997020ba81
New GELP for EPK data.
4 years ago
c414d4d40a
Merge branch 'implement-ainv' into dev-openmp
...
# Conflicts:
# base/comm/internals/psi_cswapdata_a.F90
# base/comm/internals/psi_cswaptran_a.F90
# base/comm/internals/psi_dswapdata_a.F90
# base/comm/internals/psi_dswaptran_a.F90
# base/comm/internals/psi_eswapdata_a.F90
# base/comm/internals/psi_eswaptran_a.F90
# base/comm/internals/psi_i2swapdata_a.F90
# base/comm/internals/psi_i2swaptran_a.F90
# base/comm/internals/psi_mswapdata_a.F90
# base/comm/internals/psi_mswaptran_a.F90
# base/comm/internals/psi_sswapdata_a.F90
# base/comm/internals/psi_sswaptran_a.F90
# base/comm/internals/psi_zswapdata_a.F90
# base/comm/internals/psi_zswaptran_a.F90
# base/modules/psi_i_mod.F90
4 years ago
a8ee595392
New method for checking if REPL at V descriptor
4 years ago
4674de97cf
Corrected call to use mpi
4 years ago
927bf9f103
Merge branch 'new-context' into implement-ainv
4 years ago
db9bb6ca77
Fix problem in logical recv
4 years ago
6866558372
Merge branch 'new-context' into implement-ainv
...
# Conflicts:
# prec/impl/psb_c_bjacprec_impl.f90
# prec/impl/psb_c_prec_type_impl.f90
# prec/impl/psb_d_bjacprec_impl.f90
# prec/impl/psb_d_prec_type_impl.f90
# prec/impl/psb_s_bjacprec_impl.f90
# prec/impl/psb_s_prec_type_impl.f90
# prec/impl/psb_z_bjacprec_impl.f90
# prec/impl/psb_z_prec_type_impl.f90
# prec/psb_c_prec_type.f90
# prec/psb_d_prec_type.f90
# prec/psb_s_prec_type.f90
# prec/psb_z_prec_type.f90
# test/pargen/psb_d_pde2d.f90
# test/pargen/psb_d_pde3d.f90
# test/pargen/psb_s_pde2d.f90
# test/pargen/psb_s_pde3d.f90
4 years ago
b3bcb762c5
New files from remap-coarse
4 years ago
2e0c297409
Merging remap-coarse into new-context
4 years ago
f0663512b2
Merge new-context
4 years ago
ea6e473677
Various CTXT fixes
4 years ago
e445819960
Fix wrong subroutine name ref.
4 years ago
025a4b5512
Fixes for CTXT usage, new GELP templated source.
4 years ago
42d9ad426e
Merge branch 'new-context' into dev-openmp
4 years ago
7eb653b39d
fix ICTXT vs CTXT
4 years ago
6b6ca3b6ac
Merge branch 'new-context' into dev-openmp
4 years ago
bab24cec27
Fix indx_map compilation
4 years ago
b32053d8aa
Fix base_init_null for use with remap-coarse
4 years ago
7124e8abdf
Merge branch 'new-context' into remap-coarse & fix
...
# Conflicts:
# base/modules/desc/psb_desc_mod.F90
# base/modules/penv/psi_penv_mod.F90
4 years ago
e1d859e3f5
Change return value of get_context on error.
4 years ago
9178d63ebd
Updates for C ctxt handling.
4 years ago
c52d42a50d
merge with mat-allocation
4 years ago
02b9ff5cfb
Fix Makefile dependencies
4 years ago
6a6f6ad2c2
Change name of ICTXT to CTXT
4 years ago
fbf23c3959
Added implementation in BJAC and test for ILU-type factorizations
4 years ago
8b2b86d44d
C interface. Not fully working yet.
4 years ago
b376d226cb
Fixed declaration and use of CONTEXT with new type.
4 years ago
2009ed8dbe
First round of updateds for new CONTEXT
4 years ago
2697fbe73a
Started work on encapsulating CONTEXT.
4 years ago
4bc326d06b
Add IDEST to remap interface
4 years ago
a79d853338
Debug check (to be taken out later)
4 years ago
fb422be9f4
Change interface of X_REMAP to include isrc/nrsrc
4 years ago
2090a011db
Include X_remap.
4 years ago
0adf2cbb3f
Seemingly working version of D_REMAP. Minimal test of same.
4 years ago
7c6ffcd4a2
Start work on REMAP.
4 years ago
0acf10f0d8
OpenMP in base_mat and coo
4 years ago
8d2ab75737
Added interfaces for all types of psb_nspaxpby
4 years ago
77730de80c
Put in OpenMP in CSR, first step.
