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
//! Tests for Isomorphic Layer Offloading System (Phase 3, #12)
#[cfg(test)]
mod tests {
use crate::layers::{IsomorphicOffloader, LayerPlacement, MemoryPressure};
#[test]
fn test_memory_pressure_levels() {
// Test pressure calculation from free VRAM ratio
assert_eq!(MemoryPressure::from_ratio(0.95), MemoryPressure::Low);
assert_eq!(MemoryPressure::from_ratio(0.70), MemoryPressure::Moderate);
assert_eq!(MemoryPressure::from_ratio(0.50), MemoryPressure::High);
assert_eq!(MemoryPressure::from_ratio(0.30), MemoryPressure::Critical);
}
#[test]
fn test_memory_pressure_ordering() {
// Test that pressure levels are properly ordered
assert!(MemoryPressure::Low < MemoryPressure::Moderate);
assert!(MemoryPressure::Moderate < MemoryPressure::High);
assert!(MemoryPressure::High < MemoryPressure::Critical);
// Test that equality works
assert_eq!(MemoryPressure::Low, MemoryPressure::Low);
assert_ne!(MemoryPressure::Low, MemoryPressure::Critical);
}
#[test]
fn test_isomorphic_offloader_creation_hybrid() {
// Test creating a hybrid offloader (static n_gpu)
let result = IsomorphicOffloader::new_hybrid(4);
// Should succeed or fail gracefully based on GPU availability
match result {
Ok(offloader) => {
// Check that we can get stats
let stats = offloader.stats();
assert_eq!(stats.transfers, 0);
assert_eq!(stats.offload_triggers, 0);
}
Err(e) => {
// GPU may not be available in test environment
println!("Skipping GPU test: {}", e);
}
}
}
#[test]
fn test_isomorphic_layer_init() {
// Test layer initialization
let result = IsomorphicOffloader::new_hybrid(4);
match result {
Ok(mut offloader) => {
let n_layers = 32; // Typical LLaMA-2
offloader.init_layers(n_layers).ok();
// First 4 layers should be on GPU
for i in 0..4 {
assert_eq!(
offloader.get_placement(i),
LayerPlacement::GPU,
"Layer {} should be on GPU",
i
);
}
// Layers 4+ should be on CPU
for i in 4..std::cmp::min(8, n_layers) {
assert_eq!(
offloader.get_placement(i),
LayerPlacement::CPU,
"Layer {} should be on CPU",
i
);
}
// Check GPU layers list
let gpu_layers = offloader.gpu_layers();
assert_eq!(gpu_layers.len(), 4);
}
Err(e) => {
println!("Skipping GPU test: {}", e);
}
}
}
#[test]
fn test_isomorphic_dynamic_offloader() {
// Test dynamic offloader (memory-aware)
let result = IsomorphicOffloader::new_dynamic(
512 * 1024 * 1024, // 512 MB min free
256 * 1024 * 1024, // 256 MB staging
);
match result {
Ok(mut offloader) => {
let n_layers = 32;
offloader.init_layers(n_layers).ok();
// In dynamic mode, all layers should start on CPU to minimize initial VRAM
for i in 0..n_layers {
// With dynamic strategy, placement depends on actual VRAM
// Just check that it's valid
let placement = offloader.get_placement(i);
assert!(placement == LayerPlacement::GPU || placement == LayerPlacement::CPU);
}
}
Err(e) => {
println!("Skipping GPU test: {}", e);
}
}
}
#[test]
fn test_layer_placement_get() {
// Test getting layer placements
let result = IsomorphicOffloader::new_hybrid(8);
match result {
Ok(mut offloader) => {
offloader.init_layers(16).ok();
// Test get_placement for various layers
for i in 0..8 {
assert_eq!(offloader.get_placement(i), LayerPlacement::GPU);
}
for i in 8..16 {
assert_eq!(offloader.get_placement(i), LayerPlacement::CPU);
}
// Test non-existent layer (should default to CPU)
assert_eq!(offloader.get_placement(999), LayerPlacement::CPU);
}
Err(e) => {
println!("Skipping GPU test: {}", e);
}
}
}
#[test]
fn test_stats_accumulation() {
// Test that statistics are properly tracked
let result = IsomorphicOffloader::new_hybrid(4);
match result {
Ok(offloader) => {
let stats = offloader.stats();
// Initial stats should be zero
assert_eq!(stats.transfers, 0);
assert_eq!(stats.bytes_transferred, 0);
assert_eq!(stats.offload_triggers, 0);
assert_eq!(stats.preload_triggers, 0);
// Stats should be readable
let _ = stats.peak_vram_used;
}
Err(e) => {
println!("Skipping GPU test: {}", e);
}
}
}
#[test]
fn test_offloader_target_device() {
// Test getting the target device for layers
let result = IsomorphicOffloader::new_hybrid(4);
match result {
Ok(mut offloader) => {
offloader.init_layers(8).ok();
// Get target devices
let _device_0 = offloader.get_target_device(0);
let _device_5 = offloader.get_target_device(5);
// We can't directly compare Device objects easily, but we can check
// that the function returns valid devices
// (This test mainly checks that the function doesn't panic)
}
Err(e) => {
println!("Skipping GPU test: {}", e);
}
}
}
#[test]
fn test_memory_status() {
// Test memory status reporting
let result = IsomorphicOffloader::new_hybrid(4);
match result {
Ok(mut offloader) => {
let status_result = offloader.memory_status();
// Should either succeed or fail gracefully
match status_result {
Ok((_free, used, peak)) => {
// Basic sanity checks (free and used are usize, always >= 0)
assert!(peak >= used, "Peak memory should be >= used memory");
}
Err(e) => {
println!("Memory status query failed: {}", e);
}
}
}
Err(e) => {
println!("Skipping GPU test: {}", e);
}
}
}
#[test]
#[ignore = "Requires GPU"]
fn test_gpu_layers_list() {
// Test getting list of GPU layers
let result = IsomorphicOffloader::new_hybrid(6);
match result {
Ok(mut offloader) => {
offloader.init_layers(24).ok();
let gpu_layers = offloader.gpu_layers();
// Should have 6 GPU layers
assert_eq!(gpu_layers.len(), 6);
// Should be layers 0-5
for (i, &layer_id) in gpu_layers.iter().enumerate() {
assert_eq!(layer_id, i);
}
}
Err(e) => {
println!("Skipping GPU test: {}", e);
}
}
}
#[test]
fn test_layer_placement_enum() {
// Test LayerPlacement enum properties
assert_eq!(LayerPlacement::GPU, LayerPlacement::GPU);
assert_eq!(LayerPlacement::CPU, LayerPlacement::CPU);
assert_eq!(LayerPlacement::InTransit, LayerPlacement::InTransit);
assert_ne!(LayerPlacement::GPU, LayerPlacement::CPU);
assert_ne!(LayerPlacement::CPU, LayerPlacement::InTransit);
}
#[test]
fn test_ensure_layer_ready_basic() {
// Test basic ensure_layer_ready functionality
let result = IsomorphicOffloader::new_hybrid(4);
match result {
Ok(mut offloader) => {
offloader.init_layers(8).ok();
// Try to ensure a CPU layer is ready
let ensure_result = offloader.ensure_layer_ready(5, 50 * 1024 * 1024); // 50MB
// Should handle gracefully whether it succeeds or not
let _ = ensure_result;
}
Err(e) => {
println!("Skipping GPU test: {}", e);
}
}
}
}