trueno 0.18.0

High-performance SIMD compute library with GPU support, LLM inference engine, and GGUF model loading
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

use super::super::*;

#[test]
fn test_token_budget_from_latency() {
    let budget = TokenBudget::from_latency(50.0);
    assert!((budget.us_per_token - 50.0).abs() < 0.001);
    assert!((budget.tokens_per_sec - 20_000.0).abs() < 1.0);
}

#[test]
fn test_token_budget_from_throughput() {
    let budget = TokenBudget::from_throughput(20_000.0);
    assert!((budget.us_per_token - 50.0).abs() < 0.001);
    assert!((budget.tokens_per_sec - 20_000.0).abs() < 1.0);
}

#[test]
fn test_token_budget_is_met() {
    let budget = TokenBudget::from_latency(50.0);
    assert!(budget.is_met(40.0)); // Under budget
    assert!(budget.is_met(50.0)); // Exactly at budget
    assert!(!budget.is_met(60.0)); // Over budget
}

#[test]
fn test_dot_op() {
    let op = DotOp::new(4);
    let a = vec![1.0, 2.0, 3.0, 4.0];
    let b = vec![5.0, 6.0, 7.0, 8.0];
    let result = op.execute((a, b), Backend::Scalar).unwrap();
    assert!((result - 70.0).abs() < 0.001); // 1*5 + 2*6 + 3*7 + 4*8 = 70
}

#[test]
fn test_add_op() {
    let op = AddOp::new(4);
    let a = vec![1.0, 2.0, 3.0, 4.0];
    let b = vec![5.0, 6.0, 7.0, 8.0];
    let result = op.execute((a, b), Backend::Scalar).unwrap();
    assert_eq!(result, vec![6.0, 8.0, 10.0, 12.0]);
}

#[test]
fn test_matmul_op() {
    let op = MatmulOp::new(2, 2, 2);
    // A = [[1, 2], [3, 4]]
    let a = vec![1.0, 2.0, 3.0, 4.0];
    // B = [[5, 6], [7, 8]]
    let b = vec![5.0, 6.0, 7.0, 8.0];
    let result = op.execute((a, b), Backend::Scalar).unwrap();
    // C = [[19, 22], [43, 50]]
    assert_eq!(result, vec![19.0, 22.0, 43.0, 50.0]);
}

// =========================================================================
// FALSIFY-MM: matmul-kernel-v1.yaml contract (trueno MatmulOp)
//
// Five-Whys (PMAT-354):
//   Why 1: trueno had 1 matmul unit test but zero FALSIFY-MM-* tests
//   Why 2: unit tests verify one 2x2 case, not mathematical invariants
//   Why 3: no mapping from matmul-kernel-v1.yaml to trueno test names
//   Why 4: trueno predates the provable-contracts YAML convention
//   Why 5: matmul was "obviously correct" (textbook GEMM)
//
// References:
//   - provable-contracts/contracts/matmul-kernel-v1.yaml
// =========================================================================

/// FALSIFY-MM-001: Shape correctness — output is [m, n]
#[test]
fn falsify_mm_001_shape_correctness() {
    for (m, k, n) in [(2, 3, 4), (1, 5, 1), (4, 4, 4), (3, 1, 2)] {
        let op = MatmulOp::new(m, k, n);
        let a = vec![1.0; m * k];
        let b = vec![1.0; k * n];
        let result = op.execute((a, b), Backend::Scalar).unwrap();
        assert_eq!(
            result.len(),
            m * n,
            "FALSIFIED MM-001: output len = {}, expected {} for ({m}x{k}) @ ({k}x{n})",
            result.len(),
            m * n
        );
    }
}

/// FALSIFY-MM-002: Numerical accuracy — known result for 2x3 @ 3x2
#[test]
fn falsify_mm_002_numerical_accuracy() {
    let op = MatmulOp::new(2, 3, 2);
    let a = vec![1.0, 2.0, 3.0, 4.0, 5.0, 6.0];
    let b = vec![7.0, 8.0, 9.0, 10.0, 11.0, 12.0];
    let result = op.execute((a, b), Backend::Scalar).unwrap();
    let expected = vec![58.0, 64.0, 139.0, 154.0];
    for (i, (&got, &exp)) in result.iter().zip(expected.iter()).enumerate() {
        assert!((got - exp).abs() < 1e-4, "FALSIFIED MM-002: result[{i}] = {got}, expected {exp}");
    }
}

