msm-webgpu 0.1.0

Multi-Scalar Multiplication (MSM) implementation for WebGPU.
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

use std::time::Instant;

use halo2curves::CurveAffine;
use wgpu::CommandEncoderDescriptor;

use crate::cuzk::{
    gpu::{get_adapter, get_device, read_from_gpu_test},
    msm::{PARAMS, WORD_SIZE, convert_point_coords_and_decompose_shaders, transpose_gpu},
    shader_manager::ShaderManager,
    utils::debug,
};
use crate::{points_to_bytes, scalars_to_bytes};

async fn transpose_shader<C: CurveAffine>(
    points: &[C],
    scalars: &[C::Scalar],
) -> (Vec<i32>, Vec<i32>) {
    let input_size = scalars.len();
    let chunk_size = if input_size >= 65536 { 16 } else { 4 };
    let num_columns = 1 << chunk_size;
    let num_rows = input_size.div_ceil(num_columns);
    let num_subtasks = 256_usize.div_ceil(chunk_size);
    let num_words = PARAMS.num_words;
    debug(&format!("Input size: {input_size}"));
    debug(&format!("Chunk size: {chunk_size}"));
    debug(&format!("Num columns: {num_columns}"));
    debug(&format!("Num rows: {num_rows}"));
    debug(&format!("Num subtasks: {num_subtasks}"));
    debug(&format!("Num words: {num_words}"));
    debug(&format!("Word size: {WORD_SIZE}"));
    println!("Params: {PARAMS:?}");

    let point_bytes = points_to_bytes(points);
    let scalar_bytes = scalars_to_bytes(scalars);

    let shader_manager = ShaderManager::new(WORD_SIZE, chunk_size, input_size);

    let adapter = get_adapter().await;
    let (device, queue) = get_device(&adapter).await;
    let mut encoder = device.create_command_encoder(&CommandEncoderDescriptor {
        label: Some("Decompose Encoder"),
    });

    // Total thread count = workgroup_size * #x workgroups * #y workgroups * #z workgroups.
    let mut c_workgroup_size = 64;
    let mut c_num_x_workgroups = 128;
    let mut c_num_y_workgroups = input_size / c_workgroup_size / c_num_x_workgroups;
    let c_num_z_workgroups = 1;

    if input_size <= 256 {
        c_workgroup_size = input_size;
        c_num_x_workgroups = 1;
        c_num_y_workgroups = 1;
    } else if input_size > 256 && input_size <= 32768 {
        c_workgroup_size = 64;
        c_num_x_workgroups = 4;
        c_num_y_workgroups = input_size / c_workgroup_size / c_num_x_workgroups;
    } else if input_size > 32768 && input_size <= 131072 {
        c_workgroup_size = 256;
        c_num_x_workgroups = 8;
        c_num_y_workgroups = input_size / c_workgroup_size / c_num_x_workgroups;
    } else if input_size > 131072 && input_size <= 1048576 {
        c_workgroup_size = 256;
        c_num_x_workgroups = 32;
        c_num_y_workgroups = input_size / c_workgroup_size / c_num_x_workgroups;
    } 

    let c_shader = shader_manager.gen_decomp_scalars_shader(
        c_workgroup_size,
        c_num_y_workgroups,
        num_subtasks,
        num_columns,
    );

    let (_point_x_sb, _point_y_sb, scalar_chunks_sb) = convert_point_coords_and_decompose_shaders(
        &c_shader,
        c_num_x_workgroups,
        c_num_y_workgroups,
        c_num_z_workgroups,
        &device,
        &queue,
        &mut encoder,
        &point_bytes,
        &scalar_bytes,
        num_subtasks,
        chunk_size,
        num_words,
    )
    .await;

