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 <math.h>
13 #include <stdlib.h>
14 #include <string.h>
15 #include <limits>
16
17 #include "third_party/googletest/src/googletest/include/gtest/gtest.h"
18
19 #include "./av1_rtcd.h"
20 #include "./aom_dsp_rtcd.h"
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/blockd.h"
26 #include "av1/common/scan.h"
27 #include "aom/aom_integer.h"
28 #include "aom_ports/aom_timer.h"
29
30 using libaom_test::ACMRandom;
31
32 namespace {
33 typedef void (*FwdTxfmFunc)(const int16_t *in, tran_low_t *out, int stride);
34 typedef void (*InvTxfmFunc)(const tran_low_t *in, uint8_t *out, int stride);
35 typedef void (*InvTxfmWithBdFunc)(const tran_low_t *in, uint8_t *out,
36 int stride, int bd);
37
38 template <InvTxfmFunc fn>
wrapper(const tran_low_t * in,uint8_t * out,int stride,int bd)39 void wrapper(const tran_low_t *in, uint8_t *out, int stride, int bd) {
40 (void)bd;
41 fn(in, out, stride);
42 }
43
44 typedef std::tr1::tuple<FwdTxfmFunc, InvTxfmWithBdFunc, InvTxfmWithBdFunc,
45 TX_SIZE, int, int, int>
46 PartialInvTxfmParam;
47 const int kMaxNumCoeffs = 1024;
48 const int kCountTestBlock = 10000;
49
50 class PartialIDctTest : public ::testing::TestWithParam<PartialInvTxfmParam> {
51 public:
~PartialIDctTest()52 virtual ~PartialIDctTest() {}
SetUp()53 virtual void SetUp() {
54 rnd_.Reset(ACMRandom::DeterministicSeed());
55 ftxfm_ = GET_PARAM(0);
56 full_itxfm_ = GET_PARAM(1);
57 partial_itxfm_ = GET_PARAM(2);
58 tx_size_ = GET_PARAM(3);
59 last_nonzero_ = GET_PARAM(4);
60 bit_depth_ = GET_PARAM(5);
61 pixel_size_ = GET_PARAM(6);
62 mask_ = (1 << bit_depth_) - 1;
63
64 switch (tx_size_) {
65 case TX_4X4: size_ = 4; break;
66 case TX_8X8: size_ = 8; break;
67 case TX_16X16: size_ = 16; break;
68 case TX_32X32: size_ = 32; break;
69 default: FAIL() << "Wrong Size!"; break;
70 }
71
72 // Randomize stride_ to a value less than or equal to 1024
73 stride_ = rnd_(1024) + 1;
74 if (stride_ < size_) {
75 stride_ = size_;
76 }
77 // Align stride_ to 16 if it's bigger than 16.
78 if (stride_ > 16) {
79 stride_ &= ~15;
80 }
81
82 input_block_size_ = size_ * size_;
83 output_block_size_ = size_ * stride_;
84
85 input_block_ = reinterpret_cast<tran_low_t *>(
86 aom_memalign(16, sizeof(*input_block_) * input_block_size_));
87 output_block_ = reinterpret_cast<uint8_t *>(
88 aom_memalign(16, pixel_size_ * output_block_size_));
89 output_block_ref_ = reinterpret_cast<uint8_t *>(
90 aom_memalign(16, pixel_size_ * output_block_size_));
91 }
92
TearDown()93 virtual void TearDown() {
94 aom_free(input_block_);
95 input_block_ = NULL;
96 aom_free(output_block_);
97 output_block_ = NULL;
98 aom_free(output_block_ref_);
99 output_block_ref_ = NULL;
100 libaom_test::ClearSystemState();
101 }
