You cannot select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
amg4psblas/mlprec/mld_dbjac_bld.f90

618 lines
19 KiB
Fortran

!!$
!!$
!!$ MLD2P4
!!$ MultiLevel Domain Decomposition Parallel Preconditioners Package
!!$ based on PSBLAS (Parallel Sparse BLAS v.2.0)
!!$
!!$ (C) Copyright 2007 Alfredo Buttari University of Rome Tor Vergata
!!$ Pasqua D'Ambra ICAR-CNR, Naples
!!$ Daniela di Serafino Second University of Naples
!!$ Salvatore Filippone University of Rome Tor Vergata
!!$
!!$ Redistribution and use in source and binary forms, with or without
!!$ modification, are permitted provided that the following conditions
!!$ are met:
!!$ 1. Redistributions of source code must retain the above copyright
!!$ notice, this list of conditions and the following disclaimer.
!!$ 2. Redistributions in binary form must reproduce the above copyright
!!$ notice, this list of conditions, and the following disclaimer in the
!!$ documentation and/or other materials provided with the distribution.
!!$ 3. The name of the MLD2P4 group or the names of its contributors may
!!$ not be used to endorse or promote products derived from this
!!$ software without specific written permission.
!!$
!!$ THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
!!$ ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
!!$ TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
!!$ PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE MLD2P4 GROUP OR ITS CONTRIBUTORS
!!$ BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
!!$ CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
!!$ SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
!!$ INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
!!$ CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
!!$ ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
!!$ POSSIBILITY OF SUCH DAMAGE.
!!$
!!$
! File: mld_dbjac_bld.f90.
!
! Subroutine: mld_dbjac_bld.
! Version: real.
!
! This routine builds the local extended matrix associated to an Additive
! Schwarz preconditioner and computes an LU or incomplete LU factorization
! of this matrix, according to the value of p%iprcparm(iprcparm(sub_solve_),
! set by the user through mld_dprecinit or mld_dprecset.
! Alternatively, it splits the local matrix into its block-diagonal and
! off block-diagonal parts, for the future application of multiple
! block-Jacobi sweeps.
!
! This routine is used by mld_dbaseprec_bld, to build a 'base' block-Jacobi or
! Additive Schwarz (AS) preconditioner at any level of a multilevel preconditioner,
! or a block-Jacobi or LU or ILU solver at the coarsest level of a multilevel
! preconditioner. More precisely, the routine is used to perform one of the
! following tasks:
!
! 1. construction of a block-Jacobi or Additive Schwarz preconditioner associated
! to a matrix A distributed among the processes (allowed at any level);
!
! 2. setup of block-Jacobi sweeps to compute an approximate solution of a
! linear system
! A*Y = X,
!
! distributed among the processes (allowed only at the coarsest level);
!
! 3. LU factorization of a linear system
!
! A*Y = X,
!
! distributed among the processes (allowed only at the coarsest level);
!
! 4. LU or incomplete LU factorization of a linear system
!
! A*Y = X,
!
! replicated on the processes (allowed only at the coarsest level).
!
! The following factorizations are available:
! - ILU(k), i.e. ILU factorization with fill-in level k;
! - MILU(k), i.e. modified ILU factorization with fill-in level k;
! - ILU(k,t), i.e. ILU with threshold (i.e. drop tolerance) t and k additional
! entries in each row of the L and U factors with respect to the initial
! sparsity pattern;
! - LU implemented in SuperLU version 3.0;
! - LU implemented in UMFPACK version 4.4;
! - distributed LU implemented in SuperLU_DIST version 2.0.
!
!
! Arguments:
! a - type(<psb_dspmat_type>), input.
! The sparse matrix structure containing the local part of the
! matrix to be preconditioned or factorized.
! desc_a - type(<psb_desc_type>), input.
! The communication descriptor associated to a.
! p - type(<mld_dbaseprec_type>), input/output.
