!!$ !!$ !!$ MLD2P4 version 1.0 !!$ MultiLevel Domain Decomposition Parallel Preconditioners Package !!$ based on PSBLAS (Parallel Sparse BLAS version 2.2) !!$ !!$ (C) Copyright 2008 !!$ !!$ Salvatore Filippone University of Rome Tor Vergata !!$ Alfredo Buttari University of Rome Tor Vergata !!$ Pasqua D'Ambra ICAR-CNR, Naples !!$ Daniela di Serafino Second University of Naples !!$ !!$ 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_smlprec_aply.f90 ! ! Subroutine: mld_smlprec_aply ! Version: real ! ! This routine computes ! ! Y = beta*Y + alpha*op(M^(-1))*X, ! where ! - M is a multilevel domain decomposition (Schwarz) preconditioner associated ! to a certain matrix A and stored in the array precv, ! - op(M^(-1)) is M^(-1) or its transpose, according to the value of trans, ! - X and Y are vectors, ! - alpha and beta are scalars. ! ! For each level we have as many submatrices as processes (except for the coarsest ! level where we might have a replicated index space) and each process takes care ! of one submatrix. ! ! The multilevel preconditioner M is regarded as an array of 'one-level preconditioners', ! each representing the part of the preconditioner associated to a certain level. ! For each level ilev, the preconditioner K(ilev) is stored in precv(ilev) ! and is associated to a matrix A(ilev), obtained by 'tranferring' the original ! matrix A (i.e. the matrix to be preconditioned) to the level ilev, through smoothed ! aggregation. ! ! The levels are numbered in increasing order starting from the finest one, i.e. ! level 1 is the finest level and A(1) is the matrix A. ! ! For a general description of (parallel) multilevel preconditioners see ! - B.F. Smith, P.E. Bjorstad & W.D. Gropp, ! Domain decomposition: parallel multilevel methods for elliptic partial ! differential equations, ! Cambridge University Press, 1996. ! - K. Stuben, ! Algebraic Multigrid (AMG): An Introduction with Applications, ! GMD Report N. 70, 1999. ! ! ! Arguments: ! alpha - real(psb_spk_), input. ! The scalar alpha. ! precv - type(mld_s_onelev_prec_type), dimension(:), input. ! The array of one-level preconditioner data structures containing the ! local parts of the preconditioners to be applied at each level. ! Note that nlev = size(precv) = number of levels. ! precv(ilev)%prec - type(psb_sbaseprc_type) ! The "base" preconditioner for the current level ! precv(ilev)%ac - type(psb_sspmat_type) ! The local part of the matrix A(ilev). ! precv(ilev)%desc_ac - type(psb_desc_type). ! The communication descriptor associated to the sparse ! matrix A(ilev) ! precv(ilev)%map_desc - type(psb_inter_desc_type) ! Stores the linear operators mapping between levels ! (ilev-1) and (ilev). These are the restriction and ! prolongation operators described in the sequel. ! precv(ilev)%iprcparm - integer, dimension(:), allocatable. ! The integer parameters defining the multilevel ! strategy ! precv(ilev)%rprcparm - real(psb_spk_), dimension(:), allocatable. ! The real parameters defining the multilevel strategy ! precv(ilev)%mlia - integer, dimension(:), allocatable. ! The aggregation map (ilev-1) --> (ilev). ! In case of non-smoothed aggregation, it is used ! instead of mld_sm_pr_. ! precv(ilev)%nlaggr - integer, dimension(:), allocatable. ! The number of aggregates (rows of A(ilev)) on the ! various processes. ! precv(ilev)%base_a - type(psb_sspmat_type), pointer. ! Pointer (really a pointer!) to the base matrix of ! the current level, i.e. the local part of A(ilev); ! so we have a unified