| // Split-K normalize pass. Each workgroup (row = wg.x, split index = wg.z) |
| // folds the SPLIT per-row partial sums of squares into the RMS scale, then |
| // normalizes its HIDDEN/SPLIT slice. SPLIT is small (<=64), so the serial fold |
| // avoids a third combine pass. Scale offsets follow the suffix-axis broadcast |
| // contract. |
| {% if usesF16 %} |
| enable f16; |
| {% endif %} |
| {{ env.wgsl.resourceDeclarations }} |
| |
| const HIDDEN: u32 = {{ hiddenSize }}u; |
| const EPSILON: f32 = {{ epsilon }}; |
| const WG: u32 = {{ workgroupSize }}u; |
| const SPLIT: u32 = {{ split }}u; |
| |
| {% if source.scaleRank > 0 %} |
| const X_RANK: u32 = {{ source.xRank }}u; |
| const SCALE_RANK: u32 = {{ source.scaleRank }}u; |
| const X_SHAPE: array<u32, {{ source.xRank }}> = array<u32, {{ source.xRank }}>({% for d in source.xShape %}{{ d }}u{% if not loop.last %}, {% endif %}{% endfor %}); |
| const SCALE_SHAPE: array<u32, {{ source.scaleRank }}> = array<u32, {{ source.scaleRank }}>({% for d in source.scaleShape %}{{ d }}u{% if not loop.last %}, {% endif %}{% endfor %}); |
| |
| fn x_stride(axis: u32) -> u32 { |
| var stride = 1u; |
| for (var i = axis + 1u; i < X_RANK; i += 1u) { |
| stride *= X_SHAPE[i]; |
| } |
| return stride; |
| } |
| |
| fn scale_stride(axis: u32) -> u32 { |
| var stride = 1u; |
| for (var i = axis + 1u; i < SCALE_RANK; i += 1u) { |
| stride *= SCALE_SHAPE[i]; |
| } |
| return stride; |
| } |
| |
| {% endif %} |
| fn scale_offset({% if source.scaleRank > 0 %}out_index: u32{% endif %}) -> u32 { |
| {% if source.scaleRank == 0 %} |
| return 0u; |
| {% else %} |
| var rem = out_index; |
| var offset = 0u; |
| for (var axis = 0u; axis < X_RANK; axis += 1u) { |
| let stride = x_stride(axis); |
| let coord = rem / stride; |
| rem %= stride; |
| let scale_axis = i32(axis) - i32(X_RANK - SCALE_RANK); |
| if (scale_axis >= 0) { |
| let s_axis = u32(scale_axis); |
| if (SCALE_SHAPE[s_axis] != 1u) { |
| offset += coord * scale_stride(s_axis); |
| } |
| } |
| } |
| return offset; |
| {% endif %} |
| } |
| |
| |
| @compute @workgroup_size(WG, 1, 1) |
| fn main(@builtin(workgroup_id) wg: vec3<u32>, @builtin(local_invocation_id) lid: vec3<u32>) { |
| let row = wg.x + wg.y * params.rowStride; |
| if (row >= params.rows) { |
| return; |
| } |
| let k = wg.z; |
| let tid = lid.x; |
| |
| var total = 0.0; |
| for (var i = 0u; i < SPLIT; i = i + 1u) { |
| total = total + partials[row * SPLIT + i]; |
| } |
| let inv = inverseSqrt(total / f32(HIDDEN) + EPSILON); |
| let chunk = (HIDDEN + SPLIT - 1u) / SPLIT; |
| let start = k * chunk; |
| var end = start + chunk; |
| if (end > HIDDEN) { end = HIDDEN; } |
| let base = row * HIDDEN; |
| |
| var d = start + tid; |
| loop { |
| if (d >= end) { break; } |
| let index = base + d; |
| // Preserve the ONNX stage boundary: round Normalized to X's dtype before |
| // the affine scale is applied. |
| let normalized = {{ xElement }}(f32(x[index]) * inv); |
| let value = f32(normalized) * f32(scale[scale_offset({% if source.scaleRank > 0 %}index{% endif %})]); |
| y[index] = {{ scalar }}(value); |
| d = d + WG; |
| } |
| } |
| |