! The 'base preconditioner' data structure containing the local
! part of the preconditioner or solver at the current level.
! info - integer, output.
! Error code.
!
subroutine mld_dbjac_bld(a,desc_a,p,upd,info)
use psb_base_mod
use mld_prec_mod, mld_protect_name => mld_dbjac_bld
implicit none
! Arguments
type(psb_dspmat_type), intent(in), target :: a
type(psb_desc_type), intent(in) :: desc_a
type(mld_dbaseprc_type), intent(inout) :: p
integer, intent(out) :: info
character, intent(in) :: upd
! Local Variables
integer :: i, k, m
integer :: int_err(5)
character :: trans, unitd
type(psb_dspmat_type) :: blck, atmp
logical, parameter :: debugprt=.false., debug=.false., aggr_dump=.false.
integer :: err_act, n_row, nrow_a,n_col
integer :: ictxt,np,me
character(len=20) :: name
character(len=5), parameter :: coofmt='COO', csrfmt='CSR'
if(psb_get_errstatus().ne.0) return
info=0
name='mld_dbjac_bld'
call psb_erractionsave(err_act)
ictxt=psb_cd_get_context(desc_a)
call psb_info(ictxt, me, np)
m = a%m
if (m < 0) then
info = 10
int_err(1) = 1
int_err(2) = m
call psb_errpush(info,name,i_err=int_err)
goto 9999
endif
trans = 'N'
unitd = 'U'
if (p%iprcparm(mld_n_ovr_) < 0) then
info = 11
int_err(1) = 1
int_err(2) = p%iprcparm(mld_n_ovr_)
call psb_errpush(info,name,i_err=int_err)
goto 9999
endif
call psb_nullify_sp(blck)
call psb_nullify_sp(atmp)
if(debug) write(0,*)me,': calling mld_asmat_bld',&
& p%iprcparm(mld_prec_type_),p%iprcparm(mld_n_ovr_)
if (debug) call psb_barrier(ictxt)
!
! Build the communication descriptor for the Additive Schwarz
! preconditioner and retrieves the remote pieces of the local
! extended matrix needed by that preconditioner. If the
! preconditioner is the block-Jacobi one, only a copy of the
! descriptor of the original matrix is made.
!
call mld_asmat_bld(p%iprcparm(mld_prec_type_),p%iprcparm(mld_n_ovr_),a,&
& blck,desc_a,upd,p%desc_data,info,outfmt=csrfmt)
if (debugprt) then
open(60+me)
call psb_csprt(60+me,a,head='% A')
close(60+me)
open(70+me)
call psb_csprt(70+me,blck,head='% BLCK')
close(70+me)
endif
if(info/=0) then
call psb_errpush(4010,name,a_err='mld_asmat_bld')
goto 9999
end if
if (debug) write(0,*)me,': out of mld_asmat_bld'
if (debug) call psb_barrier(ictxt)
!
! Treat separately the case the local matrix has to be reordered
! and the case this is not required.
!
select case(p%iprcparm(mld_sub_ren_))
!
! A reordering of the local matrix is required.
!
case (1:)
!
! Reorder the rows and the columns of the local extended matrix,
! according to the value of p%iprcparm(sub_ren_). The reordered
! matrix is stored into atmp, using the COO format.
!
call mld_sp_renum(a,desc_a,blck,p,atmp,info)
if (info/=0) then
call psb_errpush(4010,name,a_err='mld_sp_renum')
goto 9999
end if
!
! Clip into p%av(ap_nd_) the off block-diagonal part of the local
! matrix. The clipped matrix is then stored in CSR format.
!
call psb_sp_clip(atmp,p%av(mld_ap_nd_),info,&
& jmin=atmp%m+1,rscale=.false.,cscale=.false.)
call psb_spcnv(p%av(mld_ap_nd_),info,afmt='csr',dupl=psb_dupl_add_)
if(info /= 0) then
call psb_errpush(4010,name,a_err='psb_spcnv csr 1')
goto 9999
end if
k = psb_sp_get_nnzeros(p%av(mld_ap_nd_))
call psb_sum(ictxt,k)
if (k == 0) then
!
