hermes-simd 0.6.0

High-performance, zero-overhead SIMD abstraction library
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
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287

//! Integration tests for generic `ElementOp` strategy dispatch:
//! `transform_in_place`, `zip_into`, and invariant checks that
//! `add_assign`/`mul_assign`/`elementwise_mul` produce identical results.

use hermes_simd::{
    Add, BitAnd, BitOr, BitXor, Div, Mul, Scalar, SimdError, SimdView, Sub, Unaligned, Unmasked,
};

// ---------------------------------------------------------------------------
// Helpers
// ---------------------------------------------------------------------------

type View<'a, T> = SimdView<'a, T, Scalar, Unaligned, Unmasked, &'a [T]>;
type ViewMut<'a, T> = SimdView<'a, T, Scalar, Unaligned, Unmasked, &'a mut [T]>;

fn make_view(data: &[f32]) -> View<'_, f32> {
    SimdView::new(data).expect("Scalar/Unaligned always succeeds")
}
fn make_view_mut(data: &mut [f32]) -> ViewMut<'_, f32> {
    SimdView::new_mut(data).expect("Scalar/Unaligned always succeeds")
}

// ---------------------------------------------------------------------------
// transform_in_place — all strategies
// ---------------------------------------------------------------------------

#[test]
fn test_transform_in_place_add() {
    let a_orig = [1.0f32, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0];
    let b = [10.0f32, 20.0, 30.0, 40.0, 50.0, 60.0, 70.0, 80.0, 90.0];
    let expected: Vec<f32> = a_orig.iter().zip(b.iter()).map(|(a, b)| a + b).collect();

    let mut a = a_orig;
    let b_view = make_view(&b);
    let mut a_view = make_view_mut(&mut a);
    a_view.transform_in_place(&b_view, Add).unwrap();

    assert_eq!(&a[..], expected.as_slice());
}

#[test]
fn test_transform_in_place_sub() {
    let a_orig = [10.0f32, 20.0, 30.0, 40.0, 50.0, 7.0, 8.0, 9.0, 11.0];
    let b = [1.0f32, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0];
    let expected: Vec<f32> = a_orig.iter().zip(b.iter()).map(|(a, b)| a - b).collect();

    let mut a = a_orig;
    let b_view = make_view(&b);
    let mut a_view = make_view_mut(&mut a);
    a_view.transform_in_place(&b_view, Sub).unwrap();

    assert_eq!(&a[..], expected.as_slice());
}

#[test]
fn test_transform_in_place_mul() {
    let a_orig = [1.0f32, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0];
    let b = [2.0f32, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0, 10.0];
    let expected: Vec<f32> = a_orig.iter().zip(b.iter()).map(|(a, b)| a * b).collect();

    let mut a = a_orig;
    let b_view = make_view(&b);
    let mut a_view = make_view_mut(&mut a);
    a_view.transform_in_place(&b_view, Mul).unwrap();

    assert_eq!(&a[..], expected.as_slice());
}

#[test]
fn test_transform_in_place_div() {
    let a_orig = [10.0f32, 20.0, 30.0, 40.0, 50.0, 60.0, 70.0, 80.0, 90.0];
    let b = [2.0f32, 4.0, 5.0, 8.0, 10.0, 6.0, 7.0, 8.0, 9.0];
    let expected: Vec<f32> = a_orig.iter().zip(b.iter()).map(|(a, b)| a / b).collect();

    let mut a = a_orig;
    let b_view = make_view(&b);
    let mut a_view = make_view_mut(&mut a);
    a_view.transform_in_place(&b_view, Div).unwrap();

    for (got, exp) in a.iter().zip(expected.iter()) {
        assert!((got - exp).abs() < 1e-5, "got={got}, expected={exp}");
    }
}

// ---------------------------------------------------------------------------
// transform_in_place — delegates match (regression check)
// ---------------------------------------------------------------------------

#[test]
fn test_add_assign_matches_transform_in_place_add() {
    let a_orig = [1.0f32, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0, 10.0];
    let b = [10.0f32; 10];

    let mut a_generic = a_orig;
    {
        let b_view = make_view(&b);
        let mut view = make_view_mut(&mut a_generic);
        view.transform_in_place(&b_view, Add).unwrap();
    }

    let mut a_delegate = a_orig;
    {
        let b_view = make_view(&b);
        let mut view = make_view_mut(&mut a_delegate);
        view.add_assign(&b_view).unwrap();
    }

    assert_eq!(a_generic, a_delegate);
}

#[test]
fn test_mul_assign_matches_transform_in_place_mul() {
    let a_orig = [1.0f32, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0, 10.0];
    let b = [3.0f32; 10];

    let mut a_generic = a_orig;
    {
        let b_view = make_view(&b);
        let mut view = make_view_mut(&mut a_generic);
        view.transform_in_place(&b_view, Mul).unwrap();
    }

    let mut a_delegate = a_orig;
    {
        let b_view = make_view(&b);
        let mut view = make_view_mut(&mut a_delegate);
        view.mul_assign(&b_view).unwrap();
    }

    assert_eq!(a_generic, a_delegate);
}

// ---------------------------------------------------------------------------
// zip_into — all strategies
// ---------------------------------------------------------------------------