treatment of residuals. We ! need this to avoid passing explicitly the matrix ! A(ilev) to the routine which applies the ! preconditioner. ! precv(ilev)%base_desc - type(psb_desc_type), pointer. ! Pointer to the communication descriptor associated ! to the sparse matrix pointed by base_a. ! ! x - real(psb_spk_), dimension(:), input. ! The local part of the vector X. ! beta - real(psb_spk_), input. ! The scalar beta. ! y - real(psb_spk_), dimension(:), input/output. ! The local part of the vector Y. ! desc_data - type(psb_desc_type), input. ! The communication descriptor associated to the matrix to be ! preconditioned. ! trans - character, optional. ! If trans='N','n' then op(M^(-1)) = M^(-1); ! if trans='T','t' then op(M^(-1)) = M^(-T) (transpose of M^(-1)). ! work - real(psb_spk_), dimension (:), optional, target. ! Workspace. Its size must be at least 4*psb_cd_get_local_cols(desc_data). ! info - integer, output. ! Error code. ! ! Note that when the LU factorization of the matrix A(ilev) is computed instead of ! the ILU one, by using UMFPACK or SuperLU, the corresponding L and U factors ! are stored in data structures provided by UMFPACK or SuperLU and pointed by ! precv(ilev)%prec%iprcparm(mld_umf_ptr) or precv(ilev)%prec%iprcparm(mld_slu_ptr), ! respectively. ! subroutine mld_smlprec_aply(alpha,precv,x,beta,y,desc_data,trans,work,info) use psb_base_mod use mld_inner_mod, mld_protect_name => mld_smlprec_aply implicit none ! Arguments type(psb_desc_type),intent(in) :: desc_data type(mld_s_onelev_prec_type), intent(in) :: precv(:) real(psb_spk_),intent(in) :: alpha,beta real(psb_spk_),intent(in) :: x(:) real(psb_spk_),intent(inout) :: y(:) character, intent(in) :: trans real(psb_spk_),target :: work(:) integer, intent(out) :: info ! Local variables integer :: ictxt, np, me, err_act integer :: debug_level, debug_unit character(len=20) :: name character :: trans_ name='mld_smlprec_aply' info = 0 call psb_erractionsave(err_act) debug_unit = psb_get_debug_unit() debug_level = psb_get_debug_level() ictxt = psb_cd_get_context(desc_data) call psb_info(ictxt, me, np) if (debug_level >= psb_debug_inner_) & & write(debug_unit,*) me,' ',trim(name),& & ' Entry ', size(precv) trans_ = psb_toupper(trans) select case(precv(2)%iprcparm(mld_ml_type_)) case(mld_no_ml_) ! ! No preconditioning, should not really get here ! call psb_errpush(4001,name,a_err='mld_no_ml_ in mlprc_aply?') goto 9999 case(mld_add_ml_) ! ! Additive multilevel ! call add_ml_aply(alpha,precv,x,beta,y,desc_data,trans_,work,info) case(mld_mult_ml_) ! ! Multiplicative multilevel (multiplicative among the levels, additive inside ! each level) ! ! Pre/post-smoothing versions. ! Note that the transpose switches pre <-> post. ! select case(precv(2)%iprcparm(mld_smoother_pos_)) case(mld_post_smooth_) select case (trans_) case('N') call mlt_post_ml_aply(alpha,precv,x,beta,y,desc_data,trans_,work,info) case('T','C') call mlt_pre_ml_aply(alpha,precv,x,beta,y,desc_data,trans_,work,info) case default info = 4001 call psb_errpush(info,name,a_err='invalid trans') goto 9999 end select case(mld_pre_smooth_) select case (trans_) case('N') call mlt_pre_ml_aply(alpha,precv,x,beta,y,desc_data,trans_,work,info) case('T','C') call mlt_post_ml_aply(alpha,precv,x,beta,y,desc_data,trans_,work,info) case default info = 4001 call psb_errpush(info,name,a_err='invalid trans') goto 9999 end select case(mld_twoside_smooth_) call mlt_twoside_ml_aply(alpha,precv,x,beta,y,desc_data,trans_,work,info) case default info = 4013 call