! If the off diagonal part is emtpy, there is no point in doing
! multiple Jacobi sweeps. This is certain to happen when running
! on a single processor.
!
p%iprcparm(mld_smooth_sweeps_) = 1
end if
if (debugprt) then
call psb_barrier(ictxt)
open(40+me)
call psb_csprt(40+me,atmp,head='% Local matrix')
close(40+me)
endif
if (debug) write(0,*) me,' Factoring rows ',&
&atmp%m,a%m,blck%m,atmp%ia2(atmp%m+1)-1
!
! Compute a factorization of the diagonal block of the local matrix,
! according to the choice made by the user by setting p%iprcparm(sub_solve_)
!
select case(p%iprcparm(mld_sub_solve_))
case(mld_ilu_n_,mld_milu_n_,mld_ilu_t_)
!
! ILU(k)/MILU(k)/ILU(k,t) factorization.
!
call psb_spcnv(atmp,info,afmt='csr',dupl=psb_dupl_add_)
if (info /= 0) then
call psb_errpush(4010,name,a_err='psb_spcnv csr 2')
goto 9999
end if
call mld_ilu_bld(atmp,p%desc_data,p,upd,info)
if (info/=0) then
call psb_errpush(4010,name,a_err='mld_ilu_bld')
goto 9999
end if
if (debugprt) then
open(80+me)
call psb_csprt(80+me,p%av(mld_l_pr_),head='% Local L factor')
write(80+me,*) '% Diagonal: ',p%av(mld_l_pr_)%m
do i=1,p%av(mld_l_pr_)%m
write(80+me,*) i,i,p%d(i)
enddo
call psb_csprt(80+me,p%av(mld_u_pr_),head='% Local U factor')
close(80+me)
endif
case(mld_slu_)
!
! LU factorization through the SuperLU package.
!
call psb_spcnv(atmp,info,afmt='csr',dupl=psb_dupl_add_)
if (info /= 0) then
call psb_errpush(4010,name,a_err='psb_spcnv csr 3')
goto 9999
end if
call mld_slu_bld(atmp,p%desc_data,p,info)
if(info /= 0) then
call psb_errpush(4010,name,a_err='mld_slu_bld')
goto 9999
end if
case(mld_umf_)
!
! LU factorization through the UMFPACK package.
!
call psb_spcnv(atmp,info,afmt='csc',dupl=psb_dupl_add_)
if (info /= 0) then
call psb_errpush(4010,name,a_err='psb_spcnv csc')
goto 9999
end if
call mld_umf_bld(atmp,p%desc_data,p,info)
if(debug) write(0,*)me,': Done mld_umf_bld ',info
if (info /= 0) then
call psb_errpush(4010,name,a_err='mld_umf_bld')
goto 9999
end if
case(mld_f_none_)
!
! Error: no factorization required.
!
info=4010
call psb_errpush(info,name,a_err='Inconsistent prec mld_f_none_')
goto 9999
case default
info=4010
call psb_errpush(info,name,a_err='Unknown mld_sub_solve_')
goto 9999
end select
call psb_sp_free(atmp,info)
if(info/=0) then
call psb_errpush(4010,name,a_err='psb_sp_free')
goto 9999
end if
!
! No reordering of the local matrix is required
!
case(0)
!
! Compute a factorization of the diagonal block of the local matrix,
! according to the choice made by the user by setting p%iprcparm(sub_solve_)
!
select case(p%iprcparm(mld_sub_solve_))
case(mld_ilu_n_,mld_milu_n_,mld_ilu_t_)
!
! ILU(k)/MILU(k)/ILU(k,t) factorization.
!
!
! In case of multiple block-Jacobi sweeps, clip into p%av(ap_nd_)
! the off block-diagonal part of the local extended matrix. The
! clipped matrix is then stored in CSR format.