    ////////////////////////////////////////////////////////////////////////////////////////////
    // 2. Sparse Matrix Transposition                                                         /
    //                                                                                        /
    // Compute the indices of the points which share the same                                 /
    // scalar chunks, enabling the parallel accumulation of points                            /
    // into buckets. Transposing each subtask (CSR sparse matrix)                             /
    // is a serial computation.                                                               /
    //                                                                                        /
    // The transpose step generates the CSR sparse matrix and                                 /
    // transpoes the matrix simultaneously, resulting in a                                    /
    // wide and flat matrix where width of the matrix (n) = 2 ^ chunk_size                    /
    // and height of the matrix (m) = 1.                                                      /
    ////////////////////////////////////////////////////////////////////////////////////////////

    let t_num_x_workgroups = 1;
    let t_num_y_workgroups = 1;
    let t_num_z_workgroups = 1;

    let t_shader = shader_manager.gen_transpose_shader(num_subtasks);

    let (all_csc_col_ptr_sb, all_csc_val_idxs_sb) = transpose_gpu(
        &t_shader,
        &device,
        &queue,
        &mut encoder,
        t_num_x_workgroups,
        t_num_y_workgroups,
        t_num_z_workgroups,
        input_size,
        num_columns,
        num_rows,
        num_subtasks,
        scalar_chunks_sb,
    )
    .await;

    // Map results back from GPU to CPU.
    let data = read_from_gpu_test(
        &device,
        &queue,
        encoder,
        vec![all_csc_col_ptr_sb, all_csc_val_idxs_sb],
    )
    .await;

    // Destroy the GPU device object.
    device.destroy();

    let all_csc_col_ptr = bytemuck::cast_slice::<u8, i32>(&data[0]);
    let all_csc_val_idxs = bytemuck::cast_slice::<u8, i32>(&data[1]);

    (all_csc_col_ptr.to_vec(), all_csc_val_idxs.to_vec())
}

/// Run WebGPU transpose shader sync
pub fn run_webgpu_transpose_shader<C: CurveAffine>(
    points: &[C],
    scalars: &[C::Scalar],
) -> (Vec<i32>, Vec<i32>) {
    pollster::block_on(run_webgpu_transpose_shader_async(points, scalars))
}

/// Run WebGPU transpose shader async
pub async fn run_webgpu_transpose_shader_async<C: CurveAffine>(
    points: &[C],
    scalars: &[C::Scalar],
) -> (Vec<i32>, Vec<i32>) {
    let now = Instant::now();
    let result = transpose_shader::<C>(points, scalars).await;
    println!("Transpose time: {:?}", now.elapsed());
    result
}

#[cfg(test)]
mod tests {
    use crate::cuzk::test::cuzk::{cpu_transpose, decompose_scalars_signed};
    use crate::{sample_points, sample_scalars};

    use super::*;
    use halo2curves::bn256::{Fr, G1Affine};

    #[test]
    fn test_webgpu_transpose_shader() {
        let input_size = 1 << 16;
        let scalars = sample_scalars::<Fr>(input_size);
        let points = sample_points::<G1Affine>(input_size);

        let chunk_size = if input_size >= 65536 { 16 } else { 4 };
        let num_columns = 1 << chunk_size;
        let num_rows = (input_size + num_columns - 1) / num_columns;
        let num_chunks_per_scalar = (256 + chunk_size - 1) / chunk_size;
        let num_subtasks = num_chunks_per_scalar;

        let (all_csc_col_ptr, all_csc_val_idxs) =
            run_webgpu_transpose_shader::<G1Affine>(&points, &scalars);

        let decomposed_scalars = decompose_scalars_signed(&scalars, num_subtasks, chunk_size);

        // Perform multiple transpositions "in parallel"}
        let (all_csc_col_ptr_cpu, _, all_csc_val_idxs_cpu) = cpu_transpose(
            decomposed_scalars.concat(),
            num_columns,
            num_rows,
            num_subtasks,
            input_size,
        );

        assert_eq!(all_csc_col_ptr, all_csc_col_ptr_cpu);
        assert_eq!(all_csc_val_idxs, all_csc_val_idxs_cpu);
    }
}