leto 0.45.2

N-dimensional strided arrays, zero-copy views, and storage backends
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

#![expect(
    clippy::unwrap_used,
    reason = "test scope: failed precondition = test failure"
)]

use leto::{Array, Layout, LetoError, SliceArg, Storage, VecStorage};

#[test]
fn test_transpose() {
    let layout = Layout::c_contiguous([2, 3]).unwrap();
    let storage = VecStorage::new(vec![1, 2, 3, 4, 5, 6]);
    let array = Array::new(layout, storage).unwrap();

    let transposed = array.transpose([1, 0]).unwrap();
    assert_eq!(transposed.shape(), [3, 2]);
    assert_eq!(transposed.strides(), [1, 3]);

    // Transposed layout:
    // [0, 0] -> physical offset 0 (val 1)
    // [0, 1] -> physical offset 3 (val 4)
    // [1, 0] -> physical offset 1 (val 2)
    // [2, 1] -> physical offset 5 (val 6)
    assert_eq!(*transposed.get([0, 0]).unwrap(), 1);
    assert_eq!(*transposed.get([0, 1]).unwrap(), 4);
    assert_eq!(*transposed.get([1, 0]).unwrap(), 2);
    assert_eq!(*transposed.get([2, 1]).unwrap(), 6);
}

#[test]
fn test_broadcasting() {
    let layout = Layout::c_contiguous([1, 3]).unwrap();
    let storage = VecStorage::new(vec![100, 200, 300]);
    let array = Array::new(layout, storage).unwrap();

    let broadcasted = array.broadcast([2, 3]).unwrap();
    assert_eq!(broadcasted.shape(), [2, 3]);
    assert_eq!(broadcasted.strides(), [0, 1]); // Stride of axis 0 is 0

    assert_eq!(*broadcasted.get([0, 0]).unwrap(), 100);
    assert_eq!(*broadcasted.get([1, 0]).unwrap(), 100);
    assert_eq!(*broadcasted.get([0, 2]).unwrap(), 300);
    assert_eq!(*broadcasted.get([1, 2]).unwrap(), 300);
}

#[test]
fn test_mutable_broadcast_rejects_zero_stride_write_aliasing() {
    let layout = Layout::c_contiguous([1, 3]).unwrap();
    let storage = VecStorage::new(vec![100, 200, 300]);
    let mut array = Array::new(layout, storage).unwrap();

    let result = array.view_mut().broadcast_mut([2, 3]);
    assert!(matches!(
        result,
        Err(LetoError::IncompatibleBroadcast { .. })
    ));
}

#[test]
fn test_mutable_broadcast_permits_non_aliasing_same_shape() {
    let layout = Layout::c_contiguous([2, 3]).unwrap();
    let storage = VecStorage::new(vec![1, 2, 3, 4, 5, 6]);
    let mut array = Array::new(layout, storage).unwrap();

    let mut view = array.view_mut().broadcast_mut([2, 3]).unwrap();
    *view.get_mut([1, 2]).unwrap() = 60;

    assert_eq!(*view.get([1, 2]).unwrap(), 60);
}

#[test]
fn test_as_slice_exposes_offset_contiguous_subview() {
    // A row sliced out of a C-contiguous matrix is dense at a non-zero offset.
    let layout = Layout::c_contiguous([3, 4]).unwrap();
    let storage = VecStorage::new((0i32..12).collect::<Vec<_>>());
    let array = Array::new(layout, storage).unwrap();

    // Select row index 1 -> shape [4] starting at physical offset 4.
    let row = array
        .view()
        .slice_with::<1>(&[SliceArg::Index(1), SliceArg::All])
        .unwrap();
    assert_eq!(row.offset(), 4);
    assert!(!row.is_c_contiguous());
    assert!(row.is_contiguous());
    // Canonical as_slice now exposes the dense block independent of offset.
    assert_eq!(row.as_slice().unwrap(), &[4, 5, 6, 7]);
    assert_eq!(row.as_slice_memory_order().unwrap(), &[4, 5, 6, 7]);
}

