1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
|
# Changes: 2.1.6
```{rubric} General:
```
- Added support for Cray X1
- Added option '-suggest' to petscarch script. It suggests some
possible PETSC_ARCHes to use
- Installation instructions are rewritten
```{rubric} Vec:
```
- VecNorm now caches NORM_2
- Added VecStrideScale(), VecStrideGather(), VecStrideScatter()
- Added VecBlockNorm(),VecBlockMax(),VecBlockMin(),VecBlockScale()
```{rubric} Mat:
```
- The interface to external packages that support direct solvers has
been reorganized. This is implemented using new matrix types
MATSEQAIJSPOOLES, MATMPIAIJSPOOLES, MATSEQSBAIJSPOOLES,
MATMPISBAIJSPOOLES, MATSUPERLU, MATSUPERLU_DIST,
MATUMFPACK,MATESSL, MATLUSOL, MATAIJMUMPS, MATSBAIJMUMPS,
MATDSCPACK
- Added new convenience matrix types which automatically switch
between Seq and MPI implementations depending on the size of the
communicator used during creation: MATAIJ, MATBAIJ, MATSBAIJ,
MATMAIJ, MATDENSE, MATBDIAG
- Changed default matrix-free compute h to use WP since it uses
fewer reductions
- Replace the datatypes MatLUInfo and MatILUInfo by MatFactorInfo
- Added cholesky support to SeqAIJ
- Added support for SUBSET_NONZERO_PATTERN with MatAXPY for AIJ &
BAIJ
- Added MatIsSymmetric() for AIJ
- Added LDA support in SeqDense
- The ncols argument made optional in MatGetRow()
- Added MAT_KEEP_ZEROED_ROWS for MPIROWBS
- Removed -aij_oneindex support from AIJ code
- Added MatPartitioningSetPartitionWeights() - usable with ParMetis
- MatGetSubMatrices() now works for ROWBS
```{rubric} PC:
```
- Added PCICCSetDamping() or -pc_icc_damping
- Added -pc_constant_null_space
- Added PCILUSetShift() or -pc_ilu_shift to support damping using
the identity
- Added PCLUSetShift() or -pc_lu_shift to support damping using the
identity
- Added PCICCSetShift() or -pc_icc_shift to support damping using
the identity
```{rubric} KSP:
```
- Added a new ksp_type LGMRES
- Added KSPSetComputeSingularValues() support to ksp_type FGMRES
```{rubric} TS:
```
- Added support for Runge-Kutta using ts_type TS_RUNGE_KUTTA
```{rubric} DMMG:
```
- -dmmg_jacobian_period -1 => jacobian is frozen after initial
construction
```{rubric} DA:
```
- Allow user preallocation of jacobian using DASetGetMatrix()
- DAGetMatrix now correctly returns the requested type.
- -mat_type typename overloads the requested type for DAGetMatrix
```{rubric} SYS:
```
- Added new viewer_type PETSC_VIEWER_NETCDF, PETSC_VIEWER_HDF4
- PETSc error handlers now trap many more signals.
```{rubric} Fortran:
```
- PetscPushErrorHandler() and PetscPopErrorHandler() now supported
- PetscPrintf() and PetscSynchronizedPrintf() now work in fortran
with a single string
```{rubric} External Packages:
```
- Added TSPVodeGetParameters()
- Parmetis-3 support added using MatPartitionSetNParts(),
ISPartitionCount()
- MUMPS support added, including complex (using mat_type
MATAIJMUMPS, MATSBAIJMUMPS)
- Updated SuperLU_DIST interface to SuperLU_DIST_2.0
- Added Spooles complex support
- Added Spooles cholesky for SeqAIJ
- ADIC: ad_GRAD_MAX is now set in the application and compiled into
code when compiling appplication
- The following variable names - in the buildsystem are changed -
that correspond with SuperLU_DIST SUPERLUDIST_INCLUDE ->
SUPERLU_DIST_INCLUDE SUPERLUDIST_LIB -> SUPERLU_DIST_LIB
PETSCHAVE_SUPERLU -> PETSC_HAVE_SUPERLU
|
