// SPDX-License-Identifier: AGPL-3.0-or-later
//
// KL divergence: sum(p * ln(p / q)) (f64).
// Single-pass fused — replaces CPU log-ratio + GPU sum pipeline.
// Input distributions are assumed pre-normalized.
//
// Binding layout:
// 0: uniform { n: u32, pad: u32 }
// 1: storage, read — p: array<f64>
// 2: storage, read — q: array<f64>
// 3: storage, read_write — partials: array<f64> (one per workgroup)
struct Params {
n: u32,
pad0: u32,
}
@group(0) @binding(0) var<uniform> params: Params;
@group(0) @binding(1) var<storage, read> p: array<f64>;
@group(0) @binding(2) var<storage, read> q: array<f64>;
@group(0) @binding(3) var<storage, read_write> partials: array<f64>;
var<workgroup> wg_data: array<f64, 256>;
@compute @workgroup_size(256)
fn main(
@builtin(global_invocation_id) gid: vec3<u32>,
@builtin(local_invocation_id) lid: vec3<u32>,
@builtin(workgroup_id) wgid: vec3<u32>,
) {
let idx = gid.x;
var val = f64(0.0);
if (idx < params.n) {
let guard = f64(1e-30);
let pi = max(p[idx], guard);
let qi = max(q[idx], guard);
val = pi * log(pi / qi);
}
wg_data[lid.x] = val;
workgroupBarrier();
for (var stride = 128u; stride > 0u; stride >>= 1u) {
if (lid.x < stride) {
wg_data[lid.x] += wg_data[lid.x + stride];
}
workgroupBarrier();
}
if (lid.x == 0u) {
partials[wgid.x] = wg_data[0];
}
}