102
InitMem()103 void InitMem() {
104 memset(input_block_, 0, sizeof(*input_block_) * input_block_size_);
105 if (pixel_size_ == 1) {
106 for (int j = 0; j < output_block_size_; ++j) {
107 output_block_[j] = output_block_ref_[j] = rnd_.Rand16() & mask_;
108 }
109 } else {
110 ASSERT_EQ(2, pixel_size_);
111 uint16_t *const output = reinterpret_cast<uint16_t *>(output_block_);
112 uint16_t *const output_ref =
113 reinterpret_cast<uint16_t *>(output_block_ref_);
114 for (int j = 0; j < output_block_size_; ++j) {
115 output[j] = output_ref[j] = rnd_.Rand16() & mask_;
116 }
117 }
118 }
119
InitInput()120 void InitInput() {
121 const int max_coeff = 32766 / 4;
122 int max_energy_leftover = max_coeff * max_coeff;
123 for (int j = 0; j < last_nonzero_; ++j) {
124 int16_t coeff = static_cast<int16_t>(sqrt(1.0 * max_energy_leftover) *
125 (rnd_.Rand16() - 32768) / 65536);
126 max_energy_leftover -= coeff * coeff;
127 if (max_energy_leftover < 0) {
128 max_energy_leftover = 0;
129 coeff = 0;
130 }
131 input_block_[av1_default_scan_orders[tx_size_].scan[j]] = coeff;
132 }
133 }
134
135 protected:
136 int last_nonzero_;
137 TX_SIZE tx_size_;
138 tran_low_t *input_block_;
139 uint8_t *output_block_;
140 uint8_t *output_block_ref_;
141 int size_;
142 int stride_;
143 int pixel_size_;
144 int input_block_size_;
145 int output_block_size_;
146 int bit_depth_;
147 int mask_;
148 FwdTxfmFunc ftxfm_;
149 InvTxfmWithBdFunc full_itxfm_;
150 InvTxfmWithBdFunc partial_itxfm_;
151 ACMRandom rnd_;
152 };
153
TEST_P(PartialIDctTest,RunQuantCheck)154 TEST_P(PartialIDctTest, RunQuantCheck) {
155 DECLARE_ALIGNED(16, int16_t, input_extreme_block[kMaxNumCoeffs]);
156 DECLARE_ALIGNED(16, tran_low_t, output_ref_block[kMaxNumCoeffs]);
157
158 InitMem();
159 for (int i = 0; i < kCountTestBlock; ++i) {
160 // Initialize a test block with input range [-mask_, mask_].
161 if (i == 0) {
162 for (int k = 0; k < input_block_size_; ++k) {
163 input_extreme_block[k] = mask_;
164 }
165 } else if (i == 1) {
166 for (int k = 0; k < input_block_size_; ++k) {
167 input_extreme_block[k] = -mask_;
168 }
169 } else {
170 for (int k = 0; k < input_block_size_; ++k) {
171 input_extreme_block[k] = rnd_.Rand8() % 2 ? mask_ : -mask_;
172 }
173 }
174
175 ftxfm_(input_extreme_block, output_ref_block, size_);
176
177 // quantization with minimum allowed step sizes
178 input_block_[0] = (output_ref_block[0] / 4) * 4;
179 for (int k = 1; k < last_nonzero_; ++k) {
180 const int pos = av1_default_scan_orders[tx_size_].scan[k];
181 input_block_[pos] = (output_ref_block[pos] / 4) * 4;
182 }
183
184 ASM_REGISTER_STATE_CHECK(
185 full_itxfm_(input_block_, output_block_ref_, stride_, bit_depth_));
186 ASM_REGISTER_STATE_CHECK(
187 partial_itxfm_(input_block_, output_block_, stride_, bit_depth_));
