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
//! Configuration types for HNSW index
use crate::similarity::SimilarityMetric;
use serde::{Deserialize, Serialize};
#[cfg(feature = "gpu")]
use crate::gpu::GpuConfig;
/// Configuration for HNSW index
#[derive(Debug, Clone, Serialize, Deserialize, oxicode::Encode, oxicode::Decode)]
pub struct HnswConfig {
/// Maximum number of bi-directional links created for each node during construction (except layer 0)
pub m: usize,
/// Maximum number of bi-directional links created for each node during construction for layer 0
pub m_l0: usize,
/// Level generation factor
pub ml: f64,
/// Size of the dynamic candidate list
pub ef: usize,
/// Size of the dynamic candidate list during construction
pub ef_construction: usize,
/// Similarity metric to use
pub metric: SimilarityMetric,
/// Enable SIMD optimizations for distance calculations
pub enable_simd: bool,
/// Enable memory prefetching for improved cache performance
pub enable_prefetch: bool,
/// Enable parallel search across multiple threads
pub enable_parallel: bool,
/// Prefetch distance (number of nodes to prefetch ahead)
pub prefetch_distance: usize,
/// Enable cache-friendly data layout
pub cache_friendly_layout: bool,
/// Enable GPU acceleration for distance calculations and search
pub enable_gpu: bool,
/// GPU configuration (only used if enable_gpu is true)
#[cfg(feature = "gpu")]
pub gpu_config: Option<GpuConfig>,
/// Minimum batch size for GPU operations (smaller batches use CPU)
pub gpu_batch_threshold: usize,
/// Enable multi-GPU support for massive datasets
pub enable_multi_gpu: bool,
}
impl Default for HnswConfig {
fn default() -> Self {
Self {
m: 16,
m_l0: 32,
ml: 1.0 / (2.0_f64).ln(),
ef: 50,
ef_construction: 200,
metric: SimilarityMetric::Cosine,
enable_simd: true,
enable_prefetch: true,
enable_parallel: true,
prefetch_distance: 8,
cache_friendly_layout: true,
enable_gpu: false,
#[cfg(feature = "gpu")]
gpu_config: None,
gpu_batch_threshold: 1000,
enable_multi_gpu: false,
}
}
}
impl HnswConfig {
/// Create a performance-optimized configuration
pub fn optimized() -> Self {
Self {
m: 32,
m_l0: 64,
ef: 100,
ef_construction: 400,
enable_simd: true,
enable_prefetch: true,
enable_parallel: true,
prefetch_distance: 16,
cache_friendly_layout: true,
enable_gpu: false,
#[cfg(feature = "gpu")]
gpu_config: None,
gpu_batch_threshold: 500,
enable_multi_gpu: false,
..Default::default()
}
}
/// Create a memory-optimized configuration
pub fn memory_optimized() -> Self {
Self {
m: 8,
m_l0: 16,
ef: 32,
ef_construction: 100,
enable_simd: true,
enable_prefetch: false,
enable_parallel: false,
prefetch_distance: 4,
cache_friendly_layout: true,
enable_gpu: false,
#[cfg(feature = "gpu")]
gpu_config: None,
gpu_batch_threshold: 2000,
enable_multi_gpu: false,
..Default::default()
}
}
/// Create a GPU-accelerated configuration for maximum performance
#[cfg(feature = "gpu")]
pub fn gpu_optimized() -> Self {
Self {
m: 48,
m_l0: 96,
ef: 200,
ef_construction: 800,
enable_simd: true,
enable_prefetch: true,
enable_parallel: true,
prefetch_distance: 32,
cache_friendly_layout: true,
enable_gpu: true,
gpu_config: Some(GpuConfig {
enable_mixed_precision: true,
enable_tensor_cores: true,
batch_size: 4096,
stream_count: 8,
enable_async_execution: true,
optimization_level: crate::gpu::OptimizationLevel::Performance,
..Default::default()
}),
gpu_batch_threshold: 100,
enable_multi_gpu: false,
..Default::default()
}
}
/// Create a multi-GPU configuration for massive datasets
#[cfg(feature = "gpu")]
pub fn multi_gpu_optimized(gpu_ids: Vec<i32>) -> Self {
Self {
m: 64,
m_l0: 128,
ef: 300,
ef_construction: 1200,
enable_simd: true,
enable_prefetch: true,
enable_parallel: true,
prefetch_distance: 64,
cache_friendly_layout: true,
enable_gpu: true,
gpu_config: Some(GpuConfig {
enable_mixed_precision: true,
enable_tensor_cores: true,
batch_size: 8192,
stream_count: 16,
enable_async_execution: true,
enable_multi_gpu: true,
preferred_gpu_ids: gpu_ids,
optimization_level: crate::gpu::OptimizationLevel::Performance,
..Default::default()
}),
gpu_batch_threshold: 50,
enable_multi_gpu: true,
..Default::default()
}
}
}