HyprePreconds
(13 files, 28 lines)
HypreSolvers
(19 files, 9 lines)
[root]/Packages/Uintah/CCA/Components/Solvers
HyprePreconds
(13 files, 28 lines)
HypreSolvers
(19 files, 9 lines)
| Author | Changes | Lines of Code | Lines per Change |
|---|---|---|---|
| Totals | 464 (100.0%) | 29725 (100.0%) | 64.0 |
| livne | 393 (84.7%) | 29212 (98.3%) | 74.3 |
| worthen | 21 (4.5%) | 240 (0.8%) | 11.4 |
| harman | 29 (6.3%) | 94 (0.3%) | 3.2 |
| jas | 6 (1.3%) | 72 (0.2%) | 12.0 |
| dav | 7 (1.5%) | 58 (0.2%) | 8.2 |
| jones | 4 (0.9%) | 31 (0.1%) | 7.7 |
| dgroulx | 4 (0.9%) | 18 (0.1%) | 4.5 |
cleaned house, still more to do.
0 lines of code changed in 4 files:
Only normalize the residual during the setup phase. Theoretically, we should
normalized during both the setup and step2 of the algorithm. However, the residual
is computed slightly differently in step 2 and the residual is somehow normalized.
0 lines of code changed in 2 files:
Some AMR stuff, particularly consolidating the range-queriers for coarse/fine
level interactions. This was done in AMRICE, impAMRICE, MPMICE, MPM, and
HypreDriverSStruct
AdiabaticTable: Do errorEstimate on ICE matls
AMRICE: also add if(doICEonLevel) to some AMR functions
MPMICE: also add MPMICE::refine
Grid: Change the invalidGrid warning from checking to see
if there is an even number of cells to checking if
the interior corners of patches line up with coarse cells.
AMRInterpolate: add the range queriers.
3 lines of code changed in 1 file:
No need to normalize the residual, it already is.
0 lines of code changed in 2 files:
Added ability to normalize the residual before checking for convergence,
by default this normalization factor = 1.0
CGSolver.cc
- nuked all the #if OPT sections
58 lines of code changed in 2 files:
Turned off printing the matrix, rhs and X to a file every timestep.
3 lines of code changed in 1 file:
Changed cout_dbg to cerr in printing hypre amr solver info.
6 lines of code changed in 1 file:
AMR Updates - get imp AMR working better in parallel (still crashes in hypre code), multi-level ICE with MPM mostly works AMRICE.cc - coarsen/reflux to ice matls only, and my 'boundary' hack in an '#ifdef BRYAN' blockICE.cc - set the surrounding matl at problemSetup time, some more debugging info for when the delt is too small
180 lines of code changed in 10 files:
added iterator limits when solution is copied back, only when X is modified.
3 lines of code changed in 1 file:
added virtual function getName() which returns the solver name.
27 lines of code changed in 15 files:
AMG seems to work without crashing, matrix looks ok. Need to double check results of 2-level implicit AMR ICE example problem with Todd.
149 lines of code changed in 11 files:
Matrix is now symmetric, corrected a few bugs at C/F interface matrix construction. But C/F fluxes taken from ICE are still mis-scaled and should be fixed next.
85 lines of code changed in 1 file:
Put #ifdefs around FAC only so that CG and Hypre 1.7.7 are supported. Not clear if AMG and 1.7.7 supported yet.
36 lines of code changed in 8 files:
Working on getting the matrix to be correct in AMRSolver. Added solver component TODO list file.
147 lines of code changed in 10 files:
AMG solver produces results but diverges and corrupts sp_vol_fracX_FC
142 lines of code changed in 5 files:
Added C-F, F-C connections to graph, matrix, and FAC, AMG solver interfaces
1193 lines of code changed in 13 files:
This should keep hypre-1.9 and non-hypre-1.9 developers happy (whoops - missed this one)
7 lines of code changed in 1 file:
Along with Dav: Added #ifdef HAVE_HYPRE_1_9 in configure.ac so that if we are using Hypre 1.9.0, the AMR solver code compiles, otherwise does not
128 lines of code changed in 10 files:
Use ifdef instead of if. These will be replaced with a better mechanism soon.
2 lines of code changed in 1 file:
Hack to make SGI to compiler happy. (Same thing that we need previously in MPMICE.cc.)
33 lines of code changed in 1 file:
(59 more)