188 ASSERT_EQ(0, memcmp(output_block_ref_, output_block_,
189 pixel_size_ * output_block_size_))
190 << "Error: partial inverse transform produces different results";
191 }
192 }
193
TEST_P(PartialIDctTest,ResultsMatch)194 TEST_P(PartialIDctTest, ResultsMatch) {
195 for (int i = 0; i < kCountTestBlock; ++i) {
196 InitMem();
197 InitInput();
198
199 ASM_REGISTER_STATE_CHECK(
200 full_itxfm_(input_block_, output_block_ref_, stride_, bit_depth_));
201 ASM_REGISTER_STATE_CHECK(
202 partial_itxfm_(input_block_, output_block_, stride_, bit_depth_));
203 ASSERT_EQ(0, memcmp(output_block_ref_, output_block_,
204 pixel_size_ * output_block_size_))
205 << "Error: partial inverse transform produces different results";
206 }
207 }
208
TEST_P(PartialIDctTest,AddOutputBlock)209 TEST_P(PartialIDctTest, AddOutputBlock) {
210 for (int i = 0; i < kCountTestBlock; ++i) {
211 InitMem();
212 for (int j = 0; j < last_nonzero_; ++j) {
213 input_block_[av1_default_scan_orders[tx_size_].scan[j]] = 10;
214 }
215
216 ASM_REGISTER_STATE_CHECK(
217 full_itxfm_(input_block_, output_block_ref_, stride_, bit_depth_));
218 ASM_REGISTER_STATE_CHECK(
219 partial_itxfm_(input_block_, output_block_, stride_, bit_depth_));
220 ASSERT_EQ(0, memcmp(output_block_ref_, output_block_,
221 pixel_size_ * output_block_size_))
222 << "Error: Transform results are not correctly added to output.";
223 }
224 }
225
TEST_P(PartialIDctTest,SingleExtremeCoeff)226 TEST_P(PartialIDctTest, SingleExtremeCoeff) {
227 const int16_t max_coeff = INT16_MAX;
228 const int16_t min_coeff = INT16_MIN;
229 for (int i = 0; i < last_nonzero_; ++i) {
230 memset(input_block_, 0, sizeof(*input_block_) * input_block_size_);
231 // Run once for min and once for max.
232 for (int j = 0; j < 2; ++j) {
233 const int coeff = j ? min_coeff : max_coeff;
234
235 memset(output_block_, 0, pixel_size_ * output_block_size_);
236 memset(output_block_ref_, 0, pixel_size_ * output_block_size_);
237 input_block_[av1_default_scan_orders[tx_size_].scan[i]] = coeff;
238
239 ASM_REGISTER_STATE_CHECK(
240 full_itxfm_(input_block_, output_block_ref_, stride_, bit_depth_));
241 ASM_REGISTER_STATE_CHECK(
242 partial_itxfm_(input_block_, output_block_, stride_, bit_depth_));
243 ASSERT_EQ(0, memcmp(output_block_ref_, output_block_,
244 pixel_size_ * output_block_size_))
245 << "Error: Fails with single coeff of " << coeff << " at " << i
246 << ".";
247 }
248 }
249 }
250
TEST_P(PartialIDctTest,DISABLED_Speed)251 TEST_P(PartialIDctTest, DISABLED_Speed) {
252 // Keep runtime stable with transform size.
253 const int kCountSpeedTestBlock = 500000000 / input_block_size_;
254 InitMem();
255 InitInput();
256
257 for (int i = 0; i < kCountSpeedTestBlock; ++i) {
258 ASM_REGISTER_STATE_CHECK(
