From bbcc64772580c8a979288791afa02d30bc476d2e Mon Sep 17 00:00:00 2001 From: trav90 Date: Fri, 19 Oct 2018 21:52:15 -0500 Subject: Update aom to v1.0.0 Update aom to commit id d14c5bb4f336ef1842046089849dee4a301fbbf0. --- third_party/aom/test/lpf_test.cc | 623 +++++++++++++++++++++++++++++++++++++++ 1 file changed, 623 insertions(+) create mode 100644 third_party/aom/test/lpf_test.cc (limited to 'third_party/aom/test/lpf_test.cc') diff --git a/third_party/aom/test/lpf_test.cc b/third_party/aom/test/lpf_test.cc new file mode 100644 index 000000000..1e2862ac8 --- /dev/null +++ b/third_party/aom/test/lpf_test.cc @@ -0,0 +1,623 @@ +/* + * Copyright (c) 2016, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + */ + +#include +#include +#include + +#include "third_party/googletest/src/googletest/include/gtest/gtest.h" + +#include "config/aom_config.h" +#include "config/aom_dsp_rtcd.h" + +#include "test/acm_random.h" +#include "test/clear_system_state.h" +#include "test/register_state_check.h" +#include "test/util.h" +#include "av1/common/av1_loopfilter.h" +#include "av1/common/entropy.h" +#include "aom/aom_integer.h" + +using libaom_test::ACMRandom; + +namespace { +// Horizontally and Vertically need 32x32: 8 Coeffs preceeding filtered section +// 16 Coefs within filtered section +// 8 Coeffs following filtered section +const int kNumCoeffs = 1024; + +const int number_of_iterations = 10000; + +const int kSpeedTestNum = 500000; + +#define LOOP_PARAM \ + int p, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh +#define DUAL_LOOP_PARAM \ + int p, const uint8_t *blimit0, const uint8_t *limit0, \ + const uint8_t *thresh0, const uint8_t *blimit1, const uint8_t *limit1, \ + const uint8_t *thresh1 + +typedef void (*loop_op_t)(uint8_t *s, LOOP_PARAM); +typedef void (*dual_loop_op_t)(uint8_t *s, DUAL_LOOP_PARAM); +typedef void (*hbdloop_op_t)(uint16_t *s, LOOP_PARAM, int bd); +typedef void (*hbddual_loop_op_t)(uint16_t *s, DUAL_LOOP_PARAM, int bd); + +typedef ::testing::tuple hbdloop_param_t; +typedef ::testing::tuple + hbddual_loop_param_t; +typedef ::testing::tuple loop_param_t; +typedef ::testing::tuple dual_loop_param_t; + +template +void InitInput(Pixel_t *s, Pixel_t *ref_s, ACMRandom *rnd, const uint8_t limit, + const int mask, const int32_t p, const int i) { + uint16_t tmp_s[kNumCoeffs]; + + for (int j = 0; j < kNumCoeffs;) { + const uint8_t val = rnd->Rand8(); + if (val & 0x80) { // 50% chance to choose a new value. + tmp_s[j] = rnd->Rand16(); + j++; + } else { // 50% chance to repeat previous value in row X times. + int k = 0; + while (k++ < ((val & 0x1f) + 1) && j < kNumCoeffs) { + if (j < 1) { + tmp_s[j] = rnd->Rand16(); + } else if (val & 0x20) { // Increment by a value within the limit. + tmp_s[j] = tmp_s[j - 1] + (limit - 1); + } else { // Decrement by a value within the limit. + tmp_s[j] = tmp_s[j - 1] - (limit - 1); + } + j++; + } + } + } + + for (int j = 0; j < kNumCoeffs;) { + const uint8_t val = rnd->Rand8(); + if (val & 0x80) { + j++; + } else { // 50% chance to repeat previous value in column X times. + int k = 0; + while (k++ < ((val & 