pub fn kquant_mmvq_pair_w(
gpu: &dyn GpuBackend,
kernel: KernelHandle,
(w0, out0, n0): (DevicePtr, DevicePtr, u32),
(w1, out1, n1): (DevicePtr, DevicePtr, u32),
y_q8: DevicePtr,
k: u32,
m: u32,
stream: u64,
) -> Result<()>Expand description
2026-09-25: kquant_mmvq_q2_k_pair_w: two [n_i, k] Q2_K projections of one activation
(y_q8, block_q8_1 [m][k/32]) in one launch, out_i = [m][n_i] bf16. Grid y picks the
projection, and each runs the same device function as kquant_mmvq_w.