259 full_itxfm_(input_block_, output_block_ref_, stride_, bit_depth_));
260 }
261 aom_usec_timer timer;
262 aom_usec_timer_start(&timer);
263 for (int i = 0; i < kCountSpeedTestBlock; ++i) {
264 partial_itxfm_(input_block_, output_block_, stride_, bit_depth_);
265 }
266 libaom_test::ClearSystemState();
267 aom_usec_timer_mark(&timer);
268 const int elapsed_time =
269 static_cast<int>(aom_usec_timer_elapsed(&timer) / 1000);
270 printf("idct%dx%d_%d (bitdepth %d) time: %5d ms\n", size_, size_,
271 last_nonzero_, bit_depth_, elapsed_time);
272
273 ASSERT_EQ(0, memcmp(output_block_ref_, output_block_,
274 pixel_size_ * output_block_size_))
275 << "Error: partial inverse transform produces different results";
276 }
277
278 using std::tr1::make_tuple;
279
280 const PartialInvTxfmParam c_partial_idct_tests[] = {
281 make_tuple(&aom_fdct32x32_c, &wrapper<aom_idct32x32_1024_add_c>,
282 &wrapper<aom_idct32x32_1024_add_c>, TX_32X32, 1024, 8, 1),
283 make_tuple(&aom_fdct32x32_c, &wrapper<aom_idct32x32_1024_add_c>,
284 &wrapper<aom_idct32x32_135_add_c>, TX_32X32, 135, 8, 1),
285 make_tuple(&aom_fdct32x32_c, &wrapper<aom_idct32x32_1024_add_c>,
286 &wrapper<aom_idct32x32_34_add_c>, TX_32X32, 34, 8, 1),
287 make_tuple(&aom_fdct32x32_c, &wrapper<aom_idct32x32_1024_add_c>,
288 &wrapper<aom_idct32x32_1_add_c>, TX_32X32, 1, 8, 1),
289 make_tuple(&aom_fdct16x16_c, &wrapper<aom_idct16x16_256_add_c>,
290 &wrapper<aom_idct16x16_256_add_c>, TX_16X16, 256, 8, 1),
291 make_tuple(&aom_fdct16x16_c, &wrapper<aom_idct16x16_256_add_c>,
292 &wrapper<aom_idct16x16_38_add_c>, TX_16X16, 38, 8, 1),
293 make_tuple(&aom_fdct16x16_c, &wrapper<aom_idct16x16_256_add_c>,
294 &wrapper<aom_idct16x16_10_add_c>, TX_16X16, 10, 8, 1),
295 make_tuple(&aom_fdct16x16_c, &wrapper<aom_idct16x16_256_add_c>,
296 &wrapper<aom_idct16x16_1_add_c>, TX_16X16, 1, 8, 1),
297 make_tuple(&aom_fdct8x8_c, &wrapper<aom_idct8x8_64_add_c>,
298 &wrapper<aom_idct8x8_64_add_c>, TX_8X8, 64, 8, 1),
299 make_tuple(&aom_fdct8x8_c, &wrapper<aom_idct8x8_64_add_c>,
300 &wrapper<aom_idct8x8_12_add_c>, TX_8X8, 12, 8, 1),
301 make_tuple(&aom_fdct8x8_c, &wrapper<aom_idct8x8_64_add_c>,
302 &wrapper<aom_idct8x8_1_add_c>, TX_8X8, 1, 8, 1),
303 make_tuple(&aom_fdct4x4_c, &wrapper<aom_idct4x4_16_add_c>,
304 &wrapper<aom_idct4x4_16_add_c>, TX_4X4, 16, 8, 1),
305 make_tuple(&aom_fdct4x4_c, &wrapper<aom_idct4x4_16_add_c>,
306 &wrapper<aom_idct4x4_1_add_c>, TX_4X4, 1, 8, 1)
307 };
308
309 INSTANTIATE_TEST_CASE_P(C, PartialIDctTest,
310 ::testing::ValuesIn(c_partial_idct_tests));
311
312 #if HAVE_NEON && !CONFIG_HIGHBITDEPTH
313 const PartialInvTxfmParam neon_partial_idct_tests[] = {
314 make_tuple(&aom_fdct32x32_c, &wrapper<aom_idct32x32_1024_add_c>,
315 &wrapper<aom_idct32x32_1_add_neon>, TX_32X32, 1, 8, 1),