#[test]
fn test_zip_into_add() {
    let a = [1.0f32, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0];
    let b = [10.0f32, 20.0, 30.0, 40.0, 50.0, 60.0, 70.0, 80.0, 90.0];
    let expected: Vec<f32> = a.iter().zip(b.iter()).map(|(a, b)| a + b).collect();

    let mut out = vec![0.0f32; 9];
    let a_view = make_view(&a);
    let b_view = make_view(&b);
    a_view.zip_into(&b_view, &mut out, Add).unwrap();

    assert_eq!(out, expected);
}

#[test]
fn test_zip_into_sub() {
    let a = [10.0f32, 20.0, 30.0, 40.0, 50.0, 60.0, 70.0, 80.0, 90.0];
    let b = [1.0f32, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0];
    let expected: Vec<f32> = a.iter().zip(b.iter()).map(|(a, b)| a - b).collect();

    let mut out = vec![0.0f32; 9];
    let a_view = make_view(&a);
    let b_view = make_view(&b);
    a_view.zip_into(&b_view, &mut out, Sub).unwrap();

    assert_eq!(out, expected);
}

#[test]
fn test_zip_into_mul_matches_elementwise_mul() {
    let a = [1.0f32, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0, 10.0, 11.0];
    let b = [2.0f32; 11];

    let mut out_generic = vec![0.0f32; 11];
    let mut out_delegate = vec![0.0f32; 11];

    {
        let a_view = make_view(&a);
        let b_view = make_view(&b);
        a_view.zip_into(&b_view, &mut out_generic, Mul).unwrap();
    }
    {
        let a_view = make_view(&a);
        let b_view = make_view(&b);
        a_view.elementwise_mul(&b_view, &mut out_delegate).unwrap();
    }

    assert_eq!(out_generic, out_delegate);
}

#[test]
fn test_zip_into_div() {
    let a = [10.0f32, 20.0, 30.0, 40.0];
    let b = [2.0f32, 4.0, 5.0, 8.0];
    let expected: Vec<f32> = a.iter().zip(b.iter()).map(|(a, b)| a / b).collect();

    let mut out = vec![0.0f32; 4];
    let a_view = make_view(&a);
    let b_view = make_view(&b);
    a_view.zip_into(&b_view, &mut out, Div).unwrap();

    for (got, exp) in out.iter().zip(expected.iter()) {
        assert!((got - exp).abs() < 1e-5, "got={got}, expected={exp}");
    }
}

// ---------------------------------------------------------------------------
// Error paths
// ---------------------------------------------------------------------------

#[test]
fn test_transform_in_place_length_mismatch() {
    let a = [1.0f32; 4];
    let b = [1.0f32; 5];
    let mut a = a;
    let b_view = make_view(&b);
    let mut a_view = make_view_mut(&mut a);
    let err = a_view.transform_in_place(&b_view, Add);
    assert_eq!(err, Err(SimdError::LengthMismatch));
}

#[test]
fn test_zip_into_length_mismatch() {
    let a = [1.0f32; 4];
    let b = [1.0f32; 5];
    let mut out = vec![0.0f32; 4];
    let a_view = make_view(&a);
    let b_view = make_view(&b);
    let err = a_view.zip_into(&b_view, &mut out, Add);
    assert_eq!(err, Err(SimdError::LengthMismatch));
}

#[test]
fn test_zip_into_output_too_short() {
    let a = [1.0f32; 8];
    let b = [1.0f32; 8];
    let mut out = vec![0.0f32; 4]; // too short
    let a_view = make_view(&a);
    let b_view = make_view(&b);
    let err = a_view.zip_into(&b_view, &mut out, Add);
    assert_eq!(err, Err(SimdError::InsufficientOutputLength));
}

// ---------------------------------------------------------------------------
// ZST size invariants
// ---------------------------------------------------------------------------

#[test]
fn test_strategy_zst_sizes() {
    use core::mem::size_of;
    assert_eq!(size_of::<Add>(), 0);
    assert_eq!(size_of::<Sub>(), 0);
    assert_eq!(size_of::<Mul>(), 0);
    assert_eq!(size_of::<Div>(), 0);
    assert_eq!(size_of::<BitAnd>(), 0);
    assert_eq!(size_of::<BitOr>(), 0);
    assert_eq!(size_of::<BitXor>(), 0);
}

// ---------------------------------------------------------------------------
// Tail-element correctness (len not divisible by LANE_COUNT)
// ---------------------------------------------------------------------------

#[test]
fn test_transform_in_place_odd_length() {
    // 5 elements with Scalar (lane=1) — exercises the scalar tail path.
    let a_orig = [1.0f32, 2.0, 3.0, 4.0, 5.0];
    let b = [10.0f32, 20.0, 30.0, 40.0, 50.0];
    let expected: Vec<f32> = a_orig.iter().zip(b.iter()).map(|(x, y)| x + y).collect();

    let mut a = a_orig;
    let b_view = make_view(&b);
    let mut a_view = make_view_mut(&mut a);
    a_view.transform_in_place(&b_view, Add).unwrap();

    assert_eq!(&a[..], expected.as_slice());
}

#[test]
fn test_zip_into_odd_length() {
    let a = [1.0f32, 2.0, 3.0, 4.0, 5.0];
    let b = [1.0f32, 2.0, 3.0, 4.0, 5.0];
    let expected: Vec<f32> = a.iter().zip(b.iter()).map(|(x, y)| x * y).collect();

    let mut out = vec![0.0f32; 5];
    let a_view = make_view(&a);
    let b_view = make_view(&b);
    a_view.zip_into(&b_view, &mut out, Mul).unwrap();

    assert_eq!(out, expected);
}