pub fn moe_expert_down_act_bf16_grouped(
gpu: &dyn GpuBackend,
kernel: KernelHandle,
act: DevicePtr,
down_ptrs: DevicePtr,
output: DevicePtr,
expert_offsets: DevicePtr,
active_experts: DevicePtr,
active_count: DevicePtr,
sh_act: DevicePtr,
sh_down: DevicePtr,
sh_down_out: DevicePtr,
n: u32,
k: u32,
cap: u32,
num_tokens: u32,
stream: u64,
) -> Result<()>Expand description
2026-10-02: Grouped BF16 down over the hi + lo SiLU rows. k is the intermediate width; rows
of act and output are sorted positions, rows of sh_act and sh_down_out tokens.