316 make_tuple(&aom_fdct16x16_c, &wrapper<aom_idct16x16_256_add_c>,
317 &wrapper<aom_idct16x16_10_add_neon>, TX_16X16, 10, 8, 1),
318 make_tuple(&aom_fdct16x16_c, &wrapper<aom_idct16x16_256_add_c>,
319 &wrapper<aom_idct16x16_1_add_neon>, TX_16X16, 1, 8, 1),
320 make_tuple(&aom_fdct8x8_c, &wrapper<aom_idct8x8_64_add_c>,
321 &wrapper<aom_idct8x8_12_add_neon>, TX_8X8, 12, 8, 1),
322 make_tuple(&aom_fdct8x8_c, &wrapper<aom_idct8x8_64_add_c>,
323 &wrapper<aom_idct8x8_1_add_neon>, TX_8X8, 1, 8, 1),
324 make_tuple(&aom_fdct4x4_c, &wrapper<aom_idct4x4_16_add_c>,
325 &wrapper<aom_idct4x4_1_add_neon>, TX_4X4, 1, 8, 1)
326 };
327
328 INSTANTIATE_TEST_CASE_P(NEON, PartialIDctTest,
329 ::testing::ValuesIn(neon_partial_idct_tests));
330 #endif // HAVE_NEON && !CONFIG_HIGHBITDEPTH
331
332 #if HAVE_SSE2
333 const PartialInvTxfmParam sse2_partial_idct_tests[] = {
334 make_tuple(&aom_fdct32x32_c, &wrapper<aom_idct32x32_1024_add_c>,
335 &wrapper<aom_idct32x32_1024_add_sse2>, TX_32X32, 1024, 8, 1),
336 make_tuple(&aom_fdct32x32_c, &wrapper<aom_idct32x32_1024_add_c>,
337 &wrapper<aom_idct32x32_1024_add_sse2>, TX_32X32, 135, 8, 1),
338 make_tuple(&aom_fdct32x32_c, &wrapper<aom_idct32x32_1024_add_c>,
339 &wrapper<aom_idct32x32_34_add_sse2>, TX_32X32, 34, 8, 1),
340 make_tuple(&aom_fdct32x32_c, &wrapper<aom_idct32x32_1024_add_c>,
341 &wrapper<aom_idct32x32_1_add_sse2>, TX_32X32, 1, 8, 1),
342 make_tuple(&aom_fdct16x16_c, &wrapper<aom_idct16x16_256_add_c>,
343 &wrapper<aom_idct16x16_256_add_sse2>, TX_16X16, 256, 8, 1),
344 make_tuple(&aom_fdct16x16_c, &wrapper<aom_idct16x16_256_add_c>,
345 &wrapper<aom_idct16x16_10_add_sse2>, TX_16X16, 10, 8, 1),
346 make_tuple(&aom_fdct16x16_c, &wrapper<aom_idct16x16_256_add_c>,
347 &wrapper<aom_idct16x16_1_add_sse2>, TX_16X16, 1, 8, 1),
348 make_tuple(&aom_fdct8x8_c, &wrapper<aom_idct8x8_64_add_c>,
349 &wrapper<aom_idct8x8_64_add_sse2>, TX_8X8, 64, 8, 1),
350 make_tuple(&aom_fdct8x8_c, &wrapper<aom_idct8x8_64_add_c>,
351 &wrapper<aom_idct8x8_12_add_sse2>, TX_8X8, 12, 8, 1),
352 make_tuple(&aom_fdct8x8_c, &wrapper<aom_idct8x8_64_add_c>,
353 &wrapper<aom_idct8x8_1_add_sse2>, TX_8X8, 1, 8, 1),
354 make_tuple(&aom_fdct4x4_c, &wrapper<aom_idct4x4_16_add_c>,
355 &wrapper<aom_idct4x4_16_add_sse2>, TX_4X4, 16, 8, 1),
356 make_tuple(&aom_fdct4x4_c, &wrapper<aom_idct4x4_16_add_c>,
357 &wrapper<aom_idct4x4_1_add_sse2>, TX_4X4, 1, 8, 1)
358 };
359
360 INSTANTIATE_TEST_CASE_P(SSE2, PartialIDctTest,
361 ::testing::ValuesIn(sse2_partial_idct_tests));
362
363 #endif // HAVE_SSE2
364
365 #if HAVE_SSSE3
366 const PartialInvTxfmParam ssse3_partial_idct_tests[] = {
367 make_tuple(&aom_fdct32x32_c, &wrapper<aom_idct32x32_1024_add_c>,
368 &wrapper<aom_idct32x32_1024_add_ssse3>, TX_32X32, 1024, 8, 1),
