// SPDX-License-Identifier: AGPL-3.0-or-later
//
// Multi-shift CG: ζ recurrence + shifted scalar computation.
// Jegerlehner (hep-lat/9612014) Algorithm 1.
//
// Per iteration j, for each shift s:
// ζ_s^{j+1} = ζ_s^j * ζ_s^{j-1} * α_{j-1} /
// (β_{j-1} * α_{j-1} * (ζ_s^{j-1} - ζ_s^j) +
// ζ_s^{j-1} * α_{j-1} * (1 + σ_s * α_j))
// α_s^j = α_j * ζ_s^{j+1} / ζ_s^j
// ratio_s = ζ_s^{j+1} / ζ_s^j (β_s = ratio² * β_base, computed in p-update)
//
// Then rotates: ζ_prev ← ζ_curr, ζ_curr ← ζ_new.
//
// All dynamic scalars (α_j, β_{j-1}, α_{j-1}) read from GPU buffers.
//
// Absorbed from hotSpring lattice QCD (Mar 2026).
struct Params {
n_shifts: u32,
_pad0: u32,
_pad1: u32,
_pad2: u32,
}
@group(0) @binding(0) var<uniform> params: Params;
@group(0) @binding(1) var<storage, read> sigma: array<f64>;
@group(0) @binding(2) var<storage, read_write> zeta_curr: array<f64>;
@group(0) @binding(3) var<storage, read_write> zeta_prev: array<f64>;
@group(0) @binding(4) var<storage, read_write> alpha_s: array<f64>;
@group(0) @binding(5) var<storage, read_write> beta_ratio: array<f64>;
@group(0) @binding(6) var<storage, read> alpha_j: array<f64>;
@group(0) @binding(7) var<storage, read> beta_prev: array<f64>;
@group(0) @binding(8) var<storage, read> alpha_prev: array<f64>;
@compute @workgroup_size(64)
fn main(@builtin(global_invocation_id) gid: vec3<u32>) {
let s = gid.x;
if (s >= params.n_shifts) { return; }
let a_j = alpha_j[0];
let b_prev = beta_prev[0];
let a_prev = alpha_prev[0];
let z = zeta_curr[s];
let z_prev = zeta_prev[s];
let numer = z * z_prev * a_prev;
let denom = b_prev * a_prev * (z_prev - z)
+ z_prev * a_prev * (f64(1.0) + sigma[s] * a_j);
var z_new = f64(0.0);
if (abs(denom) > f64(1e-30)) {
z_new = numer / denom;
}
var a_s = f64(0.0);
if (abs(z) > f64(1e-30)) {
a_s = a_j * z_new / z;
}
alpha_s[s] = a_s;
var ratio = f64(0.0);
if (abs(z) > f64(1e-30)) {
ratio = z_new / z;
}
beta_ratio[s] = ratio;
zeta_prev[s] = z;
zeta_curr[s] = z_new;
}