4 years ago
23f69f58d4
Additional fixes for psi_serial_impl
4 years ago
eb513e45c6
OpenMP for base_vect
4 years ago
96a4db14fb
Rename psi_serial_impl to F90
4 years ago
069c49f20f
Rename base_vect to F90
4 years ago
4a274e43ff
cosmetic fixes
4 years ago
6e09f745ce
Merge branch 'development' into remap-coarse
4 years ago
c0c8224253
Corrected debug message in getelem
4 years ago
624a07a25b
Added getelem function to retrieve entries from distributed vectors
4 years ago
a1c2561de6
Cleanup unused variables.
4 years ago
dc36db6083
Reformat assignment of entire component.
4 years ago
7436cf564c
Start cleanup compile warnings
4 years ago
25fdc6fb73
VOLATILE and ISO_FORTRAN_ENV are now enforced, take out ifdefs
4 years ago
db7def0433
Merge branch 'development' into remap-coarse
4 years ago
64e4c194fd
Enable VECT objects in MMIO read/write.
4 years ago
7df7b6ffce
Set adjacncy list during build of halo_index for all maps.
4 years ago
e3649e1cb6
Do not use sorting on dependency lists unless dlavg<16
4 years ago
46736f9d39
Fix type handling bug in simple_triad
4 years ago
eb03797ad5
Better error messages from MAP%APPLY
4 years ago
db17057924
Cosmetic changes to fnd_owner for BLOCK.
4 years ago
97a1cfca3e
Merge branch 'unify_aggr_bld' into remap-coarse
4 years ago
76ff6c824c
Fix GEN_BLOCK distribution search for owner process: binary search on
...
repeated keys
4 years ago
6b2abed8bb
Reworked dep_list sorting. Take out obsolete functions
4 years ago
ceb79d8fd8
Fix bad argument passing for sort_dl
4 years ago
62a67a0d0e
Fix sorting of dep_list: store in CSR-like, handle self-loops
4 years ago
781f0ef083
New version of sort_dep_list.
4 years ago
56189f39fd
Start reworking of handling of dependency lists.
4 years ago
167f4157bb
Improve graph_fnd_owner
4 years ago
d9997e511f
Fixed alterations for MAP routines and new internals
4 years ago
90493218a2
Change handling of no-context in DESC for remap-coarse
4 years ago
2de1202d62
Improved error message for MAP_
4 years ago
d865bb91ca
Defined new variant of allocate_mnnz for mixed IPK/LPK
4 years ago
ad50c6e166
Add new source file for alternative CDASB.
4 years ago
5b1300b7cc
Merge branch 'fix_cdasb' into unify_aggr_bld
...
# Conflicts:
# base/internals/psi_extrct_dl.F90
4 years ago
d2f8d87204
Defined a threshold for enacting simplified exchange list build
4 years ago
2b295de9e7
New crea_index internals.
4 years ago
7fa4eae8c9
Timers in xtr_dl
4 years ago
debb03b6ae
Prepare for alternative code path
4 years ago
7de9318aa9
New timers in descriptor assembly
4 years ago
487b2c2e1d
Defined reallocate and safe_ab_cpy for intrinsic scalars.
4 years ago
e9a8814338
Change default order for DUPL constants.
5 years ago
6777e57d37
merge with merge-paraggr-newops
5 years ago
2a6107a937
Define psb_free_timers()
5 years ago
98349319e8
Cleanup DESC%FREE()
5 years ago
c0ccbe14f7
merge with unify_agg_build
5 years ago
f499ffb203
New clear_defaults function, use it in psb_exit
5 years ago
7d3e4aec06
New name qry_halo_owner to distinguish from fnd_halo_owner
5 years ago
84a8b73416
Avoid LX temporaries in X transpose
5 years ago
f28e3a9ea9
Further fixes for uninitialized vars
5 years ago
eb934e2a45
Fix uninitialized INFO in cp/mv _from
5 years ago
b56abbb270
Improve structure of Makefiles
5 years ago
58b7489db9
New fnd_owner implementation, taking into account CD%REINIT
5 years ago
055e342253
Move position of call to set_nzeros
5 years ago
26cfa837e5
Cosmetic changes for letter case conventions
5 years ago
55666cc0fe
Make sure p_adjcncy and halo_owner are preserved over desc%reinit
5 years ago
166bd219ef
Corrected checkmat bugs in compare and get_nnz routines
5 years ago
5c34e3d853
Removed orphan implementation for scale plus identity
5 years ago
9e347fae90
Removed orphan subroutines for matrix info
5 years ago
01f4f718de
Corrected implementation of scale plus identity for sparse matrices
5 years ago
91f737475e
Optimized version of X_glob_transpose: do not go through LX_glob_transpose.
