rustorch 0.6.29

Production-ready PyTorch-compatible deep learning library in Rust with special mathematical functions (gamma, Bessel, error functions), statistical distributions, Fourier transforms (FFT/RFFT), matrix decomposition (SVD/QR/LU/eigenvalue), automatic differentiation, neural networks, computer vision transforms, complete GPU acceleration (CUDA/Metal/OpenCL), SIMD optimizations, parallel processing, WebAssembly browser support, comprehensive distributed learning support, and performance validation
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

//! Common tensor operation benchmarks
//! 
//! Reusable benchmark functions for standard tensor operations

use criterion::{black_box, BenchmarkId, Criterion, Throughput};
use rustorch::tensor::Tensor;
use num_traits::Float;
use crate::common::{BenchmarkConfig, create_benchmark_group, elements_count};

/// Benchmark element-wise addition operations
pub fn bench_elementwise_add<T>(
    c: &mut Criterion,
    group_name: &str,
    sizes: &[(usize, &str)],
    config: Option<BenchmarkConfig>,
) where
    T: Float + Clone + Default + std::ops::Add<Output = T> + 'static,
{
    let mut group = create_benchmark_group(c, group_name);
    
    if let Some(cfg) = config {
        group.sample_size(cfg.sample_size);
        group.measurement_time(std::time::Duration::from_secs(cfg.measurement_time_secs));
        group.warm_up_time(std::time::Duration::from_secs(cfg.warm_up_time_secs));
    }
    
    for (size, description) in sizes {
        let tensor1 = Tensor::<T>::zeros(&[*size]);
        let tensor2 = Tensor::<T>::zeros(&[*size]);
        
        group.throughput(Throughput::Elements(*size as u64));
        
        group.bench_with_input(
            BenchmarkId::new("elementwise_add", description),
            &(*size, &tensor1, &tensor2),
            |b, (_, t1, t2)| {
                b.iter(|| {
                    let result = black_box(&**t1 + &**t2);
                    black_box(result)
                });
            },
        );
    }
    
    group.finish();
}

/// Benchmark matrix multiplication operations
pub fn bench_matrix_multiply<T>(
    c: &mut Criterion,
    group_name: &str,
    sizes: &[(usize, &str)],
    config: Option<BenchmarkConfig>,
) where
    T: Float + Clone + Default + std::ops::Add<Output = T> + std::ops::Mul<Output = T> + 'static,
{
    let mut group = create_benchmark_group(c, group_name);
    
    if let Some(cfg) = config {
        group.sample_size(cfg.sample_size);
        group.measurement_time(std::time::Duration::from_secs(cfg.measurement_time_secs));
        group.warm_up_time(std::time::Duration::from_secs(cfg.warm_up_time_secs));
    }
    
    for (size, description) in sizes {
        let tensor1 = Tensor::<T>::zeros(&[*size, *size]);
        let tensor2 = Tensor::<T>::zeros(&[*size, *size]);
        
        group.throughput(Throughput::Elements(elements_count(&[*size, *size])));
        
        group.bench_with_input(
            BenchmarkId::new("matrix_multiply", description),
            &(*size, &tensor1, &tensor2),
            |b, (_, t1, t2)| {
                b.iter(|| {
                    let result = black_box(t1.matmul(t2));
                    black_box(result)
                });
            },
        );
    }
    
    group.finish();
}

/// Benchmark tensor allocation and deallocation
pub fn bench_tensor_allocation<T>(
    c: &mut Criterion,
    group_name: &str,
    sizes: &[(usize, &str)],
    config: Option<BenchmarkConfig>,
) where
    T: Float + Clone + Default + 'static,
{
    let mut group = create_benchmark_group(c, group_name);
    
    if let Some(cfg) = config {
        group.sample_size(cfg.sample_size);
        group.measurement_time(std::time::Duration::from_secs(cfg.measurement_time_secs));
        group.warm_up_time(std::time::Duration::from_secs(cfg.warm_up_time_secs));
    }
    
    for (size, description) in sizes {
        group.throughput(Throughput::Elements(*size as u64));
        
        group.bench_with_input(
            BenchmarkId::new("tensor_allocation", description),
            size,
            |b, sz| {
                b.iter(|| {
                    let tensor = black_box(Tensor::<T>::zeros(&[*sz]));
                    black_box(tensor)
                });
            },
        );
    }
    
    group.finish();
}

/// Benchmark tensor copying operations
pub fn bench_tensor_copy<T>(
    c: &mut Criterion,
    group_name: &str,
    sizes: &[(usize, &str)],
    config: Option<BenchmarkConfig>,
) where
    T: Float + Clone + Default + 'static,
{
    let mut group = create_benchmark_group(c, group_name);
    
    if let Some(cfg) = config {
        group.sample_size(cfg.sample_size);
        group.measurement_time(std::time::Duration::from_secs(cfg.measurement_time_secs));
        group.warm_up_time(std::time::Duration::from_secs(cfg.warm_up_time_secs));
    }
    
    for (size, description) in sizes {
        let tensor = Tensor::<T>::zeros(&[*size]);
        
        group.throughput(Throughput::Elements(*size as u64));
        
        group.bench_with_input(
            BenchmarkId::new("tensor_copy", description),
            &(*size, &tensor),
            |b, (_, t)| {
                b.iter(|| {
                    let copied = black_box(t.clone());
                    black_box(copied)
                });
            },
        );
    }
    
    group.finish();
}