1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
|
/*
* Copyright (c) 2019 The WebRTC project authors. All Rights Reserved.
*
* Use of this source code is governed by a BSD-style license
* that can be found in the LICENSE file in the root of the source
* tree. An additional intellectual property rights grant can be found
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
*/
#include "modules/audio_processing/aec3/adaptive_fir_filter_erl.h"
#include <array>
#include <vector>
#include "rtc_base/system/arch.h"
#if defined(WEBRTC_ARCH_X86_FAMILY)
#include <emmintrin.h>
#endif
#include "system_wrappers/include/cpu_features_wrapper.h"
#include "test/gtest.h"
namespace webrtc {
namespace aec3 {
#if defined(WEBRTC_HAS_NEON)
// Verifies that the optimized method for echo return loss computation is
// bitexact to the reference counterpart.
TEST(AdaptiveFirFilter, UpdateErlNeonOptimization) {
const size_t kNumPartitions = 12;
std::vector<std::array<float, kFftLengthBy2Plus1>> H2(kNumPartitions);
std::array<float, kFftLengthBy2Plus1> erl;
std::array<float, kFftLengthBy2Plus1> erl_NEON;
for (size_t j = 0; j < H2.size(); ++j) {
for (size_t k = 0; k < H2[j].size(); ++k) {
H2[j][k] = k + j / 3.f;
}
}
ErlComputer(H2, erl);
ErlComputer_NEON(H2, erl_NEON);
for (size_t j = 0; j < erl.size(); ++j) {
EXPECT_FLOAT_EQ(erl[j], erl_NEON[j]);
}
}
#endif
#if defined(WEBRTC_ARCH_X86_FAMILY)
// Verifies that the optimized method for echo return loss computation is
// bitexact to the reference counterpart.
TEST(AdaptiveFirFilter, UpdateErlSse2Optimization) {
bool use_sse2 = (GetCPUInfo(kSSE2) != 0);
if (use_sse2) {
const size_t kNumPartitions = 12;
std::vector<std::array<float, kFftLengthBy2Plus1>> H2(kNumPartitions);
std::array<float, kFftLengthBy2Plus1> erl;
std::array<float, kFftLengthBy2Plus1> erl_SSE2;
for (size_t j = 0; j < H2.size(); ++j) {
for (size_t k = 0; k < H2[j].size(); ++k) {
H2[j][k] = k + j / 3.f;
}
}
ErlComputer(H2, erl);
ErlComputer_SSE2(H2, erl_SSE2);
for (size_t j = 0; j < erl.size(); ++j) {
EXPECT_FLOAT_EQ(erl[j], erl_SSE2[j]);
}
}
}
// Verifies that the optimized method for echo return loss computation is
// bitexact to the reference counterpart.
TEST(AdaptiveFirFilter, UpdateErlAvx2Optimization) {
bool use_avx2 = (GetCPUInfo(kAVX2) != 0);
if (use_avx2) {
const size_t kNumPartitions = 12;
std::vector<std::array<float, kFftLengthBy2Plus1>> H2(kNumPartitions);
std::array<float, kFftLengthBy2Plus1> erl;
std::array<float, kFftLengthBy2Plus1> erl_AVX2;
for (size_t j = 0; j < H2.size(); ++j) {
for (size_t k = 0; k < H2[j].size(); ++k) {
H2[j][k] = k + j / 3.f;
}
}
ErlComputer(H2, erl);
ErlComputer_AVX2(H2, erl_AVX2);
for (size_t j = 0; j < erl.size(); ++j) {
EXPECT_FLOAT_EQ(erl[j], erl_AVX2[j]);
}
}
}
#endif
} // namespace aec3
} // namespace webrtc
|