5 years ago
dbc20d482e
In psb_rwextd use ensure_size instead of reallocate
5 years ago
1c23654328
Added functions to compare the entries of two matrices or every entry of matrix with a value
5 years ago
ca296fc0cf
Added missign @LX@ implementation of new sparse matrix sum routines
5 years ago
3b36c2196b
Added implementation of A = alpha A + beta B for sparse matrices
5 years ago
dc55713541
Added generic interface psb_scalplusidentity for long integer spmat
5 years ago
e108416719
Added function for scaled sparse matrix plus identity
5 years ago
814e09f5db
Added interface (and C interface) to is_bld() for sparse matrices
5 years ago
d92b6c02cf
Added interface (and C interface) to is_asb() for sparse matrices
5 years ago
c4cf78e7b8
Added interface (and C interface) to is_upd() for sparse matrices
5 years ago
a6db1b566c
Added Interface (and C Interface) for nnz of a sparse matrix
5 years ago
0f309850cc
Corrected routine for weighted masked 2norm
5 years ago
9e9a2e3975
Corrected implementation of @X@_base_inv_a2_check to be compliant with NVECTOR behavior
5 years ago
4c7fa23c8a
New files for 2-bytes integer support. To be compiled and tested.
5 years ago
5504e34161
Removed the out-interface for vector scaling
5 years ago
28c1efe650
Corrected pass argument in div routines
5 years ago
ea6ad3a85f
Corrected bug in out-of-place axpby
5 years ago
44f05510bc
Added out-of-place axpby and relative c interface
5 years ago
db22e832ee
Add constants for 2-byte integers
5 years ago
beeb123042
Fixed length char component.
5 years ago
8bf0ff673b
Added psb_minquotient function with C interface
5 years ago
99dc3f5d93
New par_spmm version with 4-byte matrices (with new sphalo variant)
5 years ago
fd89f2f1bf
Added psb_gescal subroutine to entrywise scale distributed vector with C interface
5 years ago
ea9c469159
Added psb_geaddconst subroutine to add constant value to vector with C-interfaces
5 years ago
439246e643
Added out-of-place entrywise division with C interfaces
5 years ago
22e7df48da
Added out-of-place psb_mlt subroutine interface
5 years ago
0edb671d21
Added minimum of realtype distributed vectors
5 years ago
15f89400c3
Correction to the psb_geinv implementation for out flag handling
5 years ago
55989d2e8f
Do not use allocatable characters
5 years ago
7f42d63275
Added routine for logical all reduce and applied in psb_mask
5 years ago
b8ab997bc1
Added generation of mask vector for constraint tests psb_mask
5 years ago
e59932e496
chagend name to acmp for the compare routines
5 years ago
87306f4d1a
added weighted and masked 2 norm, and compare on complex vector
5 years ago
1b214f3f00
added weighted 2-norm function
5 years ago
27bf7ec666
New get_a2av_alg() in place of conditional compilation
5 years ago
6b2fa31ae1
Internal doc changes.
5 years ago
63885c36c7
psb_geabs corrected version with test
5 years ago
f5709a2254
Added psb_geabs, final output not working
5 years ago
cc9ef42464
Renamed macro SP_A2AV_MAT into SP_A2AV_TRIAD
5 years ago
cdaae0b484
Fix and unify SPHALO variants implementations.
5 years ago
d00182ee30
Fix MPI type declaration
5 years ago
193682a5a5
Fix integer KIND for some temporaries.
5 years ago
01ffca76d2
Change mat%print interface, and refactor.
5 years ago
d3e4a091e8
Cosmetic cleanup
5 years ago
d342e2d59e
Fixed borderline case for PSB_CDALL( VL=())
...
Code was hanging if VL was empty on some processes.
5 years ago
a59c79a2f8
*** empty log message ***
5 years ago
2944299aa7
Check sizes in graph_fnd_owner sampling code.
5 years ago
d031ef0763
Fix typo in internal docs.
5 years ago
76d5c5f3ae
Fixes for cases where IPK==LPK
5 years ago
936ef49e7d
Further changes for ambiguous interfaces when IPK==LPK
5 years ago
3f08263eb6
Pretty-printing fixes
5 years ago
af14080a34
Initial fixes for IPK==LPK
5 years ago
1d81cf4af9
Cleanup transpose, use coo%ensure_size
5 years ago
487fa54f06
Define coo%ensure_size method
5 years ago
ae051a2ea1
Optimize data handling in transpose.
