Function moe_expert_gate_up_shared_bf16 

pub fn moe_expert_gate_up_shared_bf16(
    gpu: &dyn GpuBackend,
    kernel: KernelHandle,
    input: DevicePtr,
    gate_weight_ptrs: DevicePtr,
    gate_out: DevicePtr,
    up_weight_ptrs: DevicePtr,
    up_out: DevicePtr,
    expert_indices: DevicePtr,
    sh_gate_weight: DevicePtr,
    sh_gate_out: DevicePtr,
    sh_up_weight: DevicePtr,
    sh_up_out: DevicePtr,
    n: u32,
    k: u32,
    top_k: u32,
    stream: u64,
) -> Result<()>

Fused gate+up expert GEMV with shared expert for BF16 weights.

BF16 variant of moe_expert_gate_up_shared_fp8: no scale tables, direct BF16 weight pointers. For models loaded via the FP8-dequant-on-load path.

Grid: (ceil(N/8), top_k+1, 2) Block: (128, 1, 1)