trueno 0.16.4

High-performance SIMD compute library with GPU support for matrix operations
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
288
289
290
291

use super::*;

#[test]
fn test_scalar_add() {
    let a = [1.0, 2.0, 3.0, 4.0];
    let b = [5.0, 6.0, 7.0, 8.0];
    let mut result = [0.0; 4];
    // SAFETY: Test code calling backend trait methods marked unsafe
    unsafe {
        ScalarBackend::add(&a, &b, &mut result);
    }
    assert_eq!(result, [6.0, 8.0, 10.0, 12.0]);
}

#[test]
fn test_scalar_mul() {
    let a = [1.0, 2.0, 3.0, 4.0];
    let b = [2.0, 3.0, 4.0, 5.0];
    let mut result = [0.0; 4];
    // SAFETY: Test code calling backend trait methods marked unsafe
    unsafe {
        ScalarBackend::mul(&a, &b, &mut result);
    }
    assert_eq!(result, [2.0, 6.0, 12.0, 20.0]);
}

#[test]
fn test_scalar_dot() {
    let a = [1.0, 2.0, 3.0];
    let b = [4.0, 5.0, 6.0];
    // SAFETY: Test code calling backend trait methods marked unsafe
    let result = unsafe { ScalarBackend::dot(&a, &b) };
    assert_eq!(result, 32.0); // 1*4 + 2*5 + 3*6 = 32
}

#[test]
fn test_scalar_sum() {
    let a = [1.0, 2.0, 3.0, 4.0];
    // SAFETY: Test code calling backend trait methods marked unsafe
    let result = unsafe { ScalarBackend::sum(&a) };
    assert_eq!(result, 10.0);
}

#[test]
fn test_scalar_max() {
    let a = [1.0, 5.0, 3.0, 2.0];
    // SAFETY: Test code calling backend trait methods marked unsafe
    let result = unsafe { ScalarBackend::max(&a) };
    assert_eq!(result, 5.0);
}

#[test]
fn test_scalar_min() {
    let a = [1.0, 5.0, 3.0, 2.0];
    // SAFETY: Test code calling backend trait methods marked unsafe
    let result = unsafe { ScalarBackend::min(&a) };
    assert_eq!(result, 1.0);
}

#[test]
fn test_scalar_sub() {
    let a = [5.0, 6.0, 7.0, 8.0];
    let b = [1.0, 2.0, 3.0, 4.0];
    let mut result = [0.0; 4];
    // SAFETY: Test code calling backend trait methods marked unsafe
    unsafe {
        ScalarBackend::sub(&a, &b, &mut result);
    }
    assert_eq!(result, [4.0, 4.0, 4.0, 4.0]);
}

#[test]
fn test_scalar_div() {
    let a = [10.0, 20.0, 30.0, 40.0];
    let b = [2.0, 4.0, 5.0, 8.0];
    let mut result = [0.0; 4];
    // SAFETY: Test code calling backend trait methods marked unsafe
    unsafe {
        ScalarBackend::div(&a, &b, &mut result);
    }
    assert_eq!(result, [5.0, 5.0, 6.0, 5.0]);
}

#[test]
fn test_scalar_argmax() {
    let a = [1.0, 5.0, 3.0, 2.0];
    // SAFETY: Test code calling backend trait methods marked unsafe
    let result = unsafe { ScalarBackend::argmax(&a) };
    assert_eq!(result, 1); // Index of 5.0
}

#[test]
fn test_scalar_argmin() {
    let a = [5.0, 1.0, 3.0, 2.0];
    // SAFETY: Test code calling backend trait methods marked unsafe
    let result = unsafe { ScalarBackend::argmin(&a) };
    assert_eq!(result, 1); // Index of 1.0
}

#[test]
fn test_scalar_sum_kahan() {
    let a = [1.0, 2.0, 3.0, 4.0];
    // SAFETY: Test code calling backend trait methods marked unsafe
    let result = unsafe { ScalarBackend::sum_kahan(&a) };
    assert_eq!(result, 10.0);
}

