aprender-compute 0.32.0

High-performance SIMD compute library with GPU support, LLM inference engine, and GGUF model loading (was: trueno)
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

#![allow(clippy::disallowed_methods, clippy::float_cmp)]
//! GPU vector operation benchmarks (vec_add, dot product)

use criterion::{BenchmarkId, Criterion, Throughput};
use std::hint::black_box;
use trueno::backends::gpu::GpuBackend;

/// Generate test data for benchmarks
fn generate_test_data(size: usize) -> Vec<f32> {
    (0..size).map(|i| (i as f32) * 0.5).collect()
}

/// Benchmark GPU vector addition vs scalar baseline
pub fn bench_gpu_vec_add(c: &mut Criterion) {
    // Skip if GPU not available
    if !GpuBackend::is_available() {
        eprintln!("GPU not available, skipping GPU benchmarks");
        return;
    }

    let mut group = c.benchmark_group("gpu_vec_add");

    for size in [1_000, 10_000, 100_000, 1_000_000].iter() {
        group.throughput(Throughput::Elements(*size as u64));

        // GPU backend
        group.bench_with_input(BenchmarkId::new("GPU", size), size, |bencher, &size| {
            let data_a = generate_test_data(size);
            let data_b = generate_test_data(size);
            let mut gpu = GpuBackend::new();

            bencher.iter(|| {
                black_box(gpu.vec_add(&data_a, &data_b).unwrap());
            });
        });

        // Scalar baseline (for speedup comparison)
        group.bench_with_input(BenchmarkId::new("Scalar", size), size, |bencher, &size| {
            let data_a = generate_test_data(size);
            let data_b = generate_test_data(size);

            bencher.iter(|| {
                let result: Vec<f32> =
                    data_a.iter().zip(data_b.iter()).map(|(a, b)| a + b).collect();
                black_box(result);
            });
        });
    }

    group.finish();
}

/// Benchmark GPU dot product vs scalar baseline
pub fn bench_gpu_dot(c: &mut Criterion) {
    // Skip if GPU not available
    if !GpuBackend::is_available() {
        eprintln!("GPU not available, skipping GPU benchmarks");
        return;
    }

    let mut group = c.benchmark_group("gpu_dot");

    for size in [1_000, 10_000, 100_000, 1_000_000].iter() {
        group.throughput(Throughput::Elements(*size as u64));

        // GPU backend
        group.bench_with_input(BenchmarkId::new("GPU", size), size, |bencher, &size| {
            let data_a = generate_test_data(size);
            let data_b = generate_test_data(size);
            let mut gpu = GpuBackend::new();

            bencher.iter(|| {
                black_box(gpu.dot(&data_a, &data_b).unwrap());
            });
        });

        // Scalar baseline (for speedup comparison)
        group.bench_with_input(BenchmarkId::new("Scalar", size), size, |bencher, &size| {
            let data_a = generate_test_data(size);
            let data_b = generate_test_data(size);

            bencher.iter(|| {
                let result: f32 = data_a.iter().zip(data_b.iter()).map(|(a, b)| a * b).sum();
                black_box(result);
            });
        });
    }

    group.finish();
}