0x1f) + 1) && j < kNumCoeffs) { + if (j < 1) { + tmp_s[j] = rnd->Rand16(); + } else if (val & 0x20) { // Increment by a value within the limit. + tmp_s[(j % 32) * 32 + j / 32] = + tmp_s[((j - 1) % 32) * 32 + (j - 1) / 32] + (limit - 1); + } else { // Decrement by a value within the limit. + tmp_s[(j % 32) * 32 + j / 32] = + tmp_s[((j - 1) % 32) * 32 + (j - 1) / 32] - (limit - 1); + } + j++; + } + } + } + + for (int j = 0; j < kNumCoeffs; j++) { + if (i % 2) { + s[j] = tmp_s[j] & mask; + } else { + s[j] = tmp_s[p * (j % p) + j / p] & mask; + } + ref_s[j] = s[j]; + } +} + +uint8_t GetOuterThresh(ACMRandom *rnd) { + return static_cast(rnd->PseudoUniform(3 * MAX_LOOP_FILTER + 5)); +} + +uint8_t GetInnerThresh(ACMRandom *rnd) { + return static_cast(rnd->PseudoUniform(MAX_LOOP_FILTER + 1)); +} + +uint8_t GetHevThresh(ACMRandom *rnd) { + return static_cast(rnd->PseudoUniform(MAX_LOOP_FILTER + 1) >> 4); +} + +template +class LoopTestParam : public ::testing::TestWithParam { + public: + virtual ~LoopTestParam() {} + virtual void SetUp() { + loopfilter_op_ = ::testing::get<0>(this->GetParam()); + ref_loopfilter_op_ = ::testing::get<1>(this->GetParam()); + bit_depth_ = ::testing::get<2>(this->GetParam()); + mask_ = (1 << bit_depth_) - 1; + } + + virtual void TearDown() { libaom_test::ClearSystemState(); } + + protected: + int bit_depth_; + int mask_; + func_type_t loopfilter_op_; + func_type_t ref_loopfilter_op_; +}; + +void call_filter(uint16_t *s, LOOP_PARAM, int bd, hbdloop_op_t op) { + op(s, p, blimit, limit, thresh, bd); +} +void call_filter(uint8_t *s, LOOP_PARAM, int bd, loop_op_t op) { + (void)bd; + op(s, p, blimit, limit, thresh); +} +void call_dualfilter(uint16_t *s, DUAL_LOOP_PARAM, int bd, + hbddual_loop_op_t op) { + op(s, p, blimit0, limit0, thresh0, blimit1, limit1, thresh1, bd); +} +void call_dualfilter(uint8_t *s, DUAL_LOOP_PARAM, int bd, dual_loop_op_t op) { + (void)bd; + op(s, p, blimit0, limit0, thresh0, blimit1, limit1, thresh1); +}; + +typedef LoopTestParam Loop8Test6Param_hbd; +typedef LoopTestParam Loop8Test6Param_lbd; +typedef LoopTestParam + Loop8Test9Param_hbd; +typedef LoopTestParam Loop8Test9Param_lbd; + +#define OPCHECK(a, b) \ + ACMRandom rnd(ACMRandom::DeterministicSeed()); \ + const int count_test_block = number_of_iterations; \ + const int32_t p = kNumCoeffs / 32; \ + DECLARE_ALIGNED(b, a, s[kNumCoeffs]); \ + DECLARE_ALIGNED(b, a, ref_s[kNumCoeffs]); \ + int err_count_total = 0; \ + int first_failure = -1; \ + for (int i = 0; i < count_test_block; ++i) { \ + int err_count = 0; \ + uint8_t tmp = GetOuterThresh(&rnd); \ + DECLARE_ALIGNED(16, const uint8_t, \ + blimit[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp, \ + tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp }; \ + tmp = GetInnerThresh(&rnd); \ + DECLARE_ALIGNED(16, const uint8_t, \ + limit[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp, \ + tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp }; \ + tmp = GetHevThresh(&rnd); \ + DECLARE_ALIGNED(16, const uint8_t, \ + thresh[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp, \ + tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp }; \ + InitInput(s, ref_s, &rnd, *limit, mask_, p, i); \ + call_filter(ref_s + 8 + p * 8, p, blimit, limit, thresh, bit_depth_, \ + ref_loopfilter_op_); \ + ASM_REGISTER_STATE_CHECK(call_filter(s + 8 + p * 8, p, blimit, limit, \ + thresh, bit_depth_, loopfilter_op_)); \ + for (int j = 0; j < kNumCoeffs; ++j) { \ + err_count += ref_s[j] != s[j]; \ + } \ + if (err_count && !err_count_total) { \ + first_failure = i; \ + } \ + err_count_total += err_count; \ + } \ + EXPECT_EQ(0, err_count_total) \ + << "Error: Loop8Test6Param, C output doesn't match SIMD " \ + "loopfilter output. " \ + << "First failed at test case " << first_failure; + +TEST_P(Loop8Test6Param_hbd, OperationCheck) { OPCHECK(uint16_t, 16); } +TEST_P(Loop8Test6Param_lbd, OperationCheck) { OPCHECK(uint8_t, 8); } + +#define VALCHECK(a, b) \ + ACMRandom rnd(ACMRandom::DeterministicSeed()); \ + const int count_test_block = number_of_iterations; \ + DECLARE_ALIGNED(b, a, s[kNumCoeffs]); \ + DECLARE_ALIGNED(b, a, ref_s[kNumCoeffs]); \ + int err_count_total = 0; \ + int first_failure = -1; \ + for (int i = 0; i < count_test_block; ++i) { \ + int err_count = 0; \ + uint8_t tmp = GetOuterThresh(&rnd); \ + DECLARE_ALIGNED(16, const uint8_t, \ + blimit[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp, \ + tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp }; \ + tmp = GetInnerThresh(&rnd); \ + DECLARE_ALIGNED(16, const uint8_t, \ + limit[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp, \ + tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp }; \ + tmp = GetHevThresh(&rnd); \ + DECLARE_ALIGNED(16, const uint8_t, \ + thresh[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp, \ + tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp }; \ + int32_t p = kNumCoeffs / 32; \ + for (int j = 0; j < kNumCoeffs; ++j) { \ + s[j] = rnd.Rand16() & mask_; \ + ref_s[j] = s[j]; \ + } \ + call_filter(ref_s + 8 + p * 8, p, blimit, limit, thresh, bit_depth_, \ + ref_loopfilter_op_); \ + ASM_REGISTER_STATE_CHECK(call_filter(s + 8 + p * 8, p, blimit, limit, \ + thresh, bit_depth_, loopfilter_op_)); \ + for (int j = 0; j < kNumCoeffs; ++j) { \ + err_count += ref_s[j] != s[j]; \ + } \ + if (err_count && !err_count_total) { \ + first_failure = i; \ + } \ + err_count_total += err_count; \ + } \ + EXPECT_EQ(0, err_count_total) \ + << "Error: Loop8Test6Param, C output doesn't match SIMD " \ + "loopfilter output. " \ + << "First failed at test case " << first_failure; + +TEST_P(Loop8Test6Param_hbd, ValueCheck) { VALCHECK(uint16_t, 16); } +TEST_P(Loop8Test6Param_lbd, ValueCheck) { VALCHECK(uint8_t, 8); } + +#define SPEEDCHECK(a, b) \ + ACMRandom rnd(ACMRandom::DeterministicSeed()); \ + const int count_test_block = kSpeedTestNum; \ + const int32_t bd = bit_depth_; \ + DECLARE_ALIGNED(b, a, s[kNumCoeffs]); \ + uint8_t tmp = GetOuterThresh(&rnd); \ + DECLARE_ALIGNED(16, const uint8_t, \ + blimit[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp, \ + tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp }; \ + tmp = GetInnerThresh(&rnd); \ + DECLARE_ALIGNED(16, const