numr 0.5.2

High-performance numerical computing with multi-backend GPU acceleration (CPU/CUDA/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

// Auto-generated gather_nonzero operations for f32

const WORKGROUP_SIZE: u32 = 256u;

struct CountParams {
    numel: u32,
}

@group(0) @binding(0) var<storage, read_write> input: array<f32>;
@group(0) @binding(1) var<storage, read_write> indices_output: array<i32>;
@group(0) @binding(2) var<storage, read_write> counter: array<atomic<u32>>;
@group(0) @binding(3) var<uniform> count_params: CountParams;

@compute @workgroup_size(256)
fn gather_nonzero_f32(@builtin(global_invocation_id) global_id: vec3<u32>) {
    let numel = count_params.numel;
    var idx = global_id.x;

    while (idx < numel) {
        if (input[idx] != 0.0) {
            let out_idx = atomicAdd(&counter[0], 1u);
            indices_output[out_idx] = i32(idx);
        }
        idx = idx + WORKGROUP_SIZE * 256u;
    }
}