#[test]
fn test_scalar_norm_l1() {
    let a = [1.0, -2.0, 3.0, -4.0];
    // SAFETY: Test code calling backend trait methods marked unsafe
    let result = unsafe { ScalarBackend::norm_l1(&a) };
    assert_eq!(result, 10.0); // |1| + |-2| + |3| + |-4| = 10
}

#[test]
fn test_scalar_norm_l2() {
    let a = [3.0, 4.0];
    // SAFETY: Test code calling backend trait methods marked unsafe
    let result = unsafe { ScalarBackend::norm_l2(&a) };
    assert_eq!(result, 5.0); // sqrt(3² + 4²) = 5
}

#[test]
fn test_scalar_scale() {
    let a = [1.0, 2.0, 3.0, 4.0];
    let mut result = [0.0; 4];
    // SAFETY: Test code calling backend trait methods marked unsafe
    unsafe {
        ScalarBackend::scale(&a, 2.0, &mut result);
    }
    assert_eq!(result, [2.0, 4.0, 6.0, 8.0]);
}

#[test]
fn test_scalar_clamp() {
    let a = [1.0, 5.0, 10.0, 15.0];
    let mut result = [0.0; 4];
    // SAFETY: Test code calling backend trait methods marked unsafe
    unsafe {
        ScalarBackend::clamp(&a, 3.0, 12.0, &mut result);
    }
    assert_eq!(result, [3.0, 5.0, 10.0, 12.0]);
}

#[test]
fn test_scalar_lerp() {
    let a = [0.0, 10.0, 20.0];
    let b = [100.0, 110.0, 120.0];
    let mut result = [0.0; 3];
    // SAFETY: Test code calling backend trait methods marked unsafe
    unsafe {
        ScalarBackend::lerp(&a, &b, 0.5, &mut result);
    }
    assert_eq!(result, [50.0, 60.0, 70.0]); // Midpoint between a and b
}

#[test]
fn test_scalar_fma() {
    let a = [1.0, 2.0, 3.0];
    let b = [2.0, 3.0, 4.0];
    let c = [5.0, 6.0, 7.0];
    let mut result = [0.0; 3];
    // SAFETY: Test code calling backend trait methods marked unsafe
    unsafe {
        ScalarBackend::fma(&a, &b, &c, &mut result);
    }
    // FMA: a*b + c
    assert_eq!(result, [7.0, 12.0, 19.0]); // [1*2+5, 2*3+6, 3*4+7]
}

#[test]
fn test_scalar_relu() {
    let a = [-3.0, -1.0, 0.0, 1.0, 3.0];
    let mut result = [0.0; 5];
    // SAFETY: Test code calling backend trait methods marked unsafe
    unsafe {
        ScalarBackend::relu(&a, &mut result);
    }
    assert_eq!(result, [0.0, 0.0, 0.0, 1.0, 3.0]);
}

#[test]
fn test_scalar_sigmoid() {
    let a = [-51.0, -1.0, 0.0, 1.0, 51.0];
    let mut result = [0.0; 5];
    // SAFETY: Test code calling backend trait methods marked unsafe
    unsafe {
        ScalarBackend::sigmoid(&a, &mut result);
    }
    // sigmoid(-51) = 0, sigmoid(0) = 0.5, sigmoid(51) = 1
    assert_eq!(result[0], 0.0); // Clamped to 0 for numerical stability
    assert!((result[1] - 0.2689).abs() < 0.001); // sigmoid(-1)
    assert_eq!(result[2], 0.5); // sigmoid(0)
    assert!((result[3] - 0.7311).abs() < 0.001); // sigmoid(1)
    assert_eq!(result[4], 1.0); // Clamped to 1 for numerical stability
}