uint8_t, \ + limit[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp, \ + tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp }; \ + tmp = GetHevThresh(&rnd); \ + DECLARE_ALIGNED(16, const uint8_t, \ + thresh[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp, \ + tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp }; \ + int32_t p = kNumCoeffs / 32; \ + for (int j = 0; j < kNumCoeffs; ++j) { \ + s[j] = rnd.Rand16() & mask_; \ + } \ + for (int i = 0; i < count_test_block; ++i) { \ + call_filter(s + 8 + p * 8, p, blimit, limit, thresh, bd, loopfilter_op_); \ + } + +TEST_P(Loop8Test6Param_hbd, DISABLED_Speed) { SPEEDCHECK(uint16_t, 16); } +TEST_P(Loop8Test6Param_lbd, DISABLED_Speed) { SPEEDCHECK(uint8_t, 8); } + +#define OPCHECKd(a, b) \ + ACMRandom rnd(ACMRandom::DeterministicSeed()); \ + const int count_test_block = number_of_iterations; \ + DECLARE_ALIGNED(b, a, s[kNumCoeffs]); \ + DECLARE_ALIGNED(b, a, ref_s[kNumCoeffs]); \ + int err_count_total = 0; \ + int first_failure = -1; \ + for (int i = 0; i < count_test_block; ++i) { \ + int err_count = 0; \ + uint8_t tmp = GetOuterThresh(&rnd); \ + DECLARE_ALIGNED(16, const uint8_t, \ + blimit0[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp, \ + tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp }; \ + tmp = GetInnerThresh(&rnd); \ + DECLARE_ALIGNED(16, const uint8_t, \ + limit0[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp, \ + tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp }; \ + tmp = GetHevThresh(&rnd); \ + DECLARE_ALIGNED(16, const uint8_t, \ + thresh0[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp, \ + tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp }; \ + tmp = GetOuterThresh(&rnd); \ + DECLARE_ALIGNED(16, const uint8_t, \ + blimit1[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp, \ + tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp }; \ + tmp = GetInnerThresh(&rnd); \ + DECLARE_ALIGNED(16, const uint8_t, \ + limit1[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp, \ + tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp }; \ + tmp = GetHevThresh(&rnd); \ + DECLARE_ALIGNED(16, const uint8_t, \ + thresh1[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp, \ + tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp }; \ + int32_t p = kNumCoeffs / 32; \ + const uint8_t limit = *limit0 < *limit1 ? *limit0 : *limit1; \ + InitInput(s, ref_s, &rnd, limit, mask_, p, i); \ + call_dualfilter(ref_s + 8 + p * 8, p, blimit0, limit0, thresh0, blimit1, \ + limit1, thresh1, bit_depth_, ref_loopfilter_op_); \ + ASM_REGISTER_STATE_CHECK( \ + call_dualfilter(s + 8 + p * 8, p, blimit0, limit0, thresh0, blimit1, \ + limit1, thresh1, bit_depth_, loopfilter_op_)); \ + for (int j = 0; j < kNumCoeffs; ++j) { \ + err_count += ref_s[j] != s[j]; \ + } \ + if (err_count && !err_count_total) { \ + first_failure = i; \ + } \ + err_count_total += err_count; \ + } \ + EXPECT_EQ(0, err_count_total) \ + << "Error: Loop8Test9Param, C output doesn't match SIMD " \ + "loopfilter output. " \ + << "First failed at test case " << first_failure; + +TEST_P(Loop8Test9Param_hbd, OperationCheck) { OPCHECKd(uint16_t, 16); } +TEST_P(Loop8Test9Param_lbd, OperationCheck) { OPCHECKd(uint8_t, 8); } + +#define VALCHECKd(a, b) \ + ACMRandom rnd(ACMRandom::DeterministicSeed()); \ + const int count_test_block = number_of_iterations; \ + DECLARE_ALIGNED(b, a, s[kNumCoeffs]); \ + DECLARE_ALIGNED(b, a, ref_s[kNumCoeffs]); \ + int err_count_total = 0; \ + int first_failure = -1; \ + for (int i = 0; i < count_test_block; ++i) { \ + int err_count = 0; \ + uint8_t tmp = GetOuterThresh(&rnd); \ + DECLARE_ALIGNED(16, const uint8_t, \ + blimit0[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp, \ + tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp }; \ + tmp = GetInnerThresh(&rnd); \ + DECLARE_ALIGNED(16, const uint8_t, \ + limit0[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp, \ + tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp }; \ + tmp = GetHevThresh(&rnd); \ + DECLARE_ALIGNED(16, const uint8_t, \ + thresh0[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp, \ + tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp }; \ + tmp = GetOuterThresh(&rnd); \ + DECLARE_ALIGNED(16, const uint8_t, \ + blimit1[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp, \ + tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp }; \ + tmp = GetInnerThresh(&rnd); \ + DECLARE_ALIGNED(16, const uint8_t, \ + limit1[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp, \ + tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp }; \ + tmp = GetHevThresh(&rnd); \ + DECLARE_ALIGNED(16, const uint8_t, \ + thresh1[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp, \ + tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp }; \ + int32_t p = kNumCoeffs / 32; \ + for (int j = 0; j < kNumCoeffs; ++j) { \ + s[j] = rnd.Rand16() & mask_; \ + ref_s[j] = s[j]; \ + } \ + call_dualfilter(ref_s + 8 + p * 8, p, blimit0, limit0, thresh0, blimit1, \ + limit1, thresh1, bit_depth_, ref_loopfilter_op_); \ + ASM_REGISTER_STATE_CHECK( \ + call_dualfilter(s + 8 + p * 8, p, blimit0, limit0, thresh0, blimit1, \ + limit1, thresh1, bit_depth_, loopfilter_op_)); \ + for (int j = 0; j < kNumCoeffs; ++j) { \ + err_count += ref_s[j] != s[j]; \ + } \ + if (err_count && !err_count_total) { \ + first_failure = i; \ + } \ + err_count_total += err_count; \ + } \ + EXPECT_EQ(0, err_count_total) \ + << "Error: Loop8Test9Param, C output doesn't match SIMD " \ + "loopfilter output. " \ + << "First failed at test case " << first_failure; + +TEST_P(Loop8Test9Param_hbd, ValueCheck) { VALCHECKd(uint16_t, 16); } +TEST_P(Loop8Test9Param_lbd, ValueCheck) { VALCHECKd(uint8_t, 8); } + +#define SPEEDCHECKd(a, b) \ + ACMRandom rnd(ACMRandom::DeterministicSeed()); \ + const int count_test_block = kSpeedTestNum; \ + DECLARE_ALIGNED(b, a, s[kNumCoeffs]); \ + uint8_t tmp = GetOuterThresh(&rnd); \ + DECLARE_ALIGNED(16, const uint8_t, \ + blimit0[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp, \ + tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp }; \ + tmp = GetInnerThresh(&rnd); \ + DECLARE_ALIGNED(16, const uint8_t, \ + limit0[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp, \ + tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp }; \ + tmp = GetHevThresh(&rnd); \ + DECLARE_ALIGNED(16, const uint8_t, \ + thresh0[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp, \ + tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp }; \ + tmp = GetOuterThresh(&rnd); \ + DECLARE_ALIGNED(16, const uint8_t, \ + blimit1[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp, \ + tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp }; \ + tmp = GetInnerThresh(&rnd); \ + DECLARE_ALIGNED(16, const uint8_t, \ + limit1[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp, \ + tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp }; \ + tmp = GetHevThresh(&rnd); \ + DECLARE_ALIGNED(16, const uint8_t, \ + thresh1[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp, \ + tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp }; \ + int32_t p = kNumCoeffs / 32; \ + for (int j = 0; j < kNumCoeffs; ++j) { \ + s[j] = rnd.Rand16() & mask_; \ + } \ + for (int i = 0; i < count_test_block; ++i) { \ + call_dualfilter(s + 8 + p * 8, p, blimit0, limit0, thresh0, blimit1, \ + limit1, thresh1, bit_depth_, loopfilter_op_); \ + } + +TEST_P(Loop8Test9Param_hbd, DISABLED_Speed) { SPEEDCHECKd(uint16_t, 16); } +TEST_P(Loop8Test9Param_lbd, DISABLED_Speed) { SPEEDCHECKd(uint8_t, 8); } + +using ::testing::make_tuple; + +#if HAVE_SSE2 + +const hbdloop_param_t kHbdLoop8Test6[] = { + make_tuple(&aom_highbd_lpf_horizontal_4_sse2, &aom_highbd_lpf_horizontal_4_c, + 8), + make_tuple(&aom_highbd_lpf_vertical_4_sse2, &aom_highbd_lpf_vertical_4_c, 8), + make_tuple(&aom_highbd_lpf_horizontal_6_sse2, &aom_highbd_lpf_horizontal_6_c, + 8), + make_tuple(&aom_highbd_lpf_horizontal_8_sse2, &aom_highbd_lpf_horizontal_8_c, + 8), + make_tuple(&aom_highbd_lpf_horizontal_14_sse2, + &aom_highbd_lpf_horizontal_14_c, 8), + make_tuple(&aom_highbd_lpf_vertical_6_sse2, &aom_highbd_lpf_vertical_6_c, 8), + make_tuple(&aom_highbd_lpf_vertical_8_sse2, &aom_highbd_lpf_vertical_8_c, 8), + + make_tuple(&aom_highbd_lpf_vertical_14_sse2, &aom_highbd_lpf_vertical_14_c, + 8), + make_tuple(&aom_highbd_lpf_horizontal_4_sse2, &aom_highbd_lpf_horizontal_4_c, + 10), + make_tuple(&aom_highbd_lpf_vertical_4_sse2, &aom_highbd_lpf_vertical_4_c, 10), + make_tuple(&aom_highbd_lpf_horizontal_6_sse2, &aom_highbd_lpf_horizontal_6_c, + 10), + make_tuple(&aom_highbd_lpf_horizontal_8_sse2, &aom_highbd_lpf_horizontal_8_c, + 10), + make_tuple(&aom_highbd_lpf_horizontal_14_sse2, + &aom_highbd_lpf_horizontal_14_c, 10), + make_tuple(&aom_highbd_lpf_vertical_6_sse2, &aom_highbd_lpf_vertical_6_c, 10), + make_tuple(&aom_highbd_lpf_vertical_8_sse2, &aom_highbd_lpf_vertical_8_c, 10), + make_tuple(&aom_highbd_lpf_vertical_14_sse2, &aom_highbd_lpf_vertical_14_c, + 10), + make_tuple(&aom_highbd_lpf_horizontal_4_sse2, &aom_highbd_lpf_horizontal_4_c, + 12), + make_tuple(&aom_highbd_lpf_vertical_4_sse2, &aom_highbd_lpf_vertical_4_c, 12), + make_tuple(&aom_highbd_lpf_horizontal_6_sse2, &aom_highbd_lpf_horizontal_6_c, + 12), + make_tuple(&aom_highbd_lpf_horizontal_8_sse2, &aom_highbd_lpf_horizontal_8_c, + 