pub fn conv1d_update_prefill(
gpu: &dyn GpuBackend,
kernel: KernelHandle,
conv1d_prefill_tp_k: KernelHandle,
conv_state: DevicePtr,
input: DevicePtr,
weight: &DenseWeight,
bias: DevicePtr,
output: DevicePtr,
d_inner: u32,
d_conv: u32,
seq_len: u32,
input_stride: u32,
output_stride: u32,
stream: u64,
) -> Result<()>Expand description
2026-09-25: Conv1d update + SiLU over seq_len prefill tokens, with
separate per-token input and output strides. Runs the token-parallel kernel
(conv1d_prefill_tp_k) when its handle is non-zero and
METRALE_CONV1D_TP is not 0, else causal_conv1d_update_prefill, one
thread per channel walking the tokens.