5 years ago
10f47d731d
Define versions of global transpose for IPK matrices
5 years ago
e2653b1c60
New clip "in place" method
5 years ago
77542b9023
New global transpose methods
5 years ago
3207596a75
Change name to psb_simple_triad
5 years ago
0aaaa477d4
New triad_a2av method
5 years ago
cea69503a4
New method for cleaning entries with negative indices in COO.
5 years ago
86872d9939
Define psb_simple_a2av
5 years ago
7c491f06f9
Fixed implementation of SPGATHER
5 years ago
4a89e4b1c7
Finish variable replacement
5 years ago
3cfdc033fd
New structure for A2A and graph_fnd_owner
...
New implementation for A2A: use mpi_allgather and mpi_reduce_scatter,
with two variants: one for variable inputs, one for fixed size
inputs.
Modified graph_fnd_owner to pass padded and uniform sized input to
A2A.
5 years ago
7c908c0849
Start mods for new fnd_owner
5 years ago
9a2c1fe0a9
Merge branch 'development' into fnd_owner
...
# Conflicts:
# base/internals/psi_extrct_dl.F90
# base/internals/psi_symm_dep_list.F90
# base/modules/desc/psb_indx_map_mod.f90
5 years ago
c3321aa981
Take out timings for now.
5 years ago
555907338e
New versions of fnd_owner: do not use MPI_alltoallv
5 years ago
77cdfd6cae
Merge branch 'merge-paraggr' into merge-paraggr-newops
...
# Conflicts:
# docs/html/index.html
# docs/html/userhtml.css
# docs/html/userhtml.html
# docs/psblas-3.7.pdf
# docs/src/psbrout.tex
5 years ago
41ef94538c
Additional improvements to internal docs.
5 years ago
e2bd21d9be
Update internal documentation.
5 years ago
92dae92582
Align with new implementation of extract_dep_list.
5 years ago
e97cab7ddc
Take out N from alloc_vect
5 years ago
4ed1ef4a0d
Additional internal docs for fnd_owner
5 years ago
c57aa2ac5c
Internal docs for fnd_owner variants.
5 years ago
e60e3c9d91
Changed default maxspace to -1
5 years ago
acae5f9278
Placed timing code in graph_fnd_owner
5 years ago
2b898aa14f
New get_mpi_comm and get_mpi_rank, code and docs
5 years ago
b9514ece40
New extract_dep_list from fnd_owner
5 years ago
43c1ccfc87
Extract_dep_list new symm
...
New dependency list symmetrizatoin, two alternatives: distributed, or
replicated after allgather.
5 years ago
0bacc130e7
Merge branch 'fnd_owner' into merge-paraggr
5 years ago
59e640c8a1
Fix error message for wrong IRST.
5 years ago
c845a7881e
Merge branch 'fnd_owner' of https://github.com/sfilippone/psblas3 into fnd_owner
5 years ago
70d19c04be
graph_fnd_owner cleanup
5 years ago
703214669a
Dead code cleanup
5 years ago
fcbe0d620a
Added vector/scalar comparison routine
5 years ago
5bd6f0cca9
Add comments to adjcncy_fnd_owner
5 years ago
d4fc54e75a
Corrected error in inv_vect implementation
5 years ago
f32f8e3387
Cosmetic changes in adjcncy_fnd_owner
5 years ago
caec98e942
New adjcncy and a2a fnd_owner
...
Reimplement adjcncy_fnd_owner to use alltoallv. Version with
sends/irecv still available under compile time constant.
Reimplement a2a_fnd_owner to use adjcncy_fnd_owner. Older version
still available under compile time constant.
5 years ago
c1f3b2d9d4
Added entrywise inversion and documentation for vec-vec operation
5 years ago
db7b882e9c
entrywise divide with zero check
5 years ago
cf3fce32c3
Update for new get_rank
5 years ago
ce6383b7ff
Added Element-by-Element divison
5 years ago
122c154bca
Merge branch 'merge-paraggr' into fnd_owner
5 years ago
e95063701b
Whitespace fixes
5 years ago
cf6db2ea0b
get_rank and get_mpicomm
...
Make psb_get_rank and psb_get_mpicomm functions, not
subroutines. Document same.
5 years ago
4ff817f4f1
Take out debug prints
5 years ago
f66d19e54b
added entrywise vector product psb_gemlt
5 years ago
374786f6ae
adjcncy_fnd_owner fix memory allocation
5 years ago
809e32a6be
adjcncy_fnd_owner with irecv
...
