kernel

method of dascore.proc.window.Windowed source

kernel(
    self ,
    data ,
    axes ,
    size ,
    stride ,
    length ,
    others ,
    factor ,
    halo ,
    inner ,
    quiet ,
    like = None,
    tiles = None,
)

Return the windows the operation made, blended, on the data’s backend.

As numpy_kernel, for any other array library, so the data stay where they are. A dask array is computed a chunk at a time, each from the windows which overlap it; see _by_dask_chunk. For any other library the windows are cut by slicing, the operation’s kernel for the library is run on each, and the results are put together by concatenation and addition, with nothing written in place, which not every array library allows.