369 make_tuple(&aom_fdct32x32_c, &wrapper<aom_idct32x32_1024_add_c>,
370 &wrapper<aom_idct32x32_135_add_ssse3>, TX_32X32, 135, 8, 1),
371 make_tuple(&aom_fdct32x32_c, &wrapper<aom_idct32x32_1024_add_c>,
372 &wrapper<aom_idct32x32_34_add_ssse3>, TX_32X32, 34, 8, 1),
373 make_tuple(&aom_fdct8x8_c, &wrapper<aom_idct8x8_64_add_c>,
374 &wrapper<aom_idct8x8_64_add_ssse3>, TX_8X8, 64, 8, 1),
375 make_tuple(&aom_fdct8x8_c, &wrapper<aom_idct8x8_64_add_c>,
376 &wrapper<aom_idct8x8_12_add_ssse3>, TX_8X8, 12, 8, 1)
377 };
378
379 INSTANTIATE_TEST_CASE_P(SSSE3, PartialIDctTest,
380 ::testing::ValuesIn(ssse3_partial_idct_tests));
381 #endif // HAVE_SSSE3
382
383 #if HAVE_AVX2
384 const PartialInvTxfmParam avx2_partial_idct_tests[] = {
385 make_tuple(&aom_fdct16x16_c, &wrapper<aom_idct16x16_256_add_c>,
386 &wrapper<aom_idct16x16_256_add_avx2>, TX_16X16, 256, 8, 1),
387 make_tuple(&aom_fdct16x16_c, &wrapper<aom_idct16x16_256_add_c>,
388 &wrapper<aom_idct16x16_38_add_avx2>, TX_16X16, 38, 8, 1),
389 make_tuple(&aom_fdct16x16_c, &wrapper<aom_idct16x16_256_add_c>,
390 &wrapper<aom_idct16x16_10_add_avx2>, TX_16X16, 10, 8, 1),
391 make_tuple(&aom_fdct16x16_c, &wrapper<aom_idct16x16_256_add_c>,
392 &wrapper<aom_idct16x16_1_add_avx2>, TX_16X16, 1, 8, 1),
393 make_tuple(&aom_fdct32x32_c, &wrapper<aom_idct32x32_1024_add_c>,
394 &wrapper<aom_idct32x32_1024_add_avx2>, TX_32X32, 1024, 8, 1),
395 make_tuple(&aom_fdct32x32_c, &wrapper<aom_idct32x32_1024_add_c>,
396 &wrapper<aom_idct32x32_135_add_avx2>, TX_32X32, 135, 8, 1),
397 make_tuple(&aom_fdct32x32_c, &wrapper<aom_idct32x32_1024_add_c>,
398 &wrapper<aom_idct32x32_34_add_avx2>, TX_32X32, 34, 8, 1),
399 make_tuple(&aom_fdct32x32_c, &wrapper<aom_idct32x32_1024_add_c>,
400 &wrapper<aom_idct32x32_1_add_avx2>, TX_32X32, 1, 8, 1),
401 };
402
403 INSTANTIATE_TEST_CASE_P(AVX2, PartialIDctTest,
404 ::testing::ValuesIn(avx2_partial_idct_tests));
405 #endif // HAVE_AVX2
406
407 #if HAVE_DSPR2 && !CONFIG_HIGHBITDEPTH
408 const PartialInvTxfmParam dspr2_partial_idct_tests[] = {
409 make_tuple(&aom_fdct32x32_c, &wrapper<aom_idct32x32_1024_add_c>,
410 &wrapper<aom_idct32x32_1024_add_dspr2>, TX_32X32, 1024, 8, 1),
411 make_tuple(&aom_fdct32x32_c, &wrapper<aom_idct32x32_1024_add_c>,
412 &wrapper<aom_idct32x32_1024_add_dspr2>, TX_32X32, 135, 8, 1),
413 make_tuple(&aom_fdct32x32_c, &wrapper<aom_idct32x32_1024_add_c>,
414 &wrapper<aom_idct32x32_34_add_dspr2>, TX_32X32, 34, 8, 1),
415 make_tuple(&aom_fdct32x32_c, &wrapper<aom_idct32x32_1024_add_c>,
416 &wrapper<aom_idct32x32_1_add_dspr2>, TX_32X32, 1, 8, 1),
417 make_tuple(&aom_fdct16x16_c, &wrapper<aom_idct16x16_256_add_c>,
418 &wrapper<aom_idct16x16_256_add_dspr2>, TX_16X16, 256, 8, 1),
419 make_tuple(&aom_fdct16x16_c, &wrapper<aom_idct16x16_256_add_c>,
