simdeez 3.0.1

SIMD library to abstract over different instruction sets and widths
Documentation 
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

use criterion::Criterion;
use simdeez::math::SimdMathF32Core;
#[cfg(not(any(target_arch = "x86_64", target_arch = "x86")))]
use simdeez::scalar::Scalar;
use simdeez::{prelude::*, simd_unsafe_generate_all};

use crate::shared::{self, BenchTargets, INPUT_LEN};

#[inline(never)]
fn scalar_log2_sum(input: &[f32]) -> f32 {
    input.iter().copied().map(f32::log2).sum()
}

#[inline(never)]
fn scalar_exp2_sum(input: &[f32]) -> f32 {
    input.iter().copied().map(f32::exp2).sum()
}

#[inline(never)]
fn scalar_ln_sum(input: &[f32]) -> f32 {
    input.iter().copied().map(f32::ln).sum()
}

#[inline(never)]
fn scalar_exp_sum(input: &[f32]) -> f32 {
    input.iter().copied().map(f32::exp).sum()
}

simd_unsafe_generate_all!(
    fn simdeez_log2_sum(input: &[f32]) -> f32 {
        shared::simdeez_sum_impl::<S>(input, |v| v.log2_u35())
    }
);

simd_unsafe_generate_all!(
    fn simdeez_exp2_sum(input: &[f32]) -> f32 {
        shared::simdeez_sum_impl::<S>(input, |v| v.exp2_u35())
    }
);

simd_unsafe_generate_all!(
    fn simdeez_ln_sum(input: &[f32]) -> f32 {
        shared::simdeez_sum_impl::<S>(input, |v| v.ln_u35())
    }
);

simd_unsafe_generate_all!(
    fn simdeez_exp_sum(input: &[f32]) -> f32 {
        shared::simdeez_sum_impl::<S>(input, |v| v.exp_u35())
    }
);

#[cfg(not(any(target_arch = "x86_64", target_arch = "x86")))]
#[inline(never)]
fn simdeez_log2_sum_scalar(input: &[f32]) -> f32 {
    shared::force_scalar_sum(input, |v: <Scalar as Simd>::Vf32| v.log2_u35())
}

#[cfg(not(any(target_arch = "x86_64", target_arch = "x86")))]
#[inline(never)]
fn simdeez_exp2_sum_scalar(input: &[f32]) -> f32 {
    shared::force_scalar_sum(input, |v: <Scalar as Simd>::Vf32| v.exp2_u35())
}

#[cfg(not(any(target_arch = "x86_64", target_arch = "x86")))]
#[inline(never)]
fn simdeez_ln_sum_scalar(input: &[f32]) -> f32 {
    shared::force_scalar_sum(input, |v: <Scalar as Simd>::Vf32| v.ln_u35())
}

#[cfg(not(any(target_arch = "x86_64", target_arch = "x86")))]
#[inline(never)]
fn simdeez_exp_sum_scalar(input: &[f32]) -> f32 {
    shared::force_scalar_sum(input, |v: <Scalar as Simd>::Vf32| v.exp_u35())
}

pub fn register(c: &mut Criterion) {
    let log_inputs = shared::make_positive_log_inputs(INPUT_LEN, 0xA11C_E001);
    let exp2_inputs = shared::make_exp2_inputs(INPUT_LEN, 0xA11C_E002);
    let exp_inputs = shared::make_exp_inputs(INPUT_LEN, 0xA11C_E003);

    shared::bench_variants(
        c,
        "simd_math/f32/log2_u35",
        &log_inputs,
        BenchTargets {
            scalar_native: scalar_log2_sum,
            simdeez_runtime: simdeez_log2_sum,
            simdeez_scalar: simdeez_log2_sum_scalar,
            #[cfg(any(target_arch = "x86_64", target_arch = "x86"))]
            simdeez_sse2: simdeez_log2_sum_sse2,
            #[cfg(any(target_arch = "x86_64", target_arch = "x86"))]
            simdeez_sse41: simdeez_log2_sum_sse41,
            #[cfg(any(target_arch = "x86_64", target_arch = "x86"))]
            simdeez_avx2: simdeez_log2_sum_avx2,
            #[cfg(any(target_arch = "x86_64", target_arch = "x86"))]
            simdeez_avx512: simdeez_log2_sum_avx512,
        },
    );

    shared::bench_variants(
        c,
        "simd_math/f32/exp2_u35",
        &exp2_inputs,
        BenchTargets {
            scalar_native: scalar_exp2_sum,
            simdeez_runtime: simdeez_exp2_sum,
            simdeez_scalar: simdeez_exp2_sum_scalar,
            #[cfg(any(target_arch = "x86_64", target_arch = "x86"))]
            simdeez_sse2: simdeez_exp2_sum_sse2,
            #[cfg(any(target_arch = "x86_64", target_arch = "x86"))]
            simdeez_sse41: simdeez_exp2_sum_sse41,
            #[cfg(any(target_arch = "x86_64", target_arch = "x86"))]
            simdeez_avx2: simdeez_exp2_sum_avx2,
            #[cfg(any(target_arch = "x86_64", target_arch = "x86"))]
            simdeez_avx512: simdeez_exp2_sum_avx512,
        },
    );

    shared::bench_variants(
        c,
        "simd_math/f32/ln_u35",
        &log_inputs,
        BenchTargets {
            scalar_native: scalar_ln_sum,
            simdeez_runtime: simdeez_ln_sum,
            simdeez_scalar: simdeez_ln_sum_scalar,
            #[cfg(any(target_arch = "x86_64", target_arch = "x86"))]
            simdeez_sse2: simdeez_ln_sum_sse2,
            #[cfg(any(target_arch = "x86_64", target_arch = "x86"))]
            simdeez_sse41: simdeez_ln_sum_sse41,
            #[cfg(any(target_arch = "x86_64", target_arch = "x86"))]
            simdeez_avx2: simdeez_ln_sum_avx2,
            #[cfg(any(target_arch = "x86_64", target_arch = "x86"))]
            simdeez_avx512: simdeez_ln_sum_avx512,
        },
    );

    shared::bench_variants(
        c,
        "simd_math/f32/exp_u35",
        &exp_inputs,
        BenchTargets {
            scalar_native: scalar_exp_sum,
            simdeez_runtime: simdeez_exp_sum,
            simdeez_scalar: simdeez_exp_sum_scalar,
            #[cfg(any(target_arch = "x86_64", target_arch = "x86"))]
            simdeez_sse2: simdeez_exp_sum_sse2,
            #[cfg(any(target_arch = "x86_64", target_arch = "x86"))]
            simdeez_sse41: simdeez_exp_sum_sse41,
            #[cfg(any(target_arch = "x86_64", target_arch = "x86"))]
            simdeez_avx2: simdeez_exp_sum_avx2,
            #[cfg(any(target_arch = "x86_64", target_arch = "x86"))]
            simdeez_avx512: simdeez_exp_sum_avx512,
        },
    );
}