numrs2 0.3.0

A Rust implementation inspired by NumPy for numerical computing (NumRS2)
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
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350

//! Tests for GPU batching operations

use numrs2::array::Array;
use numrs2::gpu::batching::{BatchConfig, BatchQueue, OperationType};
use numrs2::gpu::{new_context, GpuArray};

#[test]
#[cfg(feature = "gpu")]
fn test_batch_queue_creation() -> numrs2::error::Result<()> {
    let context = new_context()?;
    let config = BatchConfig::default();
    let queue: BatchQueue<f32> = BatchQueue::new(context, config);

    assert!(queue.is_empty()?);
    assert_eq!(queue.queue_depth()?, 0);

    Ok(())
}

#[test]
#[cfg(feature = "gpu")]
fn test_batch_queue_add_operations() -> numrs2::error::Result<()> {
    let context = new_context()?;
    let mut config = BatchConfig::default();
    config.enable_auto_flush = false; // Disable auto-flush for testing

    let mut queue: BatchQueue<f32> = BatchQueue::new(context.clone(), config);

    // Create test arrays
    let a = Array::from_vec(vec![1.0f32, 2.0, 3.0, 4.0]).reshape(&[4]);
    let b = Array::from_vec(vec![5.0f32, 6.0, 7.0, 8.0]).reshape(&[4]);

    let a_gpu = GpuArray::from_array_with_context(&a, context.clone())?;
    let b_gpu = GpuArray::from_array_with_context(&b, context.clone())?;

    // Queue operations
    queue.queue_add(&a_gpu, &b_gpu)?;
    queue.queue_multiply(&a_gpu, &b_gpu)?;

    assert_eq!(queue.queue_depth()?, 2);
    assert!(!queue.is_empty()?);

    Ok(())
}

#[test]
#[cfg(feature = "gpu")]
fn test_batch_queue_flush() -> numrs2::error::Result<()> {
    let context = new_context()?;
    let mut config = BatchConfig::default();
    config.enable_auto_flush = false;

    let mut queue: BatchQueue<f32> = BatchQueue::new(context.clone(), config);

    // Create test arrays
    let a = Array::from_vec(vec![1.0f32, 2.0, 3.0, 4.0]).reshape(&[4]);
    let b = Array::from_vec(vec![5.0f32, 6.0, 7.0, 8.0]).reshape(&[4]);

    let a_gpu = GpuArray::from_array_with_context(&a, context.clone())?;
    let b_gpu = GpuArray::from_array_with_context(&b, context.clone())?;

    // Queue operations
    queue.queue_add(&a_gpu, &b_gpu)?;
    queue.queue_multiply(&a_gpu, &b_gpu)?;

    // Flush and get results
    let results = queue.flush()?;

    assert_eq!(results.len(), 2);
    assert!(queue.is_empty()?);
    assert_eq!(queue.queue_depth()?, 0);

    // Verify results
    let add_result = &results[0];
    let mul_result = &results[1];

    assert_eq!(add_result.op_type, OperationType::Add);
    assert_eq!(mul_result.op_type, OperationType::Multiply);

    // Convert to CPU and check values
    let add_cpu = add_result.result.to_array()?;
    let mul_cpu = mul_result.result.to_array()?;

    let expected_add = vec![6.0f32, 8.0, 10.0, 12.0];
    let expected_mul = vec![5.0f32, 12.0, 21.0, 32.0];

    assert_eq!(add_cpu.to_vec(), expected_add);
    assert_eq!(mul_cpu.to_vec(), expected_mul);

    Ok(())
}

#[test]
#[cfg(feature = "gpu")]
fn test_batch_queue_statistics() -> numrs2::error::Result<()> {
    let context = new_context()?;
    let mut config = BatchConfig::default();
    config.enable_auto_flush = false;

    let mut queue: BatchQueue<f32> = BatchQueue::new(context.clone(), config);

    // Create test arrays
    let a = Array::from_vec(vec![1.0f32, 2.0, 3.0, 4.0]).reshape(&[4]);
    let b = Array::from_vec(vec![5.0f32, 6.0, 7.0, 8.0]).reshape(&[4]);

    let a_gpu = GpuArray::from_array_with_context(&a, context.clone())?;
    let b_gpu = GpuArray::from_array_with_context(&b, context.clone())?;

    // Get initial statistics
    let stats_before = queue.statistics()?;
    assert_eq!(stats_before.total_operations, 0);
    assert_eq!(stats_before.total_flushes, 0);

