Function moe_expert_gemv_gate_up 

pub fn moe_expert_gemv_gate_up(
    gpu: &dyn GpuBackend,
    kernel: KernelHandle,
    input: DevicePtr,
    gate_packed_ptrs: DevicePtr,
    gate_scale_ptrs: DevicePtr,
    gate_scale2_vals: DevicePtr,
    gate_out: DevicePtr,
    up_packed_ptrs: DevicePtr,
    up_scale_ptrs: DevicePtr,
    up_scale2_vals: DevicePtr,
    up_out: DevicePtr,
    expert_indices: DevicePtr,
    n: u32,
    k: u32,
    top_k: u32,
    stream: u64,
) -> Result<()>

Fused gate+up expert GEMV: both projections in one kernel launch.

blockIdx.z selects gate (0) vs up (1). Saves 48 launches per decode step.

Grid: (ceil(N/4), top_k, 2) Block: (128, 1, 1)