!
if (p%iprcparm(mld_smooth_sweeps_) > 1) then
n_row = psb_cd_get_local_rows(p%desc_data)
n_col = psb_cd_get_local_cols(p%desc_data)
nrow_a = a%m
! The following is known to work
! given that the output from CLIP is in COO.
call psb_sp_clip(a,p%av(mld_ap_nd_),info,&
& jmin=nrow_a+1,rscale=.false.,cscale=.false.)
call psb_sp_clip(blck,atmp,info,&
& jmin=nrow_a+1,rscale=.false.,cscale=.false.)
call psb_rwextd(n_row,p%av(mld_ap_nd_),info,b=atmp)
call psb_spcnv(p%av(mld_ap_nd_),info,afmt='csr',dupl=psb_dupl_add_)
if(info /= 0) then
call psb_errpush(4010,name,a_err='psb_spcnv csr 4')
goto 9999
end if
k = psb_sp_get_nnzeros(p%av(mld_ap_nd_))
call psb_sum(ictxt,k)
if (k == 0) then
!
! If the off block-diagonal part is emtpy, there is no point in doing
! multiple Jacobi sweeps. This is certain to happen when running
! on a single processor.
!
p%iprcparm(mld_smooth_sweeps_) = 1
end if
call psb_sp_free(atmp,info)
end if
!
! Compute the incomplete LU factorization.
!
call mld_ilu_bld(a,desc_a,p,upd,info,blck=blck)
if(info/=0) then
call psb_errpush(4010,name,a_err='mld_ilu_bld')
goto 9999
end if
if (debugprt) then
open(80+me)
call psb_csprt(80+me,p%av(mld_l_pr_),head='% Local L factor')
write(80+me,*) '% Diagonal: ',p%av(mld_l_pr_)%m
do i=1,p%av(mld_l_pr_)%m
write(80+me,*) i,i,p%d(i)
enddo
call psb_csprt(80+me,p%av(mld_u_pr_),head='% Local U factor')
close(80+me)
endif
case(mld_slu_)
!
! LU factorization through the SuperLU package.
!
call psb_spcnv(a,atmp,info,afmt='coo')
if (info /= 0) then
call psb_errpush(4010,name,a_err='psb_spcnv')
goto 9999
end if
n_row = psb_cd_get_local_rows(p%desc_data)
n_col = psb_cd_get_local_cols(p%desc_data)
call psb_rwextd(n_row,atmp,info,b=blck)
!
! In case of multiple block-Jacobi sweeps, clip into p%av(ap_nd_)
! the off block-diagonal part of the local extended matrix. The
! clipped matrix is then stored in CSR format.
!
if (p%iprcparm(mld_smooth_sweeps_) > 1) then
call psb_sp_clip(atmp,p%av(mld_ap_nd_),info,&
& jmin=atmp%m+1,rscale=.false.,cscale=.false.)
call psb_spcnv(p%av(mld_ap_nd_),info,afmt='csr',dupl=psb_dupl_add_)
if(info /= 0) then
call psb_errpush(4010,name,a_err='psb_spcnv csr 6')
goto 9999
end if
k = psb_sp_get_nnzeros(p%av(mld_ap_nd_))
call psb_sum(ictxt,k)
if (k == 0) then
!
! If the off block-diagonal part is emtpy, there is no point in doing
! multiple Jacobi sweeps. This is certain to happen when running
! on a single processor.
!
p%iprcparm(mld_smooth_sweeps_) = 1
end if
endif
!
! Compute the LU factorization.
!
if (info == 0) call psb_spcnv(atmp,info,afmt='csr',dupl=psb_dupl_add_)
if (info == 0) call mld_slu_bld(atmp,p%desc_data,p,info)
if(info /= 0) then
call psb_errpush(4010,name,a_err='mld_slu_bld')
goto 9999
end if
call psb_sp_free(atmp,info)
if(info/=0) then
call psb_errpush(4010,name,a_err='psb_sp_free')
goto 9999
end if
case(mld_sludist_)
!
