Support for Numpy convolve mode 'valid'

Feature request

I’ve been working with model simulations and there’s the need for np.convolve(a, k, mode='valid').

I wish to have support for this for use with @jit/@njit decorators.

I have this example implementation from one of the discussions about using Numba for such situation, but I’d rather not have to resort to it in projects which require valid mode.

def numba_convolve_mode_valid_as_loop(arr, kernel):
        m = arr.size
        n = kernel.size
        out_size = m - n + 1
        out = np.empty(out_size, dtype=np.complex128)
        kernel = np.flip(kernel)  # wrong results otherwise
        for i in range(out_size):
            out[i] = np.dot(arr[i:i + n], kernel)
        return out

[jeertmans]

Hello @danilo-bc, are you still working on this ? Otherwise, I would gladly give it a try!

[gmarkall]

Labelling as “good second issue” as most of the parts needed for the implementation are already present, but need connecting together. Se e.g. https://github.com/numba/numba/blob/master/numba/np/arraymath.py#L4230-L4251.