Function nchwc_pool 

pub fn nchwc_pool(
    kind: PoolKind,
    input_shape: &[i64; 4],
    kernel_shape: &[i64; 2],
    dilation_shape: &[i64; 2],
    padding: &[i64; 4],
    stride_shape: &[i64; 2],
    output_shape: &[i64; 4],
    input: &[f32],
    output: &mut [f32],
)

Execute an NCHWc blocked 2-D pool using MLAS.

input_shape / output_shape are the blocked NCHWc shapes [N, round_up(C, block), H, W]; MLAS pools each channel independently on the blocked buffer, so callers keep the activation in NCHWc across the pool with no reorder. input / output are blocked buffers. Mirrors ONNX Runtime’s NchwcTransformer handling of pooling.