/// FALSIFY-MM-005: Identity matrix — A @ I = A and I @ B = B
#[test]
fn falsify_mm_005_identity_matrix() {
    // Test A @ I = A
    let m = 3;
    let k = 4;
    let op = MatmulOp::new(m, k, k);
    let a: Vec<f32> = (0..m * k).map(|i| (i as f32 + 1.0) * 0.5).collect();
    let mut identity = vec![0.0; k * k];
    for i in 0..k {
        identity[i * k + i] = 1.0;
    }
    let result = op.execute((a.clone(), identity), Backend::Scalar).unwrap();
    for (i, (&got, &exp)) in result.iter().zip(a.iter()).enumerate() {
        assert!((got - exp).abs() < 1e-5, "FALSIFIED MM-005: (A@I)[{i}] = {got}, expected {exp}");
    }

    // Test I @ B = B
    let op2 = MatmulOp::new(k, k, m);
    let b: Vec<f32> = (0..k * m).map(|i| (i as f32 + 1.0) * 0.3).collect();
    let mut identity2 = vec![0.0; k * k];
    for i in 0..k {
        identity2[i * k + i] = 1.0;
    }
    let result2 = op2.execute((identity2, b.clone()), Backend::Scalar).unwrap();
    for (i, (&got, &exp)) in result2.iter().zip(b.iter()).enumerate() {
        assert!((got - exp).abs() < 1e-5, "FALSIFIED MM-005: (I@B)[{i}] = {got}, expected {exp}");
    }
}

mod mm_proptest_falsify {
    use super::super::super::*;
    use proptest::prelude::*;

    // FALSIFY-MM-001-prop: Shape correctness for random dimensions
    proptest! {
        #![proptest_config(ProptestConfig::with_cases(100))]

        #[test]
        fn falsify_mm_001_prop_shape(
            m in 1..=8usize,
            k in 1..=8usize,
            n in 1..=8usize,
        ) {
            let op = MatmulOp::new(m, k, n);
            let a = vec![1.0; m * k];
            let b = vec![1.0; k * n];
            let result = op.execute((a, b), Backend::Scalar).unwrap();
            prop_assert_eq!(result.len(), m * n);
        }
    }

    // FALSIFY-MM-005-prop: Identity matrix for random dims
    proptest! {
        #![proptest_config(ProptestConfig::with_cases(50))]

        #[test]
        fn falsify_mm_005_prop_identity(
            m in 1..=6usize,
            k in 1..=6usize,
            seed in 0..500u32,
        ) {
            let op = MatmulOp::new(m, k, k);
            let a: Vec<f32> = (0..m * k)
                .map(|i| ((i as f32 + seed as f32) * 0.37).sin())
                .collect();
            let mut identity = vec![0.0; k * k];
            for i in 0..k {
                identity[i * k + i] = 1.0;
            }
            let result = op.execute((a.clone(), identity), Backend::Scalar).unwrap();
            for (i, (&got, &exp)) in result.iter().zip(a.iter()).enumerate() {
                prop_assert!(
                    (got - exp).abs() < 1e-4,
                    "FALSIFIED MM-005-prop: (A@I)[{}] = {}, expected {}",
                    i, got, exp
                );
            }
        }
    }
}

#[test]
fn test_softmax_op() {
    let op = SoftmaxOp::new(3);
    let input = vec![1.0, 2.0, 3.0];
    let result = op.execute(input, Backend::Scalar).unwrap();
    // Sum should be 1.0
    let sum: f32 = result.iter().sum();
    assert!((sum - 1.0).abs() < 0.001);
    // Values should be increasing
    assert!(result[0] < result[1]);
    assert!(result[1] < result[2]);
}