multiversed 0.3.1

Attribute macros wrapping multiversion with predefined SIMD target presets
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
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207

//! Benchmark to measure multiversion dispatch overhead.
//!
//! Compares dispatch overhead between:
//! - Huge feature string (many features checked)
//! - Minimal feature string (few features checked)
//!
//! The goal is to determine if listing many features in the target string
//! adds measurable overhead to the runtime dispatch.

use criterion::{black_box, criterion_group, criterion_main, Criterion, Throughput};

// ============================================================================
// Test functions with different target string sizes
// ============================================================================

/// Minimal target: just AVX-512F
#[multiversion::multiversion(targets("x86_64+avx512f"))]
fn sum_minimal(data: &[f32]) -> f32 {
    data.iter().sum()
}

/// Medium target: x86-64-v4 core (5 AVX-512 features)
#[multiversion::multiversion(targets("x86_64+avx512f+avx512bw+avx512dq+avx512vl+avx512cd"))]
fn sum_medium(data: &[f32]) -> f32 {
    data.iter().sum()
}

/// Huge target: full psABI v4 + extras (20+ features)
#[multiversion::multiversion(targets(
    "x86_64+sse+sse2+sse3+ssse3+sse4.1+sse4.2+popcnt+cmpxchg16b+avx+avx2+bmi1+bmi2+f16c+fma+lzcnt+movbe+xsave+fxsr+avx512f+avx512bw+avx512dq+avx512vl+avx512cd+gfni+vaes+vpclmulqdq"
))]
fn sum_huge(data: &[f32]) -> f32 {
    data.iter().sum()
}

/// Multiple targets: v4, v3, v2 (dispatch chooses best)
#[multiversion::multiversion(targets(
    "x86_64+sse+sse2+sse3+ssse3+sse4.1+sse4.2+popcnt+cmpxchg16b+avx+avx2+bmi1+bmi2+f16c+fma+lzcnt+movbe+xsave+fxsr+avx512f+avx512bw+avx512dq+avx512vl+avx512cd+gfni+vaes+vpclmulqdq",
    "x86_64+sse+sse2+sse3+ssse3+sse4.1+sse4.2+popcnt+cmpxchg16b+avx+avx2+bmi1+bmi2+f16c+fma+lzcnt+movbe+xsave+fxsr",
    "x86_64+sse+sse2+sse3+ssse3+sse4.1+sse4.2+popcnt"
))]
fn sum_multi_target(data: &[f32]) -> f32 {
    data.iter().sum()
}

/// 6 different targets - maximum dispatch complexity
#[multiversion::multiversion(targets(
    // v4 with all extensions
    "x86_64+sse+sse2+sse3+ssse3+sse4.1+sse4.2+popcnt+cmpxchg16b+avx+avx2+bmi1+bmi2+f16c+fma+lzcnt+movbe+xsave+fxsr+avx512f+avx512bw+avx512dq+avx512vl+avx512cd+avx512vnni+avx512vbmi+avx512vbmi2+avx512bitalg+avx512vpopcntdq+avx512bf16+gfni+vaes+vpclmulqdq",
    // v4 core only
    "x86_64+sse+sse2+sse3+ssse3+sse4.1+sse4.2+popcnt+cmpxchg16b+avx+avx2+bmi1+bmi2+f16c+fma+lzcnt+movbe+xsave+fxsr+avx512f+avx512bw+avx512dq+avx512vl+avx512cd",
    // v3 full
    "x86_64+sse+sse2+sse3+ssse3+sse4.1+sse4.2+popcnt+cmpxchg16b+avx+avx2+bmi1+bmi2+f16c+fma+lzcnt+movbe+xsave+fxsr",
    // v3 minimal
    "x86_64+avx+avx2+fma",
    // v2
    "x86_64+sse+sse2+sse3+ssse3+sse4.1+sse4.2+popcnt",
    // v1 (baseline)
    "x86_64+sse+sse2"
))]
fn sum_6_targets(data: &[f32]) -> f32 {
    data.iter().sum()
}

/// No multiversion - baseline for comparison
fn sum_baseline(data: &[f32]) -> f32 {
    data.iter().sum()
}

