oxiblas 0.2.2

OxiBLAS - Pure Rust BLAS/LAPACK implementation for the scirs2 ecosystem
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

//! Complex GEMM (3M method) benchmark
//!
//! Measures performance of complex matrix multiplication using the 3M method
//! that leverages optimized real GEMM kernels.

use num_complex::{Complex32, Complex64};
use oxiblas::prelude::*;
use oxiblas_blas::level3::{gemm3m_c32, gemm3m_c64};
use std::time::Instant;

/// Naive complex GEMM for comparison (4 real multiplications per complex multiplication)
fn gemm_naive_c64(a: &Mat<Complex64>, b: &Mat<Complex64>, c: &mut Mat<Complex64>) {
    let m = a.nrows();
    let k = a.ncols();
    let n = b.ncols();

    for i in 0..m {
        for j in 0..n {
            let mut sum = Complex64::new(0.0, 0.0);
            for p in 0..k {
                sum += a[(i, p)] * b[(p, j)];
            }
            c[(i, j)] = sum;
        }
    }
}

fn gemm_naive_c32(a: &Mat<Complex32>, b: &Mat<Complex32>, c: &mut Mat<Complex32>) {
    let m = a.nrows();
    let k = a.ncols();
    let n = b.ncols();

    for i in 0..m {
        for j in 0..n {
            let mut sum = Complex32::new(0.0, 0.0);
            for p in 0..k {
                sum += a[(i, p)] * b[(p, j)];
            }
            c[(i, j)] = sum;
        }
    }
}

fn bench_complex_gemm_c64(sizes: &[(usize, &str)], n_warmup: usize, n_samples: usize) {
    println!("\n=== Complex64 ZGEMM (3M Method) ===");
    println!(
        "{:>8} {:>12} {:>12} {:>12}",
        "Size", "Naive", "3M Method", "Speedup"
    );
    println!("{}", "-".repeat(50));

    for &(n, label) in sizes {
        // Create complex matrices with values
        let mut a: Mat<Complex64> = Mat::zeros(n, n);
        let mut b: Mat<Complex64> = Mat::zeros(n, n);

        for i in 0..n {
            for j in 0..n {
                a[(i, j)] = Complex64::new((i + j) as f64 * 0.001, (i * j) as f64 * 0.0001);
                b[(i, j)] = Complex64::new((i + 1) as f64 * 0.002, (j + 1) as f64 * 0.0002);
            }
        }

        // Warmup naive
        for _ in 0..n_warmup {
            let mut c: Mat<Complex64> = Mat::zeros(n, n);
            gemm_naive_c64(&a, &b, &mut c);
            std::hint::black_box(&c);
        }

        // Measure naive
        let mut naive_times = Vec::with_capacity(n_samples);
        for _ in 0..n_samples {
            let mut c: Mat<Complex64> = Mat::zeros(n, n);
            let start = Instant::now();
            gemm_naive_c64(&a, &b, &mut c);
            naive_times.push(start.elapsed().as_secs_f64());
            std::hint::black_box(&c);
        }
        naive_times.sort_by(|x, y| x.partial_cmp(y).unwrap());
        let naive_median = naive_times[n_samples / 2];

        // Warmup 3M
        for _ in 0..n_warmup {
            let mut c: Mat<Complex64> = Mat::zeros(n, n);
            gemm3m_c64(
                Complex64::new(1.0, 0.0),
                a.as_ref(),
                b.as_ref(),
                Complex64::new(0.0, 0.0),
                c.as_mut(),
            );
            std::hint::black_box(&c);
        }

        // Measure 3M
        let mut opt_times = Vec::with_capacity(n_samples);
        for _ in 0..n_samples {
            let mut c: Mat<Complex64> = Mat::zeros(n, n);
            let start = Instant::now();
            gemm3m_c64(
                Complex64::new(1.0, 0.0),
                a.as_ref(),
                b.as_ref(),
                Complex64::new(0.0, 0.0),
                c.as_mut(),
            );
            opt_times.push(start.elapsed().as_secs_f64());
            std::hint::black_box(&c);
        }
        opt_times.sort_by(|x, y| x.partial_cmp(y).unwrap());
        let opt_median = opt_times[n_samples / 2];

