Function moe_w4a16_fused_gate_up_k64_n128 

pub fn moe_w4a16_fused_gate_up_k64_n128(
    gpu: &dyn GpuBackend,
    kernel: KernelHandle,
    a: DevicePtr,
    gate_packed_ptrs: DevicePtr,
    gate_scale_ptrs: DevicePtr,
    gate_scale2_vals: DevicePtr,
    up_packed_ptrs: DevicePtr,
    up_scale_ptrs: DevicePtr,
    up_scale2_vals: DevicePtr,
    c_gate: DevicePtr,
    c_up: DevicePtr,
    expert_offsets: DevicePtr,
    sorted_token_ids: DevicePtr,
    num_experts: u32,
    n_out: u32,
    k: u32,
    max_m_tiles: u32,
    stream: u64,
) -> Result<()>

K64 fused gate+up GEMM — zero pipeline stall for K=h (2048 for 35B), 32 K-steps vs 64.

Grid: (ceil(2*n_out/128), max_m_tiles, num_experts) Block: (128, 1, 1)