pub fn q1_matvec(
model: &Arc<CmfModel>,
idx: usize,
xs: &[f32],
rows: usize,
cols: usize,
out: &mut [f32],
) -> boolExpand description
q1 matvec: raw f32 activations, tile-embedded scales. Metal only for now (wgpu q1 WGSL is queued); false = CPU fallback.