#[test]
fn test_as_slice_memory_order_accepts_fortran_block() {
    // A pure transpose of a contiguous matrix is F-dense but not C-dense.
    let layout = Layout::c_contiguous([2, 3]).unwrap();
    let storage = VecStorage::new(vec![1, 2, 3, 4, 5, 6]);
    let array = Array::new(layout, storage).unwrap();
    let transposed = array.transpose([1, 0]).unwrap();

    assert!(transposed.as_slice().is_none());
    assert!(transposed.is_f_contiguous());
    assert_eq!(
        transposed.as_slice_memory_order().unwrap(),
        &[1, 2, 3, 4, 5, 6]
    );
}

#[test]
fn test_as_slice_rejects_strided_gap() {
    // Every other column is not a dense block in any memory order.
    let layout = Layout::c_contiguous([2, 4]).unwrap();
    let storage = VecStorage::new((0i32..8).collect::<Vec<_>>());
    let array = Array::new(layout, storage).unwrap();
    let strided = array
        .view()
        .slice_with::<2>(&[SliceArg::All, SliceArg::range(Some(0), Some(4), 2)])
        .unwrap();

    assert!(!strided.is_contiguous());
    assert!(strided.as_slice().is_none());
    assert!(strided.as_slice_memory_order().is_none());
}

#[test]
fn test_as_mut_slice_memory_order_writes_offset_block() {
    let layout = Layout::c_contiguous([3, 2]).unwrap();
    let storage = VecStorage::new((0i32..6).collect::<Vec<_>>());
    let mut array = Array::new(layout, storage).unwrap();

    {
        let mut row = array
            .view_mut()
            .slice_with_mut::<1>(&[SliceArg::Index(2), SliceArg::All])
            .unwrap();
        let block = row.as_mut_slice_memory_order().unwrap();
        assert_eq!(block, &[4, 5]);
        block[0] = 40;
        block[1] = 50;
    }

    assert_eq!(array.storage().as_slice(), &[0, 1, 2, 3, 40, 50]);
}

#[test]
fn test_reshape_reinterprets_dense_row_major_layout_without_copying() {
    let array = Array::from_shape_vec([2, 3], vec![1, 2, 3, 4, 5, 6]).unwrap();

    let reshaped = array.reshape([3, 2]).unwrap();

    assert_eq!(reshaped.shape(), [3, 2]);
    assert_eq!(reshaped.strides(), [2, 1]);
    assert_eq!(*reshaped.get([0, 0]).unwrap(), 1);
    assert_eq!(*reshaped.get([1, 1]).unwrap(), 4);
    assert_eq!(*reshaped.get([2, 1]).unwrap(), 6);
    assert_eq!(reshaped.as_slice().unwrap(), &[1, 2, 3, 4, 5, 6]);
}

#[test]
fn test_into_shape_preserves_owned_storage_and_new_rank() {
    let array = Array::from_shape_vec([2, 3], vec![1, 2, 3, 4, 5, 6]).unwrap();

    let reshaped = array.into_shape([6]).unwrap();

    assert_eq!(reshaped.shape(), [6]);
    assert_eq!(reshaped.strides(), [1]);
    assert_eq!(reshaped.storage().as_slice(), &[1, 2, 3, 4, 5, 6]);
}

#[test]
fn test_reshape_rejects_shape_size_mismatch_and_strided_layouts() {
    let array = Array::from_shape_vec([2, 3], vec![1, 2, 3, 4, 5, 6]).unwrap();
    let wrong_size = array.reshape([4, 2]);
    assert!(matches!(wrong_size, Err(LetoError::ShapeMismatch { .. })));

    let transposed = array.transpose([1, 0]).unwrap();
    let strided_reshape = transposed.reshape([6]);
    assert!(matches!(
        strided_reshape,
        Err(LetoError::StorageError { .. })
    ));
}

#[test]
fn test_reshape_mut_writes_through_dense_row_major_view() {
    let mut array = Array::from_shape_vec([2, 3], vec![1, 2, 3, 4, 5, 6]).unwrap();

    {
        let mut reshaped = array.reshape_mut([3, 2]).unwrap();
        *reshaped.get_mut([2, 0]).unwrap() = 50;
    }

    assert_eq!(array.storage().as_slice(), &[1, 2, 3, 4, 50, 6]);
}

#[test]
fn test_permute_is_named_transpose_alias() {
    let array = Array::from_shape_vec([2, 3], vec![1, 2, 3, 4, 5, 6]).unwrap();

    let permuted = array.permute([1, 0]).unwrap();
    let transposed = array.transpose([1, 0]).unwrap();

    assert_eq!(permuted.shape(), transposed.shape());
    assert_eq!(permuted.strides(), transposed.strides());
    assert_eq!(*permuted.get([2, 1]).unwrap(), 6);
}

