Repository navigation
Expand file tree
/
Copy pathbench_map.cpp
More file actions
158 lines (145 loc) · 6.97 KB
/
Copy pathbench_map.cpp
File metadata and controls
158 lines (145 loc) · 6.97 KB
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
/****************************************************************************
* Copyright (c) xsimd-algorithm contributors *
* *
* Distributed under the terms of the BSD 3-Clause License. *
* *
* The full license is in the file LICENSE, distributed with this software. *
****************************************************************************/
#include <cstddef>
#include <cstdint>
#include <benchmark/benchmark.h>
#include "map_binary_utils.hpp"
#include "map_unary_utils.hpp"
#include "xsimd_algorithm/map.hpp"
#include "xsimd_test_utils/map_binary_data.hpp"
#include "xsimd_test_utils/map_unary_data.hpp"
#include "xsimd_test_utils/utils.hpp"
namespace
{
using xsimd::bench::bench_binary_scalar;
using xsimd::bench::bench_map_binary;
using xsimd::bench::bench_map_unary;
using xsimd::bench::bench_scalar;
using xsimd::bench::bench_transform;
using xsimd::bench::register_binary_bench;
using xsimd::bench::register_bench;
using xsimd::alignment_options;
using xsimd::map_options;
template <typename Op>
void register_benches()
{
using input_t = typename Op::input_t;
using arch = xsimd::default_arch;
using aligned_alloc = typename xsimd::test::aligned_vector<input_t, arch>::allocator_type;
using unaligned_alloc = typename xsimd::test::unaligned_vector<input_t, arch>::allocator_type;
constexpr auto noalign = alignment_options {};
constexpr auto noopts = map_options { .unroll_factor = 1, .pure = false };
register_bench<Op, arch>("aligned/scalar", bench_scalar<Op, aligned_alloc>);
register_bench<Op, arch>("aligned/simd/transform", bench_transform<Op, aligned_alloc, arch>);
register_bench<Op, arch>(
"aligned/simd/map",
bench_map_unary<Op, aligned_alloc, arch, noalign, noopts>);
register_bench<Op, arch>(
"aligned/simd/map:pure",
bench_map_unary<
Op, aligned_alloc, arch,
noalign, map_options { .unroll_factor = 1, .pure = true }>);
register_bench<Op, arch>(
"aligned/simd/map:unroll4",
bench_map_unary<
Op, aligned_alloc, arch, noalign, map_options { .unroll_factor = 4 }>);
register_bench<Op, arch>(
"aligned/simd/map:noheader",
bench_map_unary<
Op, aligned_alloc, arch,
alignment_options { .start_aligned = true }, noopts>);
register_bench<Op, arch>(
"aligned/simd/map:noheader+pure+unroll4",
bench_map_unary<
Op, aligned_alloc, arch,
alignment_options { .start_aligned = true }, map_options { .unroll_factor = 4, .pure = true }>);
register_bench<Op, arch>("unaligned/scalar", bench_scalar<Op, unaligned_alloc>);
register_bench<Op, arch>("unaligned/simd/transform", bench_transform<Op, unaligned_alloc, arch>);
register_bench<Op, arch>(
"unaligned/simd/map",
bench_map_unary<Op, unaligned_alloc, arch, noalign, noopts>);
register_bench<Op, arch>(
"unaligned/simd/map:pure",
bench_map_unary<
Op, unaligned_alloc, arch,
noalign, map_options { .unroll_factor = 1, .pure = true }>);
register_bench<Op, arch>(
"unaligned/simd/map:unroll4",
bench_map_unary<
Op, unaligned_alloc, arch, noalign, map_options { .unroll_factor = 4 }>);
register_bench<Op, arch>(
"unaligned/simd/map:pure+unroll4",
bench_map_unary<
Op, unaligned_alloc, arch,
noalign, map_options { .unroll_factor = 4, .pure = true }>);
}
template <typename Op>
void register_binary_benches()
{
using lhs_t = typename Op::lhs_t;
using arch = xsimd::default_arch;
using aligned_alloc = typename xsimd::test::aligned_vector<lhs_t, arch>::allocator_type;
using unaligned_alloc = typename xsimd::test::unaligned_vector<lhs_t, arch>::allocator_type;
constexpr auto noalign = alignment_options {};
constexpr auto noopts = map_options { .unroll_factor = 1, .pure = false };
register_binary_bench<Op, arch>("aligned/scalar", bench_binary_scalar<Op, aligned_alloc>);
register_binary_bench<Op, arch>(
"aligned/simd/map",
bench_map_binary<Op, aligned_alloc, arch, noalign, noopts>);
register_binary_bench<Op, arch>(
"aligned/simd/map:pure",
bench_map_binary<
Op, aligned_alloc, arch,
noalign, map_options { .unroll_factor = 1, .pure = true }>);
register_binary_bench<Op, arch>(
"aligned/simd/map:unroll4",
bench_map_binary<
Op, aligned_alloc, arch, noalign, map_options { .unroll_factor = 4 }>);
register_binary_bench<Op, arch>(
"aligned/simd/map:noheader",
bench_map_binary<
Op, aligned_alloc, arch,
alignment_options { .start_aligned = true }, noopts>);
register_binary_bench<Op, arch>(
"aligned/simd/map:noheader+pure+unroll4",
bench_map_binary<
Op, aligned_alloc, arch,
alignment_options { .start_aligned = true }, map_options { .unroll_factor = 4, .pure = true }>);
register_binary_bench<Op, arch>("unaligned/scalar", bench_binary_scalar<Op, unaligned_alloc>);
register_binary_bench<Op, arch>(
"unaligned/simd/map",
bench_map_binary<Op, unaligned_alloc, arch, noalign, noopts>);
register_binary_bench<Op, arch>(
"unaligned/simd/map:pure",
bench_map_binary<
Op, unaligned_alloc, arch,
noalign, map_options { .unroll_factor = 1, .pure = true }>);
register_binary_bench<Op, arch>(
"unaligned/simd/map:unroll4",
bench_map_binary<
Op, unaligned_alloc, arch, noalign, map_options { .unroll_factor = 4 }>);
register_binary_bench<Op, arch>(
"unaligned/simd/map:pure+unroll4",
bench_map_binary<
Op, unaligned_alloc, arch,
noalign, map_options { .unroll_factor = 4, .pure = true }>);
}
bool const registered = []
{
register_benches<xsimd::test::sqrt_op<float>>();
register_benches<xsimd::test::sqrt_op<double>>();
register_benches<xsimd::test::abs_op<float>>();
register_benches<xsimd::test::abs_op<double>>();
register_benches<xsimd::test::exp_op<float>>();
register_benches<xsimd::test::exp_op<double>>();
register_binary_benches<xsimd::test::add_op<std::int32_t>>();
register_binary_benches<xsimd::test::multiply_op<float>>();
register_binary_benches<xsimd::test::mixed_multiply_op>();
return true;
}();
}