How to use the clifford.tools.g3c.cuda_products.gmt_func function in clifford

def rotor_between_objects_device(L1, L2, rotor):
    L1sqrd_val = numba.cuda.local.array(32, dtype=numba.float64)
    gp_device(L1, L1, L1sqrd_val)
    if L1sqrd_val[0] > 0:
        C_val = numba.cuda.local.array(32, dtype=numba.float64)
        sigma_val = numba.cuda.local.array(32, dtype=numba.float64)
        k_value = numba.cuda.local.array(32, dtype=numba.float64)
        gp_device(L2, L1, C_val)
        C_val[0] += 1.0
        gp_mult_with_adjoint(C_val, sigma_val)
        positive_root_device(sigma_val, k_value)
        annhilate_k_device(k_value, C_val, rotor)
    else:
        L21 = numba.cuda.local.array(32, dtype=numba.float64)
        L12 = numba.cuda.local.array(32, dtype=numba.float64)
        gp_device(L2, L1, L21)
        gp_device(L1, L2, L12)
        sumval = 0.0
        for i in range(32):
            if i == 0:
                sumval += abs(L12[i] + L21[i] - 2.0)
            else:
                sumval += abs(L12[i] + L21[i])
            rotor[i] = -L21[i]

def rotor_between_lines_device(L1, L2, rotor):
    L21_val = numba.cuda.local.array(32, dtype=numba.float64)
    L12_val = numba.cuda.local.array(32, dtype=numba.float64)

    gp_device(L2, L1, L21_val)
    gp_device(L1, L2, L12_val)

    beta_val = numba.cuda.local.array(32, dtype=numba.float64)
    K_val = numba.cuda.local.array(32, dtype=numba.float64)
    for i in range(32):
        K_val[i] = L21_val[i] + L12_val[i]
        beta_val[i] = 0.0
    K_val[0] += 2.0

    project_val_cuda(K_val, beta_val, 4)

    alpha = 2.0 * K_val[0]

    denominator = math.sqrt(alpha / 2)

    normalisation_val = numba.cuda.local.array(32, dtype=numba.float64)

def apply_rotor_device(mv, rotor, output):
    rotor_adjoint = numba.cuda.local.array(32, dtype=numba.float64)
    temp = numba.cuda.local.array(32, dtype=numba.float64)
    adjoint_device(rotor, rotor_adjoint)
    gp_device(mv, rotor_adjoint, temp)
    gp_device(rotor, temp, output)

def rotor_between_lines_device(L1, L2, rotor):
    L21_val = numba.cuda.local.array(32, dtype=numba.float64)
    L12_val = numba.cuda.local.array(32, dtype=numba.float64)

    gp_device(L2, L1, L21_val)
    gp_device(L1, L2, L12_val)

    beta_val = numba.cuda.local.array(32, dtype=numba.float64)
    K_val = numba.cuda.local.array(32, dtype=numba.float64)
    for i in range(32):
        K_val[i] = L21_val[i] + L12_val[i]
        beta_val[i] = 0.0
    K_val[0] += 2.0

    project_val_cuda(K_val, beta_val, 4)

    alpha = 2.0 * K_val[0]

    denominator = math.sqrt(alpha / 2)

    normalisation_val = numba.cuda.local.array(32, dtype=numba.float64)
    output_val = numba.cuda.local.array(32, dtype=numba.float64)

def annhilate_k_device(K_val, C_val, output):
    k_4 = numba.cuda.local.array(32, dtype=numba.float64)
    project_val_cuda(K_val, k_4, 4)
    for i in range(32):
        k_4[i] = -k_4[i]
    k_4[0] += K_val[0]
    gp_device(k_4, C_val, output)
    normalise_mv_device(output)

def gp_mult_with_adjoint(value, output):
    other_value = numba.cuda.local.array(32, dtype=numba.float64)
    adjoint_device(value, other_value)
    gp_device(value, other_value, output)