Root level parts

Handles the case when the top-level cell size increases so that the cells need to be regenerated and particles re-attached.

The technique is to resample the old cell space and use the node IDs to populate the new one. This should minimise the need for particle movement, so should be a little less expensive than a full initial partition.

If the resample fails we fall back to a full initial partition. Could be a better idea to use the same technique as in main.c, but we don't do that.

Fix for issue #31 (closed)

Added 56 commits:

• ef920fec...3e0e0c69 - 55 commits from branch master
• 9ace9101 - Merge branch 'master' into root-level-parts

Title changed from WIP: Root level parts to Root level parts

Reassigned to @nnrw56

Pedro, so this is my solution to #31 (closed). The SedovBlast MPI test runs to completion for the first time. So better than what we do at the moment. If you're happy re-assign to @matthieu. Branch can be deleted.

Added 64 commits:

• 9ace9101...7af3ec24 - 63 commits from branch master
• 4ad7e742 - Merge branch 'master' into root-level-parts

Added 302 commits:

• 4ad7e742...cbefe349 - 301 commits from branch master
• 9ab8173e - Merge branch 'master' into root-level-parts

@nnrw56, have you had a chance to look into this ?

Genetic partitioning will improve upon this but this would allow us to run without the fear of the root-level repartitionning problem.

Sorry for the long delay. Yes, this looks good to me!

Reassigned to @matthieu

Status changed to merged

mentioned in commit ce5e778a

mentioned in issue #31 (closed)