|
Mallard 0.8.0
Unstructured, GPU accelerated, high-order finite volume CFD solver
Mallard website |
Moving per-item values from their owner ranks to contiguous blocks in global-id order, for writes at global offsets. More...
#include <cstdint>#include <span>#include <utility>#include <vector>Go to the source code of this file.
Classes | |
| class | GlobalOrder |
| Items (cells, nodes) spread over the ranks, each with a global id in [0, n_global), regrouped so that rank r holds the ids [block_begin(n_global, r, p), block_begin(n_global, r + 1, p)) in order. An id may come from several ranks (shared nodes); it then takes any one of their values, so they should agree. Ids no rank sends are left zero. Without distributed, this rank holds every id and the regrouping is a local permutation (not collective). More... | |
Moving per-item values from their owner ranks to contiguous blocks in global-id order, for writes at global offsets.