pub fn mla_q_rope_writeback(
gpu: &dyn GpuBackend,
kernel: KernelHandle,
q_rope_direct: DevicePtr,
q_absorbed_buf: DevicePtr,
nq: u32,
rope: u32,
kv_lora: u32,
mla_cache_dim: u32,
stream: u64,
) -> Result<()>Expand description
MLA Q_rope writeback: scatter RoPE’d rope portions to strided layout. 1 kernel replaces 32 D2D copies.