420 &wrapper<aom_idct16x16_10_add_dspr2>, TX_16X16, 10, 8, 1),
421 make_tuple(&aom_fdct16x16_c, &wrapper<aom_idct16x16_256_add_c>,
422 &wrapper<aom_idct16x16_1_add_dspr2>, TX_16X16, 1, 8, 1),
423 make_tuple(&aom_fdct8x8_c, &wrapper<aom_idct8x8_64_add_c>,
424 &wrapper<aom_idct8x8_64_add_dspr2>, TX_8X8, 64, 8, 1),
425 make_tuple(&aom_fdct8x8_c, &wrapper<aom_idct8x8_64_add_c>,
426 &wrapper<aom_idct8x8_12_add_dspr2>, TX_8X8, 12, 8, 1),
427 make_tuple(&aom_fdct8x8_c, &wrapper<aom_idct8x8_64_add_c>,
428 &wrapper<aom_idct8x8_1_add_dspr2>, TX_8X8, 1, 8, 1),
429 make_tuple(&aom_fdct4x4_c, &wrapper<aom_idct4x4_16_add_c>,
430 &wrapper<aom_idct4x4_16_add_dspr2>, TX_4X4, 16, 8, 1),
431 make_tuple(&aom_fdct4x4_c, &wrapper<aom_idct4x4_16_add_c>,
432 &wrapper<aom_idct4x4_1_add_dspr2>, TX_4X4, 1, 8, 1)
433 };
434
435 INSTANTIATE_TEST_CASE_P(DSPR2, PartialIDctTest,
436 ::testing::ValuesIn(dspr2_partial_idct_tests));
437 #endif // HAVE_DSPR2 && !CONFIG_HIGHBITDEPTH
438
439 #if HAVE_MSA && !CONFIG_HIGHBITDEPTH
440 const PartialInvTxfmParam msa_partial_idct_tests[] = {
441 make_tuple(&aom_fdct32x32_c, &wrapper<aom_idct32x32_1024_add_c>,
442 &wrapper<aom_idct32x32_1024_add_msa>, TX_32X32, 1024, 8, 1),
443 make_tuple(&aom_fdct32x32_c, &wrapper<aom_idct32x32_1024_add_c>,
444 &wrapper<aom_idct32x32_1024_add_msa>, TX_32X32, 135, 8, 1),
445 make_tuple(&aom_fdct32x32_c, &wrapper<aom_idct32x32_1024_add_c>,
446 &wrapper<aom_idct32x32_34_add_msa>, TX_32X32, 34, 8, 1),
447 make_tuple(&aom_fdct32x32_c, &wrapper<aom_idct32x32_1024_add_c>,
448 &wrapper<aom_idct32x32_1_add_msa>, TX_32X32, 1, 8, 1),
449 make_tuple(&aom_fdct16x16_c, &wrapper<aom_idct16x16_256_add_c>,
450 &wrapper<aom_idct16x16_256_add_msa>, TX_16X16, 256, 8, 1),
451 make_tuple(&aom_fdct16x16_c, &wrapper<aom_idct16x16_256_add_c>,
452 &wrapper<aom_idct16x16_10_add_msa>, TX_16X16, 10, 8, 1),
453 make_tuple(&aom_fdct16x16_c, &wrapper<aom_idct16x16_256_add_c>,
454 &wrapper<aom_idct16x16_1_add_msa>, TX_16X16, 1, 8, 1),
455 make_tuple(&aom_fdct8x8_c, &wrapper<aom_idct8x8_64_add_c>,
456 &wrapper<aom_idct8x8_64_add_msa>, TX_8X8, 64, 8, 1),
457 make_tuple(&aom_fdct8x8_c, &wrapper<aom_idct8x8_64_add_c>,
458 &wrapper<aom_idct8x8_12_add_msa>, TX_8X8, 12, 8, 1),
459 make_tuple(&aom_fdct8x8_c, &wrapper<aom_idct8x8_64_add_c>,
460 &wrapper<aom_idct8x8_1_add_msa>, TX_8X8, 1, 8, 1),
461 make_tuple(&aom_fdct4x4_c, &wrapper<aom_idct4x4_16_add_c>,
462 &wrapper<aom_idct4x4_16_add_msa>, TX_4X4, 16, 8, 1),
463 make_tuple(&aom_fdct4x4_c, &wrapper<aom_idct4x4_16_add_c>,
464 &wrapper<aom_idct4x4_1_add_msa>, TX_4X4, 1, 8, 1)
465 };
466
467 INSTANTIATE_TEST_CASE_P(MSA, PartialIDctTest,
468 ::testing::ValuesIn(msa_partial_idct_tests));
469 #endif // HAVE_MSA && !CONFIG_HIGHBITDEPTH
470
471 } // namespace
472