#[test]
fn test_scalar_gelu() {
    let a = [-2.0, -1.0, 0.0, 1.0, 2.0];
    let mut result = [0.0; 5];
    // SAFETY: Test code calling backend trait methods marked unsafe
    unsafe {
        ScalarBackend::gelu(&a, &mut result);
    }
    // GELU approximation values
    assert!((result[0] - (-0.0454)).abs() < 0.01); // gelu(-2)
    assert!((result[1] - (-0.1588)).abs() < 0.01); // gelu(-1)
    assert_eq!(result[2], 0.0); // gelu(0) = 0
    assert!((result[3] - 0.8413).abs() < 0.01); // gelu(1)
    assert!((result[4] - 1.9545).abs() < 0.01); // gelu(2)
}

#[test]
fn test_scalar_swish() {
    let a = [-51.0, -1.0, 0.0, 1.0, 51.0];
    let mut result = [0.0; 5];
    // SAFETY: Test code calling backend trait methods marked unsafe
    unsafe {
        ScalarBackend::swish(&a, &mut result);
    }
    // swish(x) = x * sigmoid(x)
    assert_eq!(result[0], 0.0); // x * 0 = 0 (numerical stability)
    assert!((result[1] - (-0.2689)).abs() < 0.001); // -1 * sigmoid(-1)
    assert_eq!(result[2], 0.0); // 0 * sigmoid(0) = 0
    assert!((result[3] - 0.7311).abs() < 0.001); // 1 * sigmoid(1)
    assert_eq!(result[4], 51.0); // x * 1 = x (numerical stability)
}

// cuda-tile-behavior.md: Falsification test #81 - Backend equivalence
#[test]
fn test_scalar_dot_unrolled_various_sizes() {
    // Test various sizes to exercise all code paths in unrolled implementation:
    // - 0 elements (edge case)
    // - 1-3 elements (remainder only)
    // - 4 elements (exactly one unrolled chunk)
    // - 5-7 elements (one chunk + remainder)
    // - 8 elements (two chunks)
    // - 100 elements (realistic workload)
    // - 1000 elements (larger workload)
    let sizes = [0, 1, 2, 3, 4, 5, 6, 7, 8, 100, 1000];

    for &size in &sizes {
        if size == 0 {
            continue; // Empty slice edge case
        }

        let a: Vec<f32> = (0..size).map(|i| (i as f32) * 0.1).collect();
        let b: Vec<f32> = (0..size).map(|i| ((size - i) as f32) * 0.1).collect();

        // Calculate expected result using naive implementation
        let expected: f32 = a.iter().zip(&b).map(|(x, y)| x * y).sum();
        // SAFETY: CPU feature verified at runtime, slices bounds-checked
        let result = unsafe { ScalarBackend::dot(&a, &b) };

        // Tolerance accounts for FP reordering from unrolling (different accumulator summing order)
        // Use relative tolerance for larger sums where absolute error grows with magnitude
        let tolerance = (1e-5 * expected.abs()).max(1e-4);
        assert!(
            (result - expected).abs() < tolerance,
            "dot mismatch at size {}: got={}, expected={}, tolerance={}",
            size,
            result,
            expected,
            tolerance
        );
    }
}

// cuda-tile-behavior.md: Falsification test #92 - FMA single-rounding accuracy
#[test]
fn test_scalar_dot_mul_add_accuracy() {
    // Test that mul_add provides better accuracy than separate mul+add
    // FMA has single rounding vs two roundings for separate operations
    let a = vec![1.0000001_f32; 1000];
    let b = vec![1.0000001_f32; 1000];

    // SAFETY: CPU feature verified at runtime, slices bounds-checked
    let result = unsafe { ScalarBackend::dot(&a, &b) };

    // Expected: 1000 * 1.0000001 * 1.0000001 ≈ 1000.0002
    // With FMA, error should be smaller due to single rounding
    let expected = 1000.0 * 1.0000001_f32 * 1.0000001_f32;
    assert!(
        (result - expected).abs() < 1e-3,
        "FMA accuracy test: got={}, expected={}",
        result,
        expected
    );
}