psb_errpush(info,name,a_err='invalid smooth_pos',& & i_Err=(/precv(2)%iprcparm(mld_smoother_pos_),0,0,0,0/)) goto 9999 end select case default info = 4013 call psb_errpush(info,name,a_err='invalid mltype',& & i_Err=(/precv(2)%iprcparm(mld_ml_type_),0,0,0,0/)) goto 9999 end select 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 contains ! ! Subroutine: add_ml_aply ! Version: real ! Note: internal subroutine of mld_smlprec_aply. ! ! This routine computes ! ! Y = beta*Y + alpha*op(M^(-1))*X, ! where ! - M is an additive multilevel domain decomposition (Schwarz) preconditioner ! associated to a certain matrix A and stored in the array precv, ! - op(M^(-1)) is M^(-1) or its transpose, according to the value of trans, ! - X and Y are vectors, ! - alpha and beta are scalars. ! ! The preconditioner M is additive both through the levels and inside each ! level. ! ! For each level we have as many submatrices as processes (except for the coarsest ! level where we might have a replicated index space) and each process takes care ! of one submatrix. ! ! The multilevel preconditioner M is regarded as an array of 'one-level preconditioners', ! each representing the part of the preconditioner associated to a certain level. ! For each level ilev, the base preconditioner K(ilev) is stored in precv(ilev) ! and is associated to a matrix A(ilev), obtained by 'tranferring' the original ! matrix A (i.e. the matrix to be preconditioned) to the level ilev, through smoothed ! aggregation. ! ! The levels are numbered in increasing order starting from the finest one, i.e. ! level 1 is the finest level and A(1) is the matrix A. ! ! For details on the additive multilevel Schwarz preconditioner see the ! Algorithm 3.1.1 in the book: ! B.F. Smith, P.E. Bjorstad & W.D. Gropp, ! Domain decomposition: parallel multilevel methods for elliptic partial ! differential equations, Cambridge University Press, 1996. ! ! For a description of the arguments see mld_smlprec_aply. ! ! A sketch of the algorithm implemented in this routine is provided below ! (AV(ilev; sm_pr_) denotes the smoothed prolongator from level ilev to ! level ilev-1, while AV(ilev; sm_pr_t_) denotes its transpose, i.e. the ! corresponding restriction operator from level ilev-1 to level ilev). ! ! 1. ! Apply the base preconditioner at level 1. ! ! The sum over the subdomains is carried out in the ! ! application of K(1). ! X(1) = Xest ! Y(1) = (K(1)^(-1))*X(1) ! ! 2. DO ilev=2,nlev ! ! ! Transfer X(ilev-1) to the next coarser level. ! X(ilev) = AV(ilev; sm_pr_t_)*X(ilev-1) ! ! ! Apply the base preconditioner at the current level. ! ! The sum over the subdomains is carried out in the ! ! application of K(ilev). ! Y(ilev) = (K(ilev)^(-1))*X(ilev) ! ! ENDDO ! ! 3. DO ilev=nlev-1,1,-1 ! ! ! Transfer Y(ilev+1) to the next finer level. ! Y(ilev) = AV(ilev+1; sm_pr_)*Y(ilev+1) ! ! ENDDO ! ! 4. Yext = beta*Yext + alpha*Y(1) ! subroutine add_ml_aply(alpha,precv,x,beta,y,desc_data,trans,work,info) implicit none ! Arguments type(psb_desc_type),intent(in) :: desc_data type(mld_s_onelev_prec_type), intent(in) :: precv(:) real(psb_spk_),intent(in) :: alpha,beta real(psb_spk_),intent(in) :: x(:) real(psb_spk_),intent(inout) :: y(:) character, intent(in) :: trans real(psb_spk_),target :: work(:) integer, intent(out) :: info ! Local variables integer :: ictxt,np,me,i, nr2l,nc2l,err_act integer :: debug_level, debug_unit integer :: nlev, ilev character(len=20) :: name type psb_mlprec_wrk_type real(psb_spk_), allocatable :: tx(:), ty(:), x2l(:), y2l(:) end