c9e1719098
First round of changes for Richardson implementation. Zero sweeps is the identity.
5 years ago
8f73ddfbce
Streamline GS first application step when INIT='Z'
5 years ago
f496d4856b
Always make a GEN_BLOCK copy of the descriptor at fine level.
5 years ago
49259c79b3
Introduced solver%dump GLOBAL_NUM for in-house solvers
5 years ago
2894a0944b
Implement global_num for smoother%dmp.
5 years ago
9c95835ed5
Fix clone for L1_JAC
5 years ago
6259514cd1
Remove temporary implementation of L1
5 years ago
7fe0eb8580
New L1-JAC implementation.
5 years ago
d249042ea2
Modified implementation for residual check/print
5 years ago
adc5aebd6b
In L1-GS only add nonzero D1 entries.
5 years ago
108b4dd00d
Fixes for L1-JAC vs Gauss-Seidel
5 years ago
d2264f5f11
Stop criterion for aggregation.
5 years ago
8e2af97a35
Fix description of L1-GS
5 years ago
4260dc74d5
Support for L1- variants of hybrid Gauss-Seidel
5 years ago
940609564f
Fix handling of L1-BJAC in PREC%SET
5 years ago
d81bb30844
Fixes for save/restore smoothers using clone_settings in hierarchy_bld
5 years ago
d55dd1f21b
New clear_data and clone_settings methods for ILU
5 years ago
9ba74e89cd
Added clear_data and clone_settings to GS solver
5 years ago
07a2a97b13
Added clear_data to diag solver
5 years ago
dad39223ec
New clear_data and clone_settings methods for smoothers.
5 years ago
94a412e92a
Fix typo in clone_settings
5 years ago
8b0577069f
Define clone_settings and clear_data in base smoothers and solvers
5 years ago
32d550ed05
Improve use of TRIM in precinit
5 years ago
4d1a87e5d8
Added L1-JACOBI and L1-BJAC as simple preconditioners in PRECINIT
5 years ago
8f60a49fc6
Fix jac%cnv
5 years ago
de75eec402
Fix build of BJAC diagonal block factorization.
5 years ago
8f167a3295
Fix typo in cprecset.F90
5 years ago
0854eee936
New L1-BJAC smoother cleanups.
5 years ago
441c607c4a
Added L1-BJAC smoother.
5 years ago
6ac8cffa0a
Merge branch 'stopcriterion' into unify_aggr_bld
5 years ago
c0b033da57
Take out the backfix method.
...
Originally it was introduced to handle cases where the stencil was
expanded by the smoothing and multiplication process, but the
critical step is handled in AGGR%BLD_MAP, this one can be discarded.
5 years ago
e0491f1f17
Changed selection criterion between Jacobi/BJAC application
5 years ago
e680f2092f
Merge branch 'development' into stopcriterion to correct memory leak
5 years ago
67cff859ec
Fix lev%free() for sm2a
5 years ago
78e7f29371
Corrected print routine to avoid double print of last residual
5 years ago
87bcae4a28
Merge branch 'development' into stopcriterion
5 years ago
837a5b1a1d
Added default behavior and psb_toupper function in string check
5 years ago
825037761f
Reworked CSETC methods
5 years ago
6ac3e6c146
Added stop criterion on residual for BJAC as coarse solver
5 years ago
70a85c6972
Fix handling of level sweeps in the case of global solver and print warning.
5 years ago
8006626693
Mumps: use izero instead of 0 in calling sequence
5 years ago
0380a90ead
Merge branch 'unify_aggr_bld' into development
5 years ago
9ac2a27077
Merge branch 'smoothed-parm' into unify_aggr_bld
5 years ago
2f1d6e29bb
Fixes for interfacing with multiple versions of SuperLU_DIST
5 years ago
cb0e726980
New call ac_coo%fix
5 years ago
fa4623f54b
Ensure using correct NZ value
5 years ago
028ccea2e3
Fix handling of aggregator/smoother parms upon resizing
5 years ago
d2aeeb9dae
Improve structure of Makefiles
5 years ago
e4e7d8970e
Move call to ac%fix
5 years ago
6aed8608ca
New version X_spmm_bld_inner
5 years ago
Salvatore Filippone
Cirdans-Home
Salvatore Filippone