12), + make_tuple(&aom_highbd_lpf_horizontal_14_sse2, + &aom_highbd_lpf_horizontal_14_c, 12), + make_tuple(&aom_highbd_lpf_vertical_14_sse2, &aom_highbd_lpf_vertical_14_c, + 12), + make_tuple(&aom_highbd_lpf_vertical_6_sse2, &aom_highbd_lpf_vertical_6_c, 12), + make_tuple(&aom_highbd_lpf_vertical_8_sse2, &aom_highbd_lpf_vertical_8_c, 12) +}; + +INSTANTIATE_TEST_CASE_P(SSE2, Loop8Test6Param_hbd, + ::testing::ValuesIn(kHbdLoop8Test6)); + +const loop_param_t kLoop8Test6[] = { + make_tuple(&aom_lpf_horizontal_4_sse2, &aom_lpf_horizontal_4_c, 8), + make_tuple(&aom_lpf_horizontal_8_sse2, &aom_lpf_horizontal_8_c, 8), + make_tuple(&aom_lpf_horizontal_6_sse2, &aom_lpf_horizontal_6_c, 8), + make_tuple(&aom_lpf_vertical_6_sse2, &aom_lpf_vertical_6_c, 8), + make_tuple(&aom_lpf_horizontal_14_sse2, &aom_lpf_horizontal_14_c, 8), + make_tuple(&aom_lpf_vertical_4_sse2, &aom_lpf_vertical_4_c, 8), + make_tuple(&aom_lpf_vertical_8_sse2, &aom_lpf_vertical_8_c, 8), + make_tuple(&aom_lpf_vertical_14_sse2, &aom_lpf_vertical_14_c, 8), +}; + +INSTANTIATE_TEST_CASE_P(SSE2, Loop8Test6Param_lbd, + ::testing::ValuesIn(kLoop8Test6)); + +const dual_loop_param_t kLoop8Test9[] = { + make_tuple(&aom_lpf_horizontal_4_dual_sse2, &aom_lpf_horizontal_4_dual_c, 8), + make_tuple(&aom_lpf_vertical_4_dual_sse2, &aom_lpf_vertical_4_dual_c, 8), + make_tuple(&aom_lpf_horizontal_6_dual_sse2, &aom_lpf_horizontal_6_dual_c, 8), + make_tuple(&aom_lpf_vertical_6_dual_sse2, &aom_lpf_vertical_6_dual_c, 8), + make_tuple(&aom_lpf_horizontal_8_dual_sse2, &aom_lpf_horizontal_8_dual_c, 8), + make_tuple(&aom_lpf_vertical_8_dual_sse2, &aom_lpf_vertical_8_dual_c, 8), + make_tuple(&aom_lpf_horizontal_14_dual_sse2, &aom_lpf_horizontal_14_dual_c, + 8), + make_tuple(&aom_lpf_vertical_14_dual_sse2, &aom_lpf_vertical_14_dual_c, 8) +}; + +INSTANTIATE_TEST_CASE_P(SSE2, Loop8Test9Param_lbd, + ::testing::ValuesIn(kLoop8Test9)); + +#endif // HAVE_SSE2 + +#if HAVE_SSE2 +const hbddual_loop_param_t kHbdLoop8Test9[] = { + make_tuple(&aom_highbd_lpf_horizontal_4_dual_sse2, + &aom_highbd_lpf_horizontal_4_dual_c, 8), + make_tuple(&aom_highbd_lpf_horizontal_6_dual_sse2, + &aom_highbd_lpf_horizontal_6_dual_c, 8), + make_tuple(&aom_highbd_lpf_horizontal_8_dual_sse2, + &aom_highbd_lpf_horizontal_8_dual_c, 8), + make_tuple(&aom_highbd_lpf_horizontal_14_dual_sse2, + &aom_highbd_lpf_horizontal_14_dual_c, 8), + make_tuple(&aom_highbd_lpf_vertical_4_dual_sse2, + &aom_highbd_lpf_vertical_4_dual_c, 8), + make_tuple(&aom_highbd_lpf_vertical_6_dual_sse2, + &aom_highbd_lpf_vertical_6_dual_c, 8), + make_tuple(&aom_highbd_lpf_vertical_8_dual_sse2, + &aom_highbd_lpf_vertical_8_dual_c, 8), + make_tuple(&aom_highbd_lpf_vertical_14_dual_sse2, + &aom_highbd_lpf_vertical_14_dual_c, 8), + make_tuple(&aom_highbd_lpf_horizontal_4_dual_sse2, + &aom_highbd_lpf_horizontal_4_dual_c, 10), + make_tuple(&aom_highbd_lpf_horizontal_6_dual_sse2, + &aom_highbd_lpf_horizontal_6_dual_c, 10), + make_tuple(&aom_highbd_lpf_horizontal_8_dual_sse2, + &aom_highbd_lpf_horizontal_8_dual_c, 