type psb_mlprec_wrk_type type(psb_mlprec_wrk_type), allocatable :: mlprec_wrk(:) name='add_ml_aply' info = 0 call psb_erractionsave(err_act) debug_unit = psb_get_debug_unit() debug_level = psb_get_debug_level() ictxt = psb_cd_get_context(desc_data) call psb_info(ictxt, me, np) if (debug_level >= psb_debug_inner_) & & write(debug_unit,*) me,' ',trim(name),& & ' Entry ', size(precv) nlev = size(precv) allocate(mlprec_wrk(nlev),stat=info) if (info /= 0) then call psb_errpush(4010,name,a_err='Allocate') goto 9999 end if ! ! STEP 1 ! ! Apply the base preconditioner at the finest level ! allocate(mlprec_wrk(1)%x2l(size(x)),mlprec_wrk(1)%y2l(size(y)), stat=info) if (info /= 0) then info=4025 call psb_errpush(info,name,i_err=(/size(x)+size(y),0,0,0,0/),& & a_err='real(psb_spk_)') goto 9999 end if mlprec_wrk(1)%x2l(:) = x(:) mlprec_wrk(1)%y2l(:) = szero call mld_baseprec_aply(alpha,precv(1)%prec,x,beta,y,& & precv(1)%base_desc,trans,work,info) if (info /=0) then call psb_errpush(4010,name,a_err='baseprec_aply') goto 9999 end if ! ! STEP 2 ! ! For each level except the finest one ... ! do ilev = 2, nlev nc2l = psb_cd_get_local_cols(precv(ilev)%base_desc) nr2l = psb_cd_get_local_rows(precv(ilev)%base_desc) allocate(mlprec_wrk(ilev)%x2l(nc2l),mlprec_wrk(ilev)%y2l(nc2l),& & stat=info) if (info /= 0) then info=4025 call psb_errpush(info,name,i_err=(/2*nc2l,0,0,0,0/),& & a_err='real(psb_spk_)') goto 9999 end if ! Apply prolongator transpose, i.e. restriction call psb_forward_map(sone,mlprec_wrk(ilev-1)%x2l,& & szero,mlprec_wrk(ilev)%x2l,& & precv(ilev)%map_desc,info,work=work) if (info /=0) then call psb_errpush(4001,name,a_err='Error during restriction') goto 9999 end if ! ! Apply the base preconditioner ! call mld_baseprec_aply(sone,precv(ilev)%prec,& & mlprec_wrk(ilev)%x2l,szero,mlprec_wrk(ilev)%y2l,& & precv(ilev)%base_desc, trans,work,info) enddo ! ! STEP 3 ! ! For each level except the finest one ... ! do ilev =nlev,2,-1 nc2l = psb_cd_get_local_cols(precv(ilev)%base_desc) nr2l = psb_cd_get_local_rows(precv(ilev)%base_desc) ! ! Apply prolongator ! call psb_backward_map(sone,mlprec_wrk(ilev)%y2l,& & sone,mlprec_wrk(ilev-1)%y2l,& & precv(ilev)%map_desc,info,work=work) if (info /=0) then call psb_errpush(4001,name,a_err='Error during prolongation') goto 9999 end if end do ! ! STEP 4 ! ! Compute the output vector Y ! call psb_geaxpby(alpha,mlprec_wrk(1)%y2l,sone,y,precv(1)%base_desc,info) if (info /= 0) then call psb_errpush(4001,name,a_err='Error on final update') goto 9999 end if deallocate(mlprec_wrk,stat=info) if (info /= 0) then call psb_errpush(4000,name) goto 9999 end if 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 add_ml_aply ! ! Subroutine: mlt_pre_ml_aply ! Version: real ! Note: internal subroutine of mld_smlprec_aply. ! ! This routine computes ! ! Y = beta*Y + alpha*op(M^(-1))*X, ! where ! - M is a hybrid multilevel domain decomposition (Schwarz) preconditioner ! associated to a certain matrix A and stored in the array precv, ! - op(M^(-1)) is M^(-1) or its transpose, according to the value of trans, ! - X and Y are vectors, ! - alpha and beta are scalars. ! ! The preconditioner M is hybrid in the sense that it is multiplicative through the ! levels and additive inside a level; pre-smoothing only is applied at each level. ! ! For each level we have as many submatrices as processes (except for the coarsest ! level where we might have a replicated index space) and each process takes care ! of one submatrix. ! ! The multilevel