! LU factorization through the SuperLU_DIST package. This works only
! when the matrix is distributed among the processes.
! NOTE: Should have NO overlap here!!!!
!
call psb_spcnv(a,atmp,info,afmt='csr')
if (info /= 0) then
call psb_errpush(4010,name,a_err='psb_spcnv')
goto 9999
end if
n_row = psb_cd_get_local_rows(p%desc_data)
n_col = psb_cd_get_local_cols(p%desc_data)
if (info == 0) call mld_sludist_bld(atmp,p%desc_data,p,info)
if(info /= 0) then
call psb_errpush(4010,name,a_err='mld_slu_bld')
goto 9999
end if
call psb_sp_free(atmp,info)
if(info/=0) then
call psb_errpush(4010,name,a_err='psb_sp_free')
goto 9999
end if
case(mld_umf_)
!
! LU factorization through the UMFPACK package.
!
call psb_spcnv(a,atmp,info,afmt='coo')
if (info /= 0) then
call psb_errpush(4010,name,a_err='psb_spcnv')
goto 9999
end if
n_row = psb_cd_get_local_rows(p%desc_data)
n_col = psb_cd_get_local_cols(p%desc_data)
call psb_rwextd(n_row,atmp,info,b=blck)
!
! In case of multiple block-Jacobi sweeps, clip into p%av(ap_nd_)
! the off block-diagonal part of the local extended matrix. The
! clipped matrix is then stored in CSR format.
!
if (p%iprcparm(mld_smooth_sweeps_) > 1) then
call psb_sp_clip(atmp,p%av(mld_ap_nd_),info,&
& jmin=atmp%m+1,rscale=.false.,cscale=.false.)
call psb_spcnv(p%av(mld_ap_nd_),info,afmt='csr',dupl=psb_dupl_add_)
if(info /= 0) then
call psb_errpush(4010,name,a_err='psb_spcnv csr 8')
goto 9999
end if
k = psb_sp_get_nnzeros(p%av(mld_ap_nd_))
call psb_sum(ictxt,k)
if (k == 0) then
!
! If the off block-diagonal part is emtpy, there is no point in doing
! multiple Jacobi sweeps. This is certain to happen when running
! on a single processor.
!
p%iprcparm(mld_smooth_sweeps_) = 1
end if
endif
!
! Compute the LU factorization.
!
if (info == 0) call psb_ipcoo2csc(atmp,info,clshr=.true.)
if (info /= 0) then
call psb_errpush(4010,name,a_err='psb_ipcoo2csc')
goto 9999
end if
call mld_umf_bld(atmp,p%desc_data,p,info)
if(debug) write(0,*)me,': Done mld_umf_bld ',info
if (info /= 0) then
call psb_errpush(4010,name,a_err='mld_umf_bld')
goto 9999
end if
call psb_sp_free(atmp,info)
if(info/=0) then
call psb_errpush(4010,name,a_err='psb_sp_free')
goto 9999
end if
case(mld_f_none_)
!
! Error: no factorization required.
!
info=4010
call psb_errpush(info,name,a_err='Inconsistent prec mld_f_none_')
goto 9999
case default
info=4010
call psb_errpush(info,name,a_err='Unknown mld_sub_solve_')
goto 9999
end select
case default
info=4010
call psb_errpush(info,name,a_err='Invalid renum_')
goto 9999
end select
call psb_sp_free(blck,info)
if(info/=0) then
call psb_errpush(4010,name,a_err='psb_sp_free')
goto 9999
end if
if (debug) write(0,*) me,'End of ilu_bld'
call psb_erractionrestore(err_act)
return
9999 continue
call psb_erractionrestore(err_act)
if (err_act.eq.psb_act_abort_) then
call psb_error()
return
end if
return
end subroutine mld_dbjac_bld