp9_iht16x16_256_add_sse2, 2,
771 make_tuple(&vp9_fht16x16_sse2, &vp9_iht16x16_256_add_sse2, 3,
775 #if HAVE_SSSE3 && !CONFIG_VP9_HIGHBITDEPTH
776 INSTANTIATE_TEST_CASE_P(
777 SSSE3, Trans16x16DCT,
779 make_tuple(&vp9_fdct16x16_c, &vp9_idct16x16_256_add_ssse3, 0,