cubek-convolution 0.3.0

CubeK: Convolution Kernels
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

use std::marker::PhantomData;

use cubecl::{
    benchmark::{Benchmark, ProfileDuration, TimingMethod},
    client::Client,
    future,
    prelude::*,
    std::tensor::TensorHandle,
    zspace::Shape,
};
use cubek_matmul::{
    definition::MatmulElems,
    definition::{MatmulPrecision, MatrixPrecision},
};
use cubek_std::InputBinding;
use cubek_test_utils::{RunSamples, TestInput};

use crate::{ConvolutionInputs, Strategy, eval::benchmarks::problem::Conv2dProblem, launch_ref};

type LhsG<MP> = <<MP as MatmulPrecision>::Lhs as MatrixPrecision>::Global;
type LhsS<MP> = <<MP as MatmulPrecision>::Lhs as MatrixPrecision>::Stage;
type RhsG<MP> = <<MP as MatmulPrecision>::Rhs as MatrixPrecision>::Global;
type AccG<MP> = <<MP as MatmulPrecision>::Acc as MatrixPrecision>::Global;
type AccR<MP> = <<MP as MatmulPrecision>::Acc as MatrixPrecision>::Register;

pub fn bench(
    strategy: &Strategy,
    problem: &Conv2dProblem,
    num_samples: usize,
) -> Result<RunSamples, String> {
    let device = cubecl::test_device();
    let client = device.client();

    let bench = Conv2dBench::<half::f16> {
        problem: problem.clone(),
        strategy: strategy.clone(),
        device,
        client,
        samples: num_samples,
        _phantom: PhantomData,
    };

    let durations = bench
        .run(cubek_test_utils::timing_method(TimingMethod::System))
        .map_err(|e| format!("benchmark failed: {e}"))?
        .durations;

    Ok(RunSamples::new(durations))
}

struct Conv2dBench<MP> {
    problem: Conv2dProblem,
    strategy: Strategy,
    device: cubecl::Device,
    client: Client,
    samples: usize,
    _phantom: PhantomData<MP>,
}

fn make_uniform_4d(client: &Client, shape: [usize; 4], dtype: ElemType, seed: u64) -> TensorHandle {
    TestInput::builder(client.clone(), Shape::new(shape))
        .dtype(dtype)
        .uniform(seed, 0.0, 1.0)
        .generate_without_host_data()
}

impl<MP: MatmulPrecision> Benchmark for Conv2dBench<MP> {
    type Input = (TensorHandle, TensorHandle, TensorHandle);
    type Output = ();

    fn prepare(&self) -> Self::Input {
        let client = self.device.client();

        let input = make_uniform_4d(
            &client,
            self.problem.input_shape,
            LhsG::<MP>::elem_type_native(),
            0,
        );
        let weight = make_uniform_4d(
            &client,
            self.problem.weight_shape,
            RhsG::<MP>::elem_type_native(),
            1,
        );
        let bias = TestInput::builder(client.clone(), Shape::from(vec![self.problem.bias_shape]))
            .dtype(AccG::<MP>::elem_type_native())
            .layout(cubek_test_utils::StridedLayout::Explicit(vec![1]))
            .uniform(2, 0.0, 1.0)
            .generate_without_host_data();

        (input, weight, bias)
    }

    fn execute(&self, (input, weight, bias): Self::Input) -> Result<(), String> {
        let client = self.device.client();
        let [n, _, h_in, w_in] = self.problem.input_shape;
        let [c_out, _, k_h, k_w] = self.problem.weight_shape;
        let [s_h, s_w] = self.problem.args.stride;
        let [p_h, p_w] = self.problem.args.padding;
        let [d_h, d_w] = self.problem.args.dilation;

        let h_out = (h_in + 2 * p_h - d_h * (k_h - 1) - 1) / s_h + 1;
        let w_out = (w_in + 2 * p_w - d_w * (k_w - 1) - 1) / s_w + 1;

        let elems = MatmulElems::new_deprecated::<MP>();

        let out: TensorHandle =
            TensorHandle::empty(&client, vec![n, c_out, h_out, w_out], elems.acc_global);

        launch_ref::<2>(
            &self.strategy,
            &self.client,
            ConvolutionInputs::Forward {
                input: InputBinding::Normal(input.binding(), elems.lhs_global),
                weight: InputBinding::Normal(weight.binding(), elems.rhs_global),
                bias: Some(InputBinding::Normal(bias.binding(), elems.acc_global)),
                out: out.binding(),
            },
            self.problem.args.clone(),
            elems,
        )
        .map_err(|it| format!("{it:?}"))?;
        Ok(())
    }

    fn num_samples(&self) -> usize {
        self.samples
    }

    fn name(&self) -> String {
        let client = self.device.client();
        format!(
            "{}-conv2d-{}-{}-{}-{}",
            client.name(),
            LhsG::<MP>::elem_type_native(),
            LhsS::<MP>::elem_type_native(),
            AccR::<MP>::elem_type_native(),
            AccG::<MP>::elem_type_native(),
        )
        .to_lowercase()
    }

    fn sync(&self) {
        future::block_on(self.client.sync()).unwrap()
    }

    fn profile(&self, args: Self::Input) -> Result<ProfileDuration, String> {
        cubek_test_utils::profile_launch(&self.client, "conv-bench", || self.execute(args))
    }
}