        // Complex GEMM has 8 real flops per complex multiply-add
        // For C = A * B where all are n×n: 8 * n³ flops
        let flops = 8.0 * (n as f64).powi(3);
        let naive_gflops = flops / naive_median / 1e9;
        let opt_gflops = flops / opt_median / 1e9;
        let speedup = naive_median / opt_median;

        println!(
            "{:>8} {:>9.2} G/s {:>9.2} G/s {:>10.2}x",
            label, naive_gflops, opt_gflops, speedup
        );
    }
}

fn bench_complex_gemm_c32(sizes: &[(usize, &str)], n_warmup: usize, n_samples: usize) {
    println!("\n=== Complex32 CGEMM (3M Method) ===");
    println!(
        "{:>8} {:>12} {:>12} {:>12}",
        "Size", "Naive", "3M Method", "Speedup"
    );
    println!("{}", "-".repeat(50));

    for &(n, label) in sizes {
        // Create complex matrices with values
        let mut a: Mat<Complex32> = Mat::zeros(n, n);
        let mut b: Mat<Complex32> = Mat::zeros(n, n);

        for i in 0..n {
            for j in 0..n {
                a[(i, j)] = Complex32::new((i + j) as f32 * 0.001, (i * j) as f32 * 0.0001);
                b[(i, j)] = Complex32::new((i + 1) as f32 * 0.002, (j + 1) as f32 * 0.0002);
            }
        }

        // Warmup naive
        for _ in 0..n_warmup {
            let mut c: Mat<Complex32> = Mat::zeros(n, n);
            gemm_naive_c32(&a, &b, &mut c);
            std::hint::black_box(&c);
        }

        // Measure naive
        let mut naive_times = Vec::with_capacity(n_samples);
        for _ in 0..n_samples {
            let mut c: Mat<Complex32> = Mat::zeros(n, n);
            let start = Instant::now();
            gemm_naive_c32(&a, &b, &mut c);
            naive_times.push(start.elapsed().as_secs_f64());
            std::hint::black_box(&c);
        }
        naive_times.sort_by(|x, y| x.partial_cmp(y).unwrap());
        let naive_median = naive_times[n_samples / 2];

        // Warmup 3M
        for _ in 0..n_warmup {
            let mut c: Mat<Complex32> = Mat::zeros(n, n);
            gemm3m_c32(
                Complex32::new(1.0, 0.0),
                a.as_ref(),
                b.as_ref(),
                Complex32::new(0.0, 0.0),
                c.as_mut(),
            );
            std::hint::black_box(&c);
        }

        // Measure 3M
        let mut opt_times = Vec::with_capacity(n_samples);
        for _ in 0..n_samples {
            let mut c: Mat<Complex32> = Mat::zeros(n, n);
            let start = Instant::now();
            gemm3m_c32(
                Complex32::new(1.0, 0.0),
                a.as_ref(),
                b.as_ref(),
                Complex32::new(0.0, 0.0),
                c.as_mut(),
            );
            opt_times.push(start.elapsed().as_secs_f64());
            std::hint::black_box(&c);
        }
        opt_times.sort_by(|x, y| x.partial_cmp(y).unwrap());
        let opt_median = opt_times[n_samples / 2];

        // Complex GEMM has 8 real flops per complex multiply-add
        let flops = 8.0 * (n as f64).powi(3);
        let naive_gflops = flops / naive_median / 1e9;
        let opt_gflops = flops / opt_median / 1e9;
        let speedup = naive_median / opt_median;

        println!(
            "{:>8} {:>9.2} G/s {:>9.2} G/s {:>10.2}x",
            label, naive_gflops, opt_gflops, speedup
        );
    }
}

fn main() {
    println!("==============================================");
    println!("   OxiBLAS Complex GEMM (3M) Benchmark");
    println!("==============================================");
    println!();
    println!("The 3M method uses 3 optimized real GEMM operations");
    println!("with NEON SIMD, cache blocking, and prefetching.");

    let sizes = [(256, "256"), (512, "512"), (1024, "1024")];

    bench_complex_gemm_c64(&sizes, 3, 5);
    bench_complex_gemm_c32(&sizes, 3, 5);

    println!("\nNote: Complex GEMM counts 8 real flops per complex multiply-add.");
}