Function moe_expert_gate_up_shared_fp8 

pub fn moe_expert_gate_up_shared_fp8(
    gpu: &dyn GpuBackend,
    kernel: KernelHandle,
    input: DevicePtr,
    gate_weight_ptrs: DevicePtr,
    gate_scale_ptrs: DevicePtr,
    gate_out: DevicePtr,
    up_weight_ptrs: DevicePtr,
    up_scale_ptrs: DevicePtr,
    up_out: DevicePtr,
    expert_indices: DevicePtr,
    sh_gate: &Fp8Weight,
    sh_gate_out: DevicePtr,
    sh_up: &Fp8Weight,
    sh_up_out: DevicePtr,
    n: u32,
    k: u32,
    top_k: u32,
    stream: u64,
) -> Result<()>

Fused gate+up expert GEMV with shared expert for FP8 weights.

FP8 variant of moe_expert_gate_up_shared: uses 2 pointer tables per projection (weight_ptrs + scale_ptrs) instead of NVFP4’s 3 (packed + scale + scale2). Shared expert weights are passed as direct Fp8Weight pointers.

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