#[test]
fn test_to_contiguous_materializes_logical_row_major_order() {
    let array = Array::from_shape_vec([3, 4], (0i32..12).collect::<Vec<_>>()).unwrap();
    let strided = array
        .slice_with::<2>(&[SliceArg::All, SliceArg::range(Some(0), Some(4), 2)])
        .unwrap();

    let contiguous = strided.to_contiguous();

    assert_eq!(contiguous.shape(), [3, 2]);
    assert_eq!(contiguous.strides(), [2, 1]);
    assert_eq!(contiguous.storage().as_slice(), &[0, 2, 4, 6, 8, 10]);
}

#[test]
fn test_to_contiguous_materializes_broadcasted_views() {
    let row = Array::from_shape_vec([1, 3], vec![10, 20, 30]).unwrap();
    let broadcast = row.broadcast([2, 3]).unwrap();

    let contiguous = broadcast.to_contiguous();

    assert_eq!(contiguous.shape(), [2, 3]);
    assert_eq!(contiguous.storage().as_slice(), &[10, 20, 30, 10, 20, 30]);
}

#[test]
fn test_to_contiguous_materializes_f_dense_views_through_tiled_transpose() {
    // A transposed C-order matrix is F-dense and takes the tiled kernel; the
    // result must match the logical row-major odometer materialization.
    let source = Array::from_shape_vec([3, 4], (0..12).collect::<Vec<i32>>()).unwrap();
    let transposed = source.view().transpose([1, 0]).unwrap();
    assert!(transposed.is_f_dense());

    let contiguous = transposed.to_contiguous();

    assert_eq!(contiguous.shape(), [4, 3]);
    let expected: Vec<i32> = (0..4)
        .flat_map(|r| (0..3).map(move |c| *transposed.get([r, c]).unwrap()))
        .collect();
    assert_eq!(contiguous.storage().as_slice(), expected.as_slice());

    // Larger-than-tile plane exercises the blocked loops across tile seams.
    let big = Array::from_shape_vec([65, 130], (0..65 * 130).collect::<Vec<i32>>()).unwrap();
    let big_t = big.view().transpose([1, 0]).unwrap();
    let big_contig = big_t.to_contiguous();
    for r in (0..130).step_by(37) {
        for c in (0..65).step_by(13) {
            assert_eq!(
                *big_contig.get([r, c]).unwrap(),
                *big_t.get([r, c]).unwrap(),
                "tiled transpose diverges at [{r},{c}]"
            );
        }
    }
}

#[test]
fn test_scaled_add_matches_reference_on_dense_and_strided_pairs() {
    // Matching C-dense pair takes the zipped slice pass.
    let mut lhs = Array::from_shape_vec([2, 3], vec![1.0f64, 2.0, 3.0, 4.0, 5.0, 6.0]).unwrap();
    let rhs = Array::from_shape_vec([2, 3], vec![10.0f64, 20.0, 30.0, 40.0, 50.0, 60.0]).unwrap();
    lhs.scaled_add(0.5, &rhs);
    assert_eq!(
        lhs.storage().as_slice(),
        &[6.0, 12.0, 18.0, 24.0, 30.0, 36.0]
    );

    // Mismatched layouts (C-dense lhs, strided rhs view wrapped as an array)
    // take the checked per-element route; value semantics must agree.
    let base = Array::from_shape_vec([2, 6], (0..12).map(f64::from).collect::<Vec<_>>()).unwrap();
    let strided = base
        .view()
        .slice_with::<2>(&[SliceArg::All, SliceArg::range(Some(0), Some(6), 2)])
        .unwrap();
    let mut target =
        Array::from_shape_vec([2, 3], vec![100.0f64, 100.0, 100.0, 100.0, 100.0, 100.0]).unwrap();
    let strided_rhs = strided.as_array();
    target.scaled_add(2.0, &strided_rhs);
    // strided elements: rows [0,2,4] and [6,8,10] → target += 2*those.
    assert_eq!(
        target.storage().as_slice(),
        &[100.0, 104.0, 108.0, 112.0, 116.0, 120.0]
    );
}