// SPDX-License-Identifier: AGPL-3.0-or-later
// cosine_similarity.wgsl - Cosine Similarity
//
// Computes cosine similarity between pairs of vectors
// Similarity = (a · b) / (||a|| * ||b||)
//
// Used in recommendation systems, similarity search, clustering
struct Params {
num_vectors_a: u32,
num_vectors_b: u32,
vector_dim: u32,
_padding: u32,
}
@group(0) @binding(0) var<storage, read> vectors_a: array<f32>; // [num_vectors_a, dim]
@group(0) @binding(1) var<storage, read> vectors_b: array<f32>; // [num_vectors_b, dim]
@group(0) @binding(2) var<storage, read_write> output: array<f32>; // [num_vectors_a, num_vectors_b]
@group(0) @binding(3) var<uniform> params: Params;
@compute @workgroup_size(16, 16)
fn main(@builtin(global_invocation_id) global_id: vec3<u32>) {
let i = global_id.x;
let j = global_id.y;
if (i >= params.num_vectors_a || j >= params.num_vectors_b) {
return;
}
var dot_product: f32 = 0.0;
var norm_a: f32 = 0.0;
var norm_b: f32 = 0.0;
for (var d: u32 = 0u; d < params.vector_dim; d = d + 1u) {
let a_val = vectors_a[i * params.vector_dim + d];
let b_val = vectors_b[j * params.vector_dim + d];
dot_product = dot_product + a_val * b_val;
norm_a = norm_a + a_val * a_val;
norm_b = norm_b + b_val * b_val;
}
// Cosine similarity
let similarity = dot_product / (sqrt(norm_a) * sqrt(norm_b) + 1e-8);
output[i * params.num_vectors_b + j] = similarity;
}