    // Queue and flush operations
    queue.queue_add(&a_gpu, &b_gpu)?;
    queue.queue_multiply(&a_gpu, &b_gpu)?;
    queue.queue_subtract(&a_gpu, &b_gpu)?;

    let stats_after_queue = queue.statistics()?;
    assert_eq!(stats_after_queue.total_operations, 3);
    assert_eq!(stats_after_queue.current_queue_depth, 3);

    queue.flush()?;

    let stats_after_flush = queue.statistics()?;
    assert_eq!(stats_after_flush.total_flushes, 1);
    assert_eq!(stats_after_flush.total_executed, 3);
    assert_eq!(stats_after_flush.current_queue_depth, 0);
    assert!(stats_after_flush.avg_batch_size > 0.0);

    Ok(())
}

#[test]
#[cfg(feature = "gpu")]
fn test_batch_queue_auto_flush() -> numrs2::error::Result<()> {
    let context = new_context()?;
    let mut config = BatchConfig::default();
    config.enable_auto_flush = true;
    config.max_batch_size = 2; // Small batch size for testing

    let mut queue: BatchQueue<f32> = BatchQueue::new(context.clone(), config);

    // Create test arrays
    let a = Array::from_vec(vec![1.0f32, 2.0, 3.0, 4.0]).reshape(&[4]);
    let b = Array::from_vec(vec![5.0f32, 6.0, 7.0, 8.0]).reshape(&[4]);

    let a_gpu = GpuArray::from_array_with_context(&a, context.clone())?;
    let b_gpu = GpuArray::from_array_with_context(&b, context.clone())?;

    // Queue operations - should auto-flush after 2 operations
    queue.queue_add(&a_gpu, &b_gpu)?;
    queue.queue_multiply(&a_gpu, &b_gpu)?;

    // Check that auto-flush occurred
    let stats = queue.statistics()?;
    assert!(stats.total_flushes > 0);

    Ok(())
}

#[test]
#[cfg(feature = "gpu")]
fn test_batch_queue_matmul() -> numrs2::error::Result<()> {
    let context = new_context()?;
    let mut config = BatchConfig::default();
    config.enable_auto_flush = false;

    let mut queue: BatchQueue<f32> = BatchQueue::new(context.clone(), config);

    // Create test matrices
    let a = Array::from_vec(vec![
        1.0f32, 2.0,
        3.0, 4.0,
    ]).reshape(&[2, 2]);

    let b = Array::from_vec(vec![
        5.0f32, 6.0,
        7.0, 8.0,
    ]).reshape(&[2, 2]);

    let a_gpu = GpuArray::from_array_with_context(&a, context.clone())?;
    let b_gpu = GpuArray::from_array_with_context(&b, context.clone())?;

    // Queue matmul operation
    queue.queue_matmul(&a_gpu, &b_gpu)?;

    // Flush and get result
    let results = queue.flush()?;

    assert_eq!(results.len(), 1);
    assert_eq!(results[0].op_type, OperationType::MatMul);

    // Convert to CPU and verify
    let result_cpu = results[0].result.to_array()?;
    let result_vec = result_cpu.to_vec();

    // Expected: [[1*5 + 2*7, 1*6 + 2*8], [3*5 + 4*7, 3*6 + 4*8]]
    //         = [[19, 22], [43, 50]]
    let expected = vec![19.0f32, 22.0, 43.0, 50.0];

    for (i, (&actual, &expected)) in result_vec.iter().zip(expected.iter()).enumerate() {
        assert!(
            (actual - expected).abs() < 1e-5,
            "Mismatch at index {}: {} != {}",
            i,
            actual,
            expected
        );
    }

    Ok(())
}

#[test]
#[cfg(feature = "gpu")]
fn test_batch_queue_clear() -> numrs2::error::Result<()> {
    let context = new_context()?;
    let mut config = BatchConfig::default();
    config.enable_auto_flush = false;

    let mut queue: BatchQueue<f32> = BatchQueue::new(context.clone(), config);

    // Create test arrays
    let a = Array::from_vec(vec![1.0f32, 2.0, 3.0, 4.0]).reshape(&[4]);
    let b = Array::from_vec(vec![5.0f32, 6.0, 7.0, 8.0]).reshape(&[4]);

    let a_gpu = GpuArray::from_array_with_context(&a, context.clone())?;
    let b_gpu = GpuArray::from_array_with_context(&b, context.clone())?;

