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
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
|
// Copyright 2013 The Chromium Authors. All rights reserved.
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file.
#include "base/cpu.h"
#include "base/memory/aligned_memory.h"
#include "base/memory/scoped_ptr.h"
#include "base/time/time.h"
#include "media/base/vector_math.h"
#include "media/base/vector_math_testing.h"
#include "testing/gtest/include/gtest/gtest.h"
#include "testing/perf/perf_test.h"
using base::TimeTicks;
using std::fill;
namespace media {
static const int kBenchmarkIterations = 200000;
static const int kEWMABenchmarkIterations = 50000;
static const float kScale = 0.5;
static const int kVectorSize = 8192;
class VectorMathPerfTest : public testing::Test {
public:
VectorMathPerfTest() {
// Initialize input and output vectors.
input_vector_.reset(static_cast<float*>(base::AlignedAlloc(
sizeof(float) * kVectorSize, vector_math::kRequiredAlignment)));
output_vector_.reset(static_cast<float*>(base::AlignedAlloc(
sizeof(float) * kVectorSize, vector_math::kRequiredAlignment)));
fill(input_vector_.get(), input_vector_.get() + kVectorSize, 1.0f);
fill(output_vector_.get(), output_vector_.get() + kVectorSize, 0.0f);
}
void RunBenchmark(void (*fn)(const float[], float, int, float[]),
bool aligned,
const std::string& test_name,
const std::string& trace_name) {
TimeTicks start = TimeTicks::HighResNow();
for (int i = 0; i < kBenchmarkIterations; ++i) {
fn(input_vector_.get(),
kScale,
kVectorSize - (aligned ? 0 : 1),
output_vector_.get());
}
double total_time_milliseconds =
(TimeTicks::HighResNow() - start).InMillisecondsF();
perf_test::PrintResult(test_name,
"",
trace_name,
kBenchmarkIterations / total_time_milliseconds,
"runs/ms",
true);
}
void RunBenchmark(
std::pair<float, float> (*fn)(float, const float[], int, float),
int len,
const std::string& test_name,
const std::string& trace_name) {
TimeTicks start = TimeTicks::HighResNow();
for (int i = 0; i < kEWMABenchmarkIterations; ++i) {
fn(0.5f, input_vector_.get(), len, 0.1f);
}
double total_time_milliseconds =
(TimeTicks::HighResNow() - start).InMillisecondsF();
perf_test::PrintResult(test_name,
"",
trace_name,
kEWMABenchmarkIterations / total_time_milliseconds,
"runs/ms",
true);
}
protected:
scoped_ptr<float, base::AlignedFreeDeleter> input_vector_;
scoped_ptr<float, base::AlignedFreeDeleter> output_vector_;
DISALLOW_COPY_AND_ASSIGN(VectorMathPerfTest);
};
// Define platform independent function name for FMAC* perf tests.
#if defined(ARCH_CPU_X86_FAMILY)
#define FMAC_FUNC FMAC_SSE
#elif defined(ARCH_CPU_ARM_FAMILY) && defined(USE_NEON)
#define FMAC_FUNC FMAC_NEON
#endif
// Benchmark for each optimized vector_math::FMAC() method.
TEST_F(VectorMathPerfTest, FMAC) {
// Benchmark FMAC_C().
RunBenchmark(
vector_math::FMAC_C, true, "vector_math_fmac", "unoptimized");
#if defined(FMAC_FUNC)
#if defined(ARCH_CPU_X86_FAMILY)
ASSERT_TRUE(base::CPU().has_sse());
#endif
// Benchmark FMAC_FUNC() with unaligned size.
ASSERT_NE((kVectorSize - 1) % (vector_math::kRequiredAlignment /
sizeof(float)), 0U);
RunBenchmark(
vector_math::FMAC_FUNC, false, "vector_math_fmac", "optimized_unaligned");
// Benchmark FMAC_FUNC() with aligned size.
ASSERT_EQ(kVectorSize % (vector_math::kRequiredAlignment / sizeof(float)),
0U);
RunBenchmark(
vector_math::FMAC_FUNC, true, "vector_math_fmac", "optimized_aligned");
#endif
}
#undef FMAC_FUNC
// Define platform independent function name for FMULBenchmark* tests.
#if defined(ARCH_CPU_X86_FAMILY)
#define FMUL_FUNC FMUL_SSE
#elif defined(ARCH_CPU_ARM_FAMILY) && defined(USE_NEON)
#define FMUL_FUNC FMUL_NEON
#endif
// Benchmark for each optimized vector_math::FMUL() method.
TEST_F(VectorMathPerfTest, FMUL) {
// Benchmark FMUL_C().
RunBenchmark(
vector_math::FMUL_C, true, "vector_math_fmul", "unoptimized");
#if defined(FMUL_FUNC)
#if defined(ARCH_CPU_X86_FAMILY)
ASSERT_TRUE(base::CPU().has_sse());
#endif
// Benchmark FMUL_FUNC() with unaligned size.
ASSERT_NE((kVectorSize - 1) % (vector_math::kRequiredAlignment /
sizeof(float)), 0U);
RunBenchmark(
vector_math::FMUL_FUNC, false, "vector_math_fmul", "optimized_unaligned");
// Benchmark FMUL_FUNC() with aligned size.
ASSERT_EQ(kVectorSize % (vector_math::kRequiredAlignment / sizeof(float)),
0U);
RunBenchmark(
vector_math::FMUL_FUNC, true, "vector_math_fmul", "optimized_aligned");
#endif
}
#undef FMUL_FUNC
#if defined(ARCH_CPU_X86_FAMILY)
#define EWMAAndMaxPower_FUNC EWMAAndMaxPower_SSE
#elif defined(ARCH_CPU_ARM_FAMILY) && defined(USE_NEON)
#define EWMAAndMaxPower_FUNC EWMAAndMaxPower_NEON
#endif
// Benchmark for each optimized vector_math::EWMAAndMaxPower() method.
TEST_F(VectorMathPerfTest, EWMAAndMaxPower) {
// Benchmark EWMAAndMaxPower_C().
RunBenchmark(vector_math::EWMAAndMaxPower_C,
kVectorSize,
"vector_math_ewma_and_max_power",
"unoptimized");
#if defined(EWMAAndMaxPower_FUNC)
#if defined(ARCH_CPU_X86_FAMILY)
ASSERT_TRUE(base::CPU().has_sse());
#endif
// Benchmark EWMAAndMaxPower_FUNC() with unaligned size.
ASSERT_NE((kVectorSize - 1) % (vector_math::kRequiredAlignment /
sizeof(float)), 0U);
RunBenchmark(vector_math::EWMAAndMaxPower_FUNC,
kVectorSize - 1,
"vector_math_ewma_and_max_power",
"optimized_unaligned");
// Benchmark EWMAAndMaxPower_FUNC() with aligned size.
ASSERT_EQ(kVectorSize % (vector_math::kRequiredAlignment / sizeof(float)),
0U);
RunBenchmark(vector_math::EWMAAndMaxPower_FUNC,
kVectorSize,
"vector_math_ewma_and_max_power",
"optimized_aligned");
#endif
}
#undef EWMAAndMaxPower_FUNC
} // namespace media
|
