tenflowers-core 0.1.1

Core tensor operations and execution engine for TenfloweRS
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
288
289
290

//! Tests for Binary Operations
//!
//! Comprehensive test suite for binary operations including same-shape operations,
//! broadcasting, SIMD optimizations, and performance verification.

#[cfg(test)]
#[allow(irrefutable_let_patterns)] // Pattern matching on TensorStorage is irrefutable when GPU feature is disabled
mod tests {
    use super::super::convenience::*;
    use crate::tensor::TensorStorage;
    use crate::Tensor;

    #[test]
    fn test_add_same_shape() {
        let a = Tensor::<f32>::from_vec(vec![1.0, 2.0, 3.0], &[3])
            .expect("test: from_vec should succeed");
        let b = Tensor::<f32>::from_vec(vec![4.0, 5.0, 6.0], &[3])
            .expect("test: from_vec should succeed");

        let c = add(&a, &b).expect("test: add should succeed");
        let expected = vec![5.0, 7.0, 9.0];

        // Compare results
        if let TensorStorage::Cpu(arr) = &c.storage {
            assert_eq!(
                arr.as_slice().expect("tensor should be contiguous"),
                &expected
            );
        }
    }

    #[test]
    fn test_mul_broadcast() {
        let a = Tensor::<f32>::from_vec(vec![1.0, 2.0, 3.0], &[3, 1])
            .expect("test: from_vec should succeed");
        let b = Tensor::<f32>::from_vec(vec![2.0, 3.0], &[1, 2])
            .expect("test: from_vec should succeed");

        let c = mul(&a, &b).expect("test: mul should succeed");
        assert_eq!(c.shape().dims(), &[3, 2]);

        // Expected: [[2, 3], [4, 6], [6, 9]]
        let expected = vec![2.0, 3.0, 4.0, 6.0, 6.0, 9.0];
        if let TensorStorage::Cpu(arr) = &c.storage {
            assert_eq!(
                arr.as_slice().expect("tensor should be contiguous"),
                &expected
            );
        }
    }

    #[test]
    fn test_pow() {
        let a = Tensor::<f32>::from_vec(vec![2.0, 3.0, 4.0], &[3])
            .expect("test: from_vec should succeed");
        let b = Tensor::<f32>::from_vec(vec![2.0, 2.0, 2.0], &[3])
            .expect("test: from_vec should succeed");

        let c = pow(&a, &b).expect("test: pow should succeed");
        let expected = vec![4.0, 9.0, 16.0];

        if let TensorStorage::Cpu(arr) = &c.storage {
            assert_eq!(
                arr.as_slice().expect("tensor should be contiguous"),
                &expected
            );
        }

        // Test broadcasting
        let scalar =
            Tensor::<f32>::from_vec(vec![3.0], &[1]).expect("test: from_vec should succeed");
        let d = pow(&a, &scalar).expect("test: pow should succeed");
        let expected = vec![8.0, 27.0, 64.0];

        if let TensorStorage::Cpu(arr) = &d.storage {
            assert_eq!(
                arr.as_slice().expect("tensor should be contiguous"),
                &expected
            );
        }
    }

    #[test]
    #[cfg(feature = "simd")]
    fn test_simd_add_optimization() {
        use crate::Device;

        // Test SIMD-optimized addition for f32 tensors
        let a = Tensor::from_vec(vec![1.0f32, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0], &[8])
            .expect("test: from_vec should succeed");
        let b = Tensor::from_vec(vec![2.0f32, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0], &[8])
            .expect("test: from_vec should succeed");

        let result = add(&a, &b).expect("test: add should succeed");
        let expected = vec![3.0f32, 5.0, 7.0, 9.0, 11.0, 13.0, 15.0, 17.0];

        if let TensorStorage::Cpu(arr) = &result.storage {
            assert_eq!(
                arr.as_slice().expect("tensor should be contiguous"),
                &expected
            );
        }
    }

    #[test]
    #[cfg(feature = "simd")]
    fn test_simd_mul_optimization() {
        use crate::Device;

        // Test SIMD-optimized multiplication for f32 tensors
        let a = Tensor::from_vec(vec![2.0f32, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0], &[8])
            .expect("test: from_vec should succeed");
        let b = Tensor::from_vec(vec![1.5f32, 2.0, 2.5, 3.0, 3.5, 4.0, 4.5, 5.0], &[8])
            .expect("test: from_vec should succeed");

        let result = mul(&a, &b).expect("test: mul should succeed");
        let expected = vec![3.0f32, 6.0, 10.0, 15.0, 21.0, 28.0, 36.0, 45.0];

        if let TensorStorage::Cpu(arr) = &result.storage {
            assert_eq!(
                arr.as_slice().expect("tensor should be contiguous"),
                &expected
            );
        }
    }

    #[test]
    #[cfg(feature = "autograd")]
    fn test_memory_profiling_integration() {
        // NOTE(v0.2): Implement profiler integration
        // if let Ok(mut p) = get_profiler().lock() {
        //     p.operations.clear();
        //     p.allocations = 0;
        // }

        // Perform a binary operation
        let a = Tensor::from_vec(vec![1.0f32, 2.0, 3.0, 4.0], &[4])
            .expect("test: from_vec should succeed");
        let b = Tensor::from_vec(vec![2.0f32, 3.0, 4.0, 5.0], &[4])
            .expect("test: from_vec should succeed");
        let _result = add(&a, &b).expect("test: add should succeed");