    // Queue operations
    queue.queue_add(&a_gpu, &b_gpu)?;
    queue.queue_multiply(&a_gpu, &b_gpu)?;

    assert_eq!(queue.queue_depth()?, 2);

    // Clear queue
    queue.clear()?;

    assert_eq!(queue.queue_depth()?, 0);
    assert!(queue.is_empty()?);

    Ok(())
}

#[test]
#[cfg(feature = "gpu")]
fn test_batch_queue_mixed_operations() -> numrs2::error::Result<()> {
    let context = new_context()?;
    let mut config = BatchConfig::default();
    config.enable_auto_flush = false;

    let mut queue: BatchQueue<f32> = BatchQueue::new(context.clone(), config);

    // Create test arrays
    let a = Array::from_vec(vec![4.0f32, 9.0, 16.0, 25.0]).reshape(&[4]);
    let b = Array::from_vec(vec![2.0f32, 3.0, 4.0, 5.0]).reshape(&[4]);

    let a_gpu = GpuArray::from_array_with_context(&a, context.clone())?;
    let b_gpu = GpuArray::from_array_with_context(&b, context.clone())?;

    // Queue various operations
    queue.queue_add(&a_gpu, &b_gpu)?;
    queue.queue_subtract(&a_gpu, &b_gpu)?;
    queue.queue_multiply(&a_gpu, &b_gpu)?;
    queue.queue_divide(&a_gpu, &b_gpu)?;

    // Flush and verify
    let results = queue.flush()?;

    assert_eq!(results.len(), 4);

    let add_result = results[0].result.to_array()?.to_vec();
    let sub_result = results[1].result.to_array()?.to_vec();
    let mul_result = results[2].result.to_array()?.to_vec();
    let div_result = results[3].result.to_array()?.to_vec();

    // Verify each result
    assert_eq!(add_result, vec![6.0f32, 12.0, 20.0, 30.0]);
    assert_eq!(sub_result, vec![2.0f32, 6.0, 12.0, 20.0]);
    assert_eq!(mul_result, vec![8.0f32, 27.0, 64.0, 125.0]);
    assert_eq!(div_result, vec![2.0f32, 3.0, 4.0, 5.0]);

    Ok(())
}

#[test]
#[cfg(feature = "gpu")]
fn test_batch_queue_dynamic_optimization() -> numrs2::error::Result<()> {
    let context = new_context()?;
    let mut config = BatchConfig::default();
    config.enable_dynamic_optimization = true;
    config.enable_auto_flush = false;
    config.max_batch_size = 16;

    let mut queue: BatchQueue<f32> = BatchQueue::new(context.clone(), config);

    // Create test arrays
    let a = Array::from_vec(vec![1.0f32; 100]).reshape(&[100]);
    let b = Array::from_vec(vec![2.0f32; 100]).reshape(&[100]);

    let a_gpu = GpuArray::from_array_with_context(&a, context.clone())?;
    let b_gpu = GpuArray::from_array_with_context(&b, context.clone())?;

    // Queue multiple batches to allow optimization to kick in
    for _ in 0..5 {
        for _ in 0..8 {
            queue.queue_add(&a_gpu, &b_gpu)?;
        }
        queue.flush()?;
    }

    let stats = queue.statistics()?;

    // Verify that optimization ran
    assert!(stats.total_flushes >= 5);
    assert!(stats.total_executed >= 40);
    assert!(stats.estimated_gpu_occupancy >= 0.0);

    Ok(())
}

#[test]
#[cfg(feature = "gpu")]
fn test_operation_type_properties() {
    assert!(OperationType::MatMul.is_batchable());
    assert!(OperationType::Add.is_batchable());
    assert!(OperationType::Conv2D.is_batchable());

    assert!(OperationType::MatMul.cost_factor() > OperationType::Add.cost_factor());
    assert!(OperationType::Conv2D.cost_factor() > OperationType::Multiply.cost_factor());
}

#[test]
#[cfg(feature = "gpu")]
fn test_batch_config_custom() {
    let config = BatchConfig {
        max_batch_size: 64,
        batch_timeout: std::time::Duration::from_millis(20),
        min_batch_size: 8,
        enable_dynamic_optimization: false,
        enable_auto_flush: false,
        target_occupancy: 0.9,
    };

    assert_eq!(config.max_batch_size, 64);
    assert_eq!(config.min_batch_size, 8);
    assert!(!config.enable_dynamic_optimization);
    assert!(!config.enable_auto_flush);
}