lp:~amcg-stokes/fluidity/fluidity-block-velocity

Branch merges

Propose for merging

No branches dependent on this one.

Related source package recipes

No recipes using this branch. (?)

Create packaging recipe

Related snap packages

Related bugs

Related blueprints

3865. By Rhodri Davies on 2014-05-29

Merge in trunk.

3864. By Stephan Kramer on 2012-10-25

Change MatCreateSeqAIJ to the new MatCreate,MatSetSizes,MatSetType,MatSeqAIJSetPreallocation and also call MatSetBlockSizes() for each SeqAIJ that is created.
Block sizes are set to (1,1) normally and (u%dim,u%dim) for the velocity matrix.
group_size (the block size seen by petsc) is now added as attribute in petsc_numbering
This is now picked up in full_projection to get the group_size for ct_m from inner_m.

3863. By Stephan Kramer on 2012-10-25

Temporarily switching off baij format as it doesn't actually seem to work with gamg. Still keeping the new velocity ordering though.

3862. By Stephan Kramer on 2012-10-25

In full_projection we also need to tell the petsc_numbering about blocking the velocity components.

3861. By Stephan Kramer on 2012-10-24

Fix for resolved conflict: MAT_IGNORE_ZERO_ENTRIES should not be set for baij matrices.

Currently failing tests, after a "sed -i 's/eisenstat/sor/' tests/*/*.flml":
Stokes_mms_cg_p2p1_test_cty_cv.xml: FFFPPPFFFFFFFFFPPFFFFFFFF
Stokes_mms_cg_p1bp1.xml: F
Stokes_Subduction_VK_Case_1b.xml: PFFFFFFFFF
channel_wind_drag_rotated.xml: F
turbine_dirichlet_3d_moving_mesh.xml: F
turbine_dirichlet_3d.xml: F

Passes: 1357
Failures: 33
Warnings: 0

This is all to do with "mg" no working on baij velocity solves (last 4 tests) and sor for some reason being much slower than eisenstat so I had to kill it (first 2).

3860. By Stephan Kramer on 2012-10-24

Merging in from trunk.

3859. By Stephan Kramer on 2012-10-24

Reverting the last two commits:
1) changes in mg to deal with block matrices
2) an unfinished attempt at rescaling a la May
These may be useful at some point in the future, so we can restore them from history.

3858. By Stephan Kramer on 2012-01-04

Some work on scaling of the system. Does it work? No idea...

3857. By Stephan Kramer on 2011-12-14

The block based aggregation now sort of works (for one problem). The strong-connection criterion needs fixing though.

3856. By Stephan Kramer on 2011-12-11

Actually start using SeqBAIJ instead of SeqAIJ for velocity matrix.