10), + make_tuple(&aom_highbd_lpf_horizontal_14_dual_sse2, + &aom_highbd_lpf_horizontal_14_dual_c, 10), + make_tuple(&aom_highbd_lpf_vertical_4_dual_sse2, + &aom_highbd_lpf_vertical_4_dual_c, 10), + make_tuple(&aom_highbd_lpf_vertical_6_dual_sse2, + &aom_highbd_lpf_vertical_6_dual_c, 10), + make_tuple(&aom_highbd_lpf_vertical_8_dual_sse2, + &aom_highbd_lpf_vertical_8_dual_c, 10), + make_tuple(&aom_highbd_lpf_vertical_14_dual_sse2, + &aom_highbd_lpf_vertical_14_dual_c, 10), + make_tuple(&aom_highbd_lpf_horizontal_4_dual_sse2, + &aom_highbd_lpf_horizontal_4_dual_c, 12), + make_tuple(&aom_highbd_lpf_horizontal_6_dual_sse2, + &aom_highbd_lpf_horizontal_6_dual_c, 12), + make_tuple(&aom_highbd_lpf_horizontal_8_dual_sse2, + &aom_highbd_lpf_horizontal_8_dual_c, 12), + make_tuple(&aom_highbd_lpf_horizontal_14_dual_sse2, + &aom_highbd_lpf_horizontal_14_dual_c, 12), + make_tuple(&aom_highbd_lpf_vertical_4_dual_sse2, + &aom_highbd_lpf_vertical_4_dual_c, 12), + make_tuple(&aom_highbd_lpf_vertical_6_dual_sse2, + &aom_highbd_lpf_vertical_6_dual_c, 12), + make_tuple(&aom_highbd_lpf_vertical_8_dual_sse2, + &aom_highbd_lpf_vertical_8_dual_c, 12), + make_tuple(&aom_highbd_lpf_vertical_14_dual_sse2, + &aom_highbd_lpf_vertical_14_dual_c, 12), +}; + +INSTANTIATE_TEST_CASE_P(SSE2, Loop8Test9Param_hbd, + ::testing::ValuesIn(kHbdLoop8Test9)); + +#endif // HAVE_SSE2 + +#if HAVE_NEON +const loop_param_t kLoop8Test6[] = { + make_tuple(&aom_lpf_vertical_14_neon, &aom_lpf_vertical_14_c, 8), + make_tuple(&aom_lpf_vertical_8_neon, &aom_lpf_vertical_8_c, 8), + make_tuple(&aom_lpf_horizontal_8_neon, &aom_lpf_horizontal_8_c, 8), + make_tuple(&aom_lpf_horizontal_6_neon, &aom_lpf_horizontal_6_c, 8) +}; + +INSTANTIATE_TEST_CASE_P(NEON, Loop8Test6Param_lbd, + ::testing::ValuesIn(kLoop8Test6)); +#endif // HAVE_NEON + +#if HAVE_AVX2 +const hbddual_loop_param_t kHbdLoop8Test9Avx2[] = { + make_tuple(&aom_highbd_lpf_horizontal_4_dual_avx2, + &aom_highbd_lpf_horizontal_4_dual_c, 8), + make_tuple(&aom_highbd_lpf_horizontal_4_dual_avx2, + &aom_highbd_lpf_horizontal_4_dual_c, 10), + make_tuple(&aom_highbd_lpf_horizontal_4_dual_avx2, + &aom_highbd_lpf_horizontal_4_dual_c, 12), + make_tuple(&aom_highbd_lpf_horizontal_8_dual_avx2, + &aom_highbd_lpf_horizontal_8_dual_c, 8), + make_tuple(&aom_highbd_lpf_horizontal_8_dual_avx2, + &aom_highbd_lpf_horizontal_8_dual_c, 10), + make_tuple(&aom_highbd_lpf_horizontal_8_dual_avx2, + &aom_highbd_lpf_horizontal_8_dual_c, 12), + make_tuple(&aom_highbd_lpf_vertical_4_dual_avx2, + &aom_highbd_lpf_vertical_4_dual_c, 8), + make_tuple(&aom_highbd_lpf_vertical_4_dual_avx2, + &aom_highbd_lpf_vertical_4_dual_c, 10), + make_tuple(&aom_highbd_lpf_vertical_4_dual_avx2, + &aom_highbd_lpf_vertical_4_dual_c, 12), + make_tuple(&aom_highbd_lpf_vertical_8_dual_avx2, + &aom_highbd_lpf_vertical_8_dual_c, 8), + make_tuple(&aom_highbd_lpf_vertical_8_dual_avx2, + &aom_highbd_lpf_vertical_8_dual_c, 10), + make_tuple(&aom_highbd_lpf_vertical_8_dual_avx2, + &aom_highbd_lpf_vertical_8_dual_c, 12), +}; + +INSTANTIATE_TEST_CASE_P(AVX2, Loop8Test9Param_hbd, + ::testing::ValuesIn(kHbdLoop8Test9Avx2)); +#endif +} // namespace -- cgit v1.2.3