Fluidity

lp:~amcg-stokes/fluidity/multimaterial_diagnostic_dependencies

Created by Cian Wilson and last modified
Get this branch:
bzr branch lp:~amcg-stokes/fluidity/multimaterial_diagnostic_dependencies
Members of AMCG Stokes can upload to this branch. Log in for directions.

Branch merges

Related bugs

Related blueprints

Branch information

Owner:
AMCG Stokes
Project:
Fluidity
Status:
Merged

Recent revisions

4138. By Cian Wilson

Merging in changes from lp:fluidity.

4137. By Cian Wilson

A fix for the failing test case. Do not consider all states unless the algorithm has multiple material_phase_support.

4136. By Rhodri Davies

Add field names to write statments.

4135. By Cian Wilson

Fix now that there is an extra inner loop over states we're cycling over.

4134. By Cian Wilson

Merging in changes from lp:fluidity.

4133. By Cian Wilson

Diagnostic dependencies with unspecified state names are now checked in all available states rather than just the current state. This allows bulk multimaterial properties to correctly automatically work out their dependencies when we don't know what the state names are going to be. This has been tested on bulk_viscosity when it depends on MaterialViscosity in multiple states with unknown names. Needs to be run through buildbot to see how it affects other tests.

4132. By Mark Filipiak

Changes the mesh decomposition tool from fldecomp to flredecomp.

Previously, fldecomp was used for the initial decomposition of a mesh
 before running fluidity and flredecomp was used for re-decomposition
 after a checkpoint. fldecomp is a stand-alone program; flredecomp is
 part of Fluidity, using its input routines and its interface to the
 Zoltan partitioner. Vertex and element re-numbering is being
 developed in Fluidity to improve the NUMA and cache performance.
 Using flredecomp for all mesh decomposition means that the
 re-numbering needs to be implemented in Fluidity only and not in
 fldecomp as well.

Most of the change is simply converting the tests from the fldecomp
 work-flow:

<generate mesh>
 fldecomp -n <nparts> <other options> mesh
 mpiexec -n <nparts> fluidity <other options> sim.flml

to the flredecomp work-flow:

<generate mesh>
 mpiexec -n <nparts> flredecomp -i 1 -o <nparts> sim sim_flredecomp
 mpiexec -n <nparts> fluidity <other options> sim_flredecomp.flml

These changes have been implemented in all of the parallel tests,
 longtests, and examples.

Code changes are:

1/ For spherical meshes, the spatial dimension as well as the
 topological dimension is broadcast by insert_external_mesh
 (Populate_State.F90). Both dimensions are needed in flredecomp.

Users will still have fldecomp in their work-flow, so fldecomp has not
 yet been removed. It now prints a message recommending that the
 decomposition is changed to use flredecomp, and warning that fldecomp
 will be removed in a future release of Fluidity.

flredecomp cannot decompose Terreno meshes, so fldecomp has to be
 retained until any Fluidity simulations using Terreno meshes are
 converted to using the mesh extrusion available in Fluidity.

4131. By Patrick Farrell

Look for a CGAL function in CGAL, not cos. [Build fix for --enable-cgal flag]

4130. By Mark Filipiak

Input of environmental options (e.g. forcing fields) is moved from
Usage.cpp into ocean_forcing/InitialiseOceanForcing.F90, to be
accessible in flredecomp as well as fluidity. Relevant Fortran-C
interfaces have been converted from using F77_FUNC to use
iso_c_binding.

4129. By Adam Candy

Improvements to the backward-facing step 2d kepsilon example - hopefully this will fix the buildbot failure.

Branch metadata

Branch format:
Branch format 7
Repository format:
Bazaar repository format 2a (needs bzr 1.16 or later)
Stacked on:
lp:fluidity
This branch contains Public information 
Everyone can see this information.

Subscribers