// ============================================================================
// Benchmarks
// ============================================================================

fn bench_dispatch_overhead(c: &mut Criterion) {
    // Small data to emphasize dispatch overhead vs actual work
    let small_data: Vec<f32> = (0..64).map(|i| i as f32).collect();
    // Medium data
    let medium_data: Vec<f32> = (0..1024).map(|i| i as f32).collect();
    // Large data where dispatch overhead should be negligible
    let large_data: Vec<f32> = (0..65536).map(|i| i as f32).collect();

    let mut group = c.benchmark_group("dispatch_overhead_small");
    group.throughput(Throughput::Elements(small_data.len() as u64));

    group.bench_function("baseline_no_multiversion", |b| {
        b.iter(|| sum_baseline(black_box(&small_data)))
    });
    group.bench_function("minimal_1_feature", |b| {
        b.iter(|| sum_minimal(black_box(&small_data)))
    });
    group.bench_function("medium_5_features", |b| {
        b.iter(|| sum_medium(black_box(&small_data)))
    });
    group.bench_function("huge_27_features", |b| {
        b.iter(|| sum_huge(black_box(&small_data)))
    });
    group.bench_function("multi_target_3_versions", |b| {
        b.iter(|| sum_multi_target(black_box(&small_data)))
    });
    group.bench_function("multi_target_6_versions", |b| {
        b.iter(|| sum_6_targets(black_box(&small_data)))
    });
    group.finish();

    let mut group = c.benchmark_group("dispatch_overhead_medium");
    group.throughput(Throughput::Elements(medium_data.len() as u64));

    group.bench_function("baseline_no_multiversion", |b| {
        b.iter(|| sum_baseline(black_box(&medium_data)))
    });
    group.bench_function("minimal_1_feature", |b| {
        b.iter(|| sum_minimal(black_box(&medium_data)))
    });
    group.bench_function("huge_27_features", |b| {
        b.iter(|| sum_huge(black_box(&medium_data)))
    });
    group.bench_function("multi_target_3_versions", |b| {
        b.iter(|| sum_multi_target(black_box(&medium_data)))
    });
    group.bench_function("multi_target_6_versions", |b| {
        b.iter(|| sum_6_targets(black_box(&medium_data)))
    });
    group.finish();

    let mut group = c.benchmark_group("dispatch_overhead_large");
    group.throughput(Throughput::Elements(large_data.len() as u64));

    group.bench_function("baseline_no_multiversion", |b| {
        b.iter(|| sum_baseline(black_box(&large_data)))
    });
    group.bench_function("minimal_1_feature", |b| {
        b.iter(|| sum_minimal(black_box(&large_data)))
    });
    group.bench_function("huge_27_features", |b| {
        b.iter(|| sum_huge(black_box(&large_data)))
    });
    group.bench_function("multi_target_3_versions", |b| {
        b.iter(|| sum_multi_target(black_box(&large_data)))
    });
    group.bench_function("multi_target_6_versions", |b| {
        b.iter(|| sum_6_targets(black_box(&large_data)))
    });
    group.finish();
}

/// Benchmark repeated calls to measure per-call dispatch overhead
fn bench_repeated_dispatch(c: &mut Criterion) {
    let data: Vec<f32> = (0..64).map(|i| i as f32).collect();

    let mut group = c.benchmark_group("repeated_dispatch");

    // Call the function many times in a loop to amortize any one-time costs
    group.bench_function("baseline_1000_calls", |b| {
        b.iter(|| {
            let mut total = 0.0f32;
            for _ in 0..1000 {
                total += sum_baseline(black_box(&data));
            }
            total
        })
    });

    group.bench_function("minimal_1000_calls", |b| {
        b.iter(|| {
            let mut total = 0.0f32;
            for _ in 0..1000 {
                total += sum_minimal(black_box(&data));
            }
            total
        })
    });

    group.bench_function("huge_1000_calls", |b| {
        b.iter(|| {
            let mut total = 0.0f32;
            for _ in 0..1000 {
                total += sum_huge(black_box(&data));
            }
            total
        })
    });

    group.bench_function("multi_target_3_1000_calls", |b| {
        b.iter(|| {
            let mut total = 0.0f32;
            for _ in 0..1000 {
                total += sum_multi_target(black_box(&data));
            }
            total
        })
    });

    group.bench_function("multi_target_6_1000_calls", |b| {
        b.iter(|| {
            let mut total = 0.0f32;
            for _ in 0..1000 {
                total += sum_6_targets(black_box(&data));
            }
            total
        })
    });

    group.finish();
}

criterion_group!(benches, bench_dispatch_overhead, bench_repeated_dispatch);
criterion_main!(benches);