Implement adjcncy_fnd_owner with irecv to improve performance instead
of psb_snd/psb_rcv. To be fully debugged.
5 years ago
410c0c9995
graph_fnd_owner sweep
...
Implement initial sweep in graph_fnd_owner based on user-provided
topology.
Cleanup old indx_map implementation.
New test program just for testing cdasb.
5 years ago
79a25a80ed
Cosmetic changes to graph_fnd_owner
5 years ago
6258b9ff66
use desc_mod instead of indx_map_mod to pull interface of cd_get_maxspace
5 years ago
31ed911b1e
New set/get_maxspace()
5 years ago
e2bd101ded
Fix typos in source.
5 years ago
1da39b8bbf
Factored symmetrization of del_list into its own routine.
5 years ago
fc90423305
Make adj() argument allocatable so it can be adjusted for symmetry
5 years ago
6732106bc2
First version of adjncy_fnd_owner. To be fully tested.
5 years ago
6c31765b3f
First draft of psi_graph_fnd_owner. Need to revise choice of algorithmic thresholds.
5 years ago
95dffe2f76
Refactor fnd_owner into graph_fnd_owner which may call a2a_ and adjcncy_
5 years ago
4507a6522a
New indxmap component for process adjacency list, setters/getters.
5 years ago
ce115b1318
Put in a safety margin in reallocation of dep_list. To be reworked completely.
5 years ago
8ae408fa03
Improved memory allocation in extract_dep_list
5 years ago
3ae846edb5
Start refactoring of crea_index
5 years ago
65092a42b1
Split hash_impl from desc_impl.
5 years ago
2529225730
Changed fnd_owner internal names.
5 years ago
40204a7be6
Moved fnd_owner to internals impl. subdir
5 years ago
1d04fffbd4
Refactor map_fnd_owner, first step towards new version.
5 years ago
b3d6a970ee
Add adjncy component
5 years ago
4366ac4948
Move psb_indx_map_fnd_owner in modules/desc subdir, update makefiles.
5 years ago
b83ec80619
Fix scan_sum, define V version
5 years ago
d189083816
Check and enforce allocation of indxmap%halo_owner
5 years ago
c54076ed97
Kludge: defined indx_map%cpy method, use same in %clone. To be reviewed
5 years ago
62c21836f0
Fix REINIT use of array constants with implicit loops using IPK/LPK
5 years ago
768344d93f
Add version of CDINS with lpk NZ
5 years ago
c338e7b799
New LPK set_nzeros for COO
5 years ago
319d2831b7
New methods for LX%set_nrows with IPK/LPK
5 years ago
634c2e0aa5
Implement clip_diag for LX
5 years ago
ece66421bf
Add len arg in declaration.
5 years ago
a65626c7b0
Merge fnd_halo_owner method and setup
5 years ago
10b802743e
Change constant name to psb_c_Xpk_
5 years ago
71059dc783
Version of sphalo for LX_csr.
5 years ago
40755d6279
New method par_csr_spspmm.
5 years ago
22d76d1e48
Fix SYM handling at spmat level.
5 years ago
dcedab8ae0
New usehash argument in CDALL. Still need to put in a heuristics for
...
gen_block with too many cores
5 years ago
2f07fc5220
New scan_sum and exscan_sum collectives.
5 years ago
b97bdf3d2e
Added timers facility.
5 years ago
5a61fbd2c0
Duplicated matrix entries should be ADDED by default.
5 years ago
7e2ffbf34d
New handling of SYMmetry
5 years ago
239f25a913
Memory allocation for COO/CSC/CSR.
5 years ago
57d08dc4d7
Formatting source code fixes.
5 years ago
242b9ec91f
New variant of psb_spins in CSR mode. Document same.
5 years ago
9a6d9ac5c4
Aux space spspmm estimate was wrong, reverted.
5 years ago
5abe99196c
Fix calls to checkvect in data exchange functions, take 2.
5 years ago
113478ea64
Fix calls to checkvect in data exchange functions.
5 years ago
08063d4f48
Deallocate output if needed in indxmap%clone.
5 years ago
3bafaa242a
New checkvect interface, and use.
6 years ago
3ecdb7d026
Explicitly free output descriptor on desc%clone.
6 years ago
4789277b19
Fixed switch to build state for overlap descriptor.
6 years ago
51e50120bd
Add move_alloc method to sparse matrix.
6 years ago
fd8ce4f3de
New linmap internal structure.
6 years ago
300762ed29
Fix map build for gen_linear.
6 years ago
Salvatore Filippone
Salvatore Filippone
Cirdans-Home