        // Check that simple profiling captured the operation
        // if let Ok(p) = get_profiler().lock() {
        //     // Should have recorded the binary_Add operation
        //     assert!(p.operations.contains_key("binary_Add"));
        //     // Should have recorded at least one allocation
        //     assert!(p.allocations > 0);
        // }
    }

    #[test]
    fn test_subtraction() {
        let a = Tensor::<f32>::from_vec(vec![5.0, 7.0, 9.0], &[3])
            .expect("test: from_vec should succeed");
        let b = Tensor::<f32>::from_vec(vec![1.0, 2.0, 3.0], &[3])
            .expect("test: from_vec should succeed");

        let c = sub(&a, &b).expect("test: sub should succeed");
        let expected = vec![4.0, 5.0, 6.0];

        if let TensorStorage::Cpu(arr) = &c.storage {
            assert_eq!(
                arr.as_slice().expect("tensor should be contiguous"),
                &expected
            );
        }
    }

    #[test]
    fn test_division() {
        let a = Tensor::<f32>::from_vec(vec![6.0, 8.0, 10.0], &[3])
            .expect("test: from_vec should succeed");
        let b = Tensor::<f32>::from_vec(vec![2.0, 4.0, 5.0], &[3])
            .expect("test: from_vec should succeed");

        let c = div(&a, &b).expect("test: div should succeed");
        let expected = vec![3.0, 2.0, 2.0];

        if let TensorStorage::Cpu(arr) = &c.storage {
            assert_eq!(
                arr.as_slice().expect("tensor should be contiguous"),
                &expected
            );
        }
    }

    #[test]
    fn test_min_max() {
        let a = Tensor::<f32>::from_vec(vec![1.0, 5.0, 3.0], &[3])
            .expect("test: from_vec should succeed");
        let b = Tensor::<f32>::from_vec(vec![2.0, 3.0, 4.0], &[3])
            .expect("test: from_vec should succeed");

        let min_result = min(&a, &b).expect("test: min should succeed");
        let expected_min = vec![1.0, 3.0, 3.0];

        if let TensorStorage::Cpu(arr) = &min_result.storage {
            assert_eq!(
                arr.as_slice().expect("tensor should be contiguous"),
                &expected_min
            );
        }

        let max_result = max(&a, &b).expect("test: max should succeed");
        let expected_max = vec![2.0, 5.0, 4.0];

        if let TensorStorage::Cpu(arr) = &max_result.storage {
            assert_eq!(
                arr.as_slice().expect("tensor should be contiguous"),
                &expected_max
            );
        }
    }

    #[test]
    fn test_scalar_add() {
        let tensor = Tensor::<f32>::from_vec(vec![1.0, 2.0, 3.0], &[3])
            .expect("test: from_vec should succeed");
        let result = scalar_add(&tensor, 5.0).expect("test: scalar_add should succeed");
        let expected = vec![6.0, 7.0, 8.0];

        if let TensorStorage::Cpu(arr) = &result.storage {
            assert_eq!(
                arr.as_slice().expect("tensor should be contiguous"),
                &expected
            );
        }
    }

    #[test]
    fn test_clamp() {
        let tensor = Tensor::<f32>::from_vec(vec![-1.0, 0.5, 2.0, 5.0], &[4])
            .expect("test: from_vec should succeed");
        let result = clamp(&tensor, 0.0, 3.0).expect("test: clamp should succeed");
        let expected = vec![0.0, 0.5, 2.0, 3.0];

        if let TensorStorage::Cpu(arr) = &result.storage {
            assert_eq!(
                arr.as_slice().expect("tensor should be contiguous"),
                &expected
            );
        }
    }

    #[test]
    fn test_ultra_performance_functions() {
        let a = Tensor::<f32>::from_vec(vec![1.0, 2.0, 3.0], &[3])
            .expect("test: from_vec should succeed");
        let b = Tensor::<f32>::from_vec(vec![4.0, 5.0, 6.0], &[3])
            .expect("test: from_vec should succeed");

        // Test ultra functions (they should behave the same but with metrics)
        let ultra_add_result = ultra_add(&a, &b).expect("test: ultra_add should succeed");
        let ultra_mul_result = ultra_mul(&a, &b).expect("test: ultra_mul should succeed");
        let ultra_sub_result = ultra_sub(&b, &a).expect("test: ultra_sub should succeed");
        let ultra_div_result = ultra_div(&b, &a).expect("test: ultra_div should succeed");

        // Expected results
        let expected_add = vec![5.0, 7.0, 9.0];
        let expected_mul = vec![4.0, 10.0, 18.0];
        let expected_sub = vec![3.0, 3.0, 3.0];
        let expected_div = vec![4.0, 2.5, 2.0];

        if let TensorStorage::Cpu(arr) = &ultra_add_result.storage {
            assert_eq!(
                arr.as_slice().expect("tensor should be contiguous"),
                &expected_add
            );
        }
        if let TensorStorage::Cpu(arr) = &ultra_mul_result.storage {
            assert_eq!(
                arr.as_slice().expect("tensor should be contiguous"),
                &expected_mul
            );
        }
        if let TensorStorage::Cpu(arr) = &ultra_sub_result.storage {
            assert_eq!(
                arr.as_slice().expect("tensor should be contiguous"),
                &expected_sub
            );
        }
        if let TensorStorage::Cpu(arr) = &ultra_div_result.storage {
            assert_eq!(
                arr.as_slice().expect("tensor should be contiguous"),
                &expected_div
            );
        }
    }
}