# tree_d_3d.kernels.evaluate_groups.evaluate_groups_numpy

*module*

The version of the ``evaluate_groups`` kernel for machines without Pyccel.

The walk of ``evaluate_groups_pyccel.py`` with the same arguments and the same terms for
every target, vectorized over (target, cell) pairs instead of walking one group at a time:
all targets descend the tree together, level by level, each pair is accepted as a
multipole, summed directly (a leaf) or replaced by the cell's four children. The sums
run in another order, so results agree with the Pyccel version to rounding. This is the
version Pyodide (the browser) runs.

[View source](https://github.com/max-models/tree-d-3d/blob/devel/src/tree_d_3d/kernels/evaluate_groups/evaluate_groups_numpy.py#L1-L1)

## tree_d_3d.kernels.evaluate_groups.evaluate_groups_numpy.evaluate_groups

*function*

```python
def evaluate_groups(targets, exclude, group_start, log_kernel, with_potential, spos, scharge, sindex, node_center, node_half, node_start, node_end, node_child, node_q, node_qabs, node_p, node_s, moment_re, moment_im, theta, softening, stack_size, phi, field)
```

Potential and field at the targets, without ``k_e``: see ``evaluate_groups_pyccel``.

The groups (``group_start``) and ``stack_size`` are not needed here: every target walks
the tree on its own, all of them at once.

[View source](https://github.com/max-models/tree-d-3d/blob/devel/src/tree_d_3d/kernels/evaluate_groups/evaluate_groups_numpy.py#L111-L205)
