CFD/hypre
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
9,383 Commits 1 Branch 84 Tags 207 MiB
acdd1951ca
Commit Graph

1105 Commits

Author SHA1 Message Date
falgout
46488e8cbc Added HYPRE_Complex and HYPRE_Real types in place of double. 
Added an example code to test CG on a 4D HYPRE_SSTRUCT complex problem.
Added regression tests for bigint, maxdim, and complex.
Added a test to make sure double types are not added to the source.
See [Issue995] in the tracker for more details.
 2013-10-11 19:48:06 +00:00
kolev1
ebed231c65 In AMS, do not assume that the entries of the discrete gradient G are +1/-1. 
This allows the solver to work for AMR problems on quad/hex meshes, where the
reduced G can have more than two nonzero entries per row incorporating the AMR
vertex interpolation weights.
 2013-08-10 18:55:38 +00:00
kolev1
9323e57379 Patch from Jerome Solberg to fix zero rows in the Pi matrices, which can 
appear due to contact boundary conditions.
 2013-07-02 20:55:05 +00:00
kolev1
d6869b1689 Removed white space between backslash and newline (PGI compiler warning). 2013-04-19 17:43:06 +00:00
kolev1
28e747336f Updated the HYPRE_AMSSet*AMGCoarseRelaxType() documentation to list all 
coarse grid direct solver options.
 2013-02-14 19:40:11 +00:00
ulrikey
49631c7107 added new relaxation types for old Gaussian elimination on coarsest level: 
19 for plain GE, 29 for GE with pivoting
 2013-02-14 18:52:28 +00:00
ulrikey
551d0d1935 fixed a memory bug 2013-02-13 01:09:16 +00:00
ulrikey
2e073c2391 fixed bug when using L1-Jacobi on coarsest grid with CF turned on 2013-01-28 22:05:29 +00:00
kolev1
6eba36ed7d Added user functions HYPRE_AMSSet*AMGCoarseRelaxType for specifying the coarsest 
level relaxation for the AMG subspace solvers in AMS.

