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

// 3 -> 2
// storage -> workgroup
// second stage

// NUM_INVOCATIONS workgroups of 1 thread each
@compute @workgroup_size(1)
fn run_bucket_accumulation_phase(
    @builtin(global_invocation_id) global_id: vec3<u32>,
    @builtin(local_invocation_id) local_id: vec3<u32>
) {
    let gidx = global_id.x;
    let lidx = local_id.x;

    if (gidx >= msm_len.val) {
        return;
    }

    bucket_accumulation_phase(gidx);
}

@compute @workgroup_size(1)
fn run_bucket_reduction_phase(
    @builtin(global_invocation_id) global_id: vec3<u32>,
    @builtin(local_invocation_id) local_id: vec3<u32>
) {
    let gidx = global_id.x;
    let lidx = local_id.x;
    if (gidx >= msm_len.val) {
        return;
    }
    bucket_reduction_phase(gidx);
}

@compute @workgroup_size(1)
fn run_final_reduction_phase(
    @builtin(global_invocation_id) global_id: vec3<u32>,
    @builtin(local_invocation_id) local_id: vec3<u32>
) {
    let gidx = global_id.x;
    let lidx = local_id.x;
    if (gidx >= msm_len.val) {
        return;
    }
    result[gidx] = final_reduction_phase(gidx);
}


// Assuming a workgroup size of 256 (as in your original code)
@compute @workgroup_size(1)
fn aggregate(
    @builtin(global_invocation_id) global_id: vec3<u32>,
    @builtin(local_invocation_id) local_id: vec3<u32>
) {
    let lidx = local_id.x;
    var n = num_invocations.val;
   
    // if (lidx == 0) {
    //     while (n > 1) {
    //         let l = (n + 1) / 2;
    //         let pairs = n / 2;
    //         for (var i = 0u; i < pairs; i = i + 1u) {
    //             result[i] = jacobian_add(
    //                 result[2 * i],
    //                 result[2 * i + 1]
    //             );
    //         }
    //         if (n % 2 == 1) {
    //             result[pairs] = result[n - 1];
    //         }
    //         n = (n + 1) / 2;
    //     }
    // }
}