The default value corresponds to l1-GS on the coarsest level. This is different
than BoomerAMG's default (exact solve), since the AMS subspace matrices are
often singular in applications.
 2013-01-23 23:34:51 +00:00
ulrikey
f3465a64f7 fixed a potentially uninitialized variable 2013-01-15 15:27:23 +00:00
ulrikey
23df99bdbb fixed use of uninitialized variable 2013-01-11 15:13:26 +00:00
ulrikey
b5a7aa32f8 added check for unfreed memory and free it if necessary for the case of calling repeated Setups 
without calling Destroy.
 2013-01-10 19:14:33 +00:00
falgout
7038621993 Apparently 'cp -fa' does not work on Solaris, so changing to 'cp -fR' after all. 2013-01-10 15:59:09 +00:00
falgout
d696f7c535 Changed 'cp -fpPR' to 'cp -fa' in all of makefiles. 
The PETSc developers have had trouble on some platforms with -fpPR, but the -fa
option seems better than the -fR option they suggested because it will also
preserve attributes.  See [issue1022].
 2013-01-10 00:32:15 +00:00
ulrikey
048753bd29 deleted a semicolon 2013-01-08 16:03:24 +00:00
ulrikey
8aedae3214 fixed various klocwork issues 2013-01-07 23:57:24 +00:00
kolev1
529b5f2d50 Made sure that one can select the l1-GS smoother only on the coarsest grid. 2012-12-08 01:24:30 +00:00
kolev1
b7bbc6d215 Changed the code slightly so it doesn't break for matrices with zeros on the 
diagonal.
 2012-12-07 19:06:38 +00:00
falgout
2a88dea460 Changed a comment so that the 'check-mpi' script wouldn't complain. 2012-11-21 16:04:52 +00:00
ulrikey
431c900054 fixed several memory leaks 2012-11-20 15:42:02 +00:00
ulrikey
521f0320e4 added hypre-prefix 2012-11-20 15:17:25 +00:00
ulrikey
684df1fce4 added a switch to allow for agglomeration instead of redundant coarse grid solve. 
Redundant coarse grid solve now requires additional call to HYPRE_BoomerAMGSetRedundant.
 2012-11-19 21:50:10 +00:00
ulrikey
9af3123b15 fixed bug which potentially caused problems for distance-two interpolation used with nodal coarsening. 2012-11-05 18:45:25 +00:00
falgout
b5ffd1ed56 Changed the (diagonal==0) case in a couple of interpolation routines to avoid 
possible divide-by-zero by setting interpolation in that row to all zeros.
 2012-10-18 17:46:30 +00:00
falgout
7da71151f0 Added some debug printing of interpolation operators 2012-10-17 17:01:47 +00:00
kolev1
2469fbf2f4 Fixed a bug in the l1-GS smoother with OpenMP. 2012-09-25 21:23:47 +00:00
kolev1
9f385dd9fa Fixed a bug. 2012-06-18 16:19:15 +00:00
falgout
b335ec98aa Fixed some strict checking errors associated with type hypre_MPI_User_function. 
Updated the check-mpi autotest script to catch MPI_Comm_size, etc.
 2012-05-25 21:00:52 +00:00
ulrikey
a1398aa709 fixed bug and memory leak 2012-05-25 17:19:52 +00:00
ulrikey
163a13bbb0 fixed a memory leak 2012-05-24 23:56:22 +00:00
ulrikey
ea4b7462f9 generalize communicator in assumed partition and moved Gussian elimination 
matrix iteration into setup phase
 2012-05-24 22:46:26 +00:00
kolev1
fe88064d43 Added BoomerAMGSetMinCoarseSize(2) calls to the subspace solvers. 2012-05-10 20:33:25 +00:00
kolev1
6d84a472e9 Initialize the matrix pointers SN and AN in hypre_BoomerAMGSetup with NULL. 2012-05-10 20:32:48 +00:00
ulrikey
ca180796c9 changed default for min_coarse_size 2012-05-08 22:00:18 +00:00
ulrikey
df01ecf036 added new option which allows to set a minimal coarse grid size 2012-05-08 21:33:00 +00:00
falgout
a478ca3974 Made a few more klocwork related fixes. 2012-05-04 17:47:55 +00:00
falgout
6cec0baa42 Made a number of Klocwork fixes and removed old 'ierr' stuff as well. 2012-05-03 17:17:14 +00:00
kolev1
f45e65df25 Fixed a klocwork warning in AMEDestroy(). 2012-04-30 18:30:52 +00:00
kolev1
891e32d672 A few changes to address some issues found by the Klocwork static analyzer. 2012-04-25 22:37:16 +00:00
schroder2
50a94d0efc See issue 910. Fixed accidental usages of the built-in "int" type in 
par_multi_interp.c to use HYPRE_Int.
 2012-04-25 15:41:54 +00:00
schroder2
396ae05798 Added threading to par_multi_interp.c (multi-pass interpolation routine). 
Both phases of the multi-pass routine have some threading logic, the
sparsity structure computation phase and interpolation weight computation
phase.
 2012-04-24 22:23:57 +00:00
kolev1
953c25420d Testing for cvs commit quota error. 2012-04-23 20:48:45 +00:00
falgout
78ae8340d3 Added some variable initialization to fix the Visual Studio openmp compile. 
It's not clear that these were actually errors.
 2012-04-14 00:26:10 +00:00
falgout
0332374491 Changed the OpenMP implementation from include-file based to hard-coded omp 
pragmas in order to get around a compiler issue with Visual Studio.
 2012-03-16 22:46:05 +00:00
falgout
de1c289538 Made some changes to get additional configuration options working. Not sure yet 
how to handle setting MPI flags in general.
 2012-03-14 00:30:30 +00:00
falgout
b5557b0985 Removed references to 'headers.h' from some Makefiles and a few C files. 2012-03-06 21:46:12 +00:00
falgout
61fd2bdf64 Removing headers.h files that are no longer needed. 2012-03-06 21:31:43 +00:00
falgout
52796d3aea First implementation of CMake build. 2012-03-06 19:15:29 +00:00
schroder2
062e06e519 Fix for issue 864, basically removed use of the "collapse" flag in an 
OMP parallel for statement.  This flag is only supported in OMP v3.0 and later,
hence the error on dawndev.  This flag was simply removed because it was
unnecessary.  I was trying to tell OMP to only collapse the outermost for
loop, but this is already the default behavior.
 2012-02-23 01:58:50 +00:00
schroder2
f68b8c4097 Added threading to par_lr_interp.c:hypre_BoomerAMGBuildExtPIInterp() 2012-02-16 01:17:48 +00:00