preconditioner M is regarded as an array of 'one-level preconditioners', ! each representing the part of the preconditioner associated to a certain level. ! For each level ilev, the base preconditioner K(ilev) is stored in precv(ilev) ! and is associated to a matrix A(ilev), obtained by 'tranferring' the original ! matrix A (i.e. the matrix to be preconditioned) to the level ilev, through smoothed ! aggregation. ! ! The levels are numbered in increasing order starting from the finest one, i.e. ! level 1 is the finest level and A(1) is the matrix A. ! ! For details on the pre-smoothed hybrid multiplicative multilevel Schwarz ! preconditioner, see the Algorithm 3.2.1 in the book: ! B.F. Smith, P.E. Bjorstad & W.D. Gropp, ! Domain decomposition: parallel multilevel methods for elliptic partial ! differential equations, Cambridge University Press, 1996. ! ! For a description of the arguments see mld_smlprec_aply. ! ! A sketch of the algorithm implemented in this routine is provided below ! (AV(ilev; sm_pr_) denotes the smoothed prolongator from level ilev to ! level ilev-1, while AV(ilev; sm_pr_t_) denotes its transpose, i.e. the ! corresponding restriction operator from level ilev-1 to level ilev). ! ! 1. X(1) = Xext ! ! 2. ! Apply the base preconditioner at the finest level. ! Y(1) = (K(1)^(-1))*X(1) ! ! 3. ! Compute the residual at the finest level. ! TX(1) = X(1) - A(1)*Y(1) ! ! 4. DO ilev=2, nlev ! ! ! Transfer the residual to the current (coarser) level. ! X(ilev) = AV(ilev; sm_pr_t_)*TX(ilev-1) ! ! ! Apply the base preconditioner at the current level. ! ! The sum over the subdomains is carried out in the ! ! application of K(ilev). ! Y(ilev) = (K(ilev)^(-1))*X(ilev) ! ! ! Compute the residual at the current level (except at ! ! the coarsest level). ! IF (ilev < nlev) ! TX(ilev) = (X(ilev)-A(ilev)*Y(ilev)) ! ! ENDDO ! ! 5. DO ilev=nlev-1,1,-1 ! ! ! Transfer Y(ilev+1) to the next finer level ! Y(ilev) = Y(ilev) + AV(ilev+1; sm_pr_)*Y(ilev+1) ! ! ENDDO ! ! 6. Yext = beta*Yext + alpha*Y(1) ! ! subroutine mlt_pre_ml_aply(alpha,precv,x,beta,y,desc_data,trans,work,info) implicit none ! Arguments type(psb_desc_type),intent(in) :: desc_data type(mld_s_onelev_prec_type), intent(in) :: precv(:) real(psb_spk_),intent(in) :: alpha,beta real(psb_spk_),intent(in) :: x(:) real(psb_spk_),intent(inout) :: y(:) character, intent(in) :: trans real(psb_spk_),target :: work(:) integer, intent(out) :: info ! Local variables integer :: ictxt,np,me,i, nr2l,nc2l,err_act integer :: debug_level, debug_unit integer :: nlev, ilev character(len=20) :: name type psb_mlprec_wrk_type real(psb_spk_), allocatable :: tx(:), ty(:), x2l(:), y2l(:) end type psb_mlprec_wrk_type type(psb_mlprec_wrk_type), allocatable :: mlprec_wrk(:) name='mlt_pre_ml_aply' info = 0 call psb_erractionsave(err_act) debug_unit = psb_get_debug_unit() debug_level = psb_get_debug_level() ictxt = psb_cd_get_context(desc_data) call psb_info(ictxt, me, np) if (debug_level >= psb_debug_inner_) & & write(debug_unit,*) me,' ',trim(name),& & ' Entry ', size(precv) nlev = size(precv) allocate(mlprec_wrk(nlev),stat=info) if (info /= 0) then call psb_errpush(4010,name,a_err='Allocate') goto 9999 end if ! ! STEP 1 ! ! Copy the input vector X ! nc2l = psb_cd_get_local_cols(precv(1)%base_desc) allocate(mlprec_wrk(1)%x2l(nc2l),mlprec_wrk(1)%y2l(nc2l), & & mlprec_wrk(1)%tx(nc2l), stat=info) if (info /= 0) then info=4025 call psb_errpush(info,name,i_err=(/4*nc2l,0,0,0,0/),& & a_err='real(psb_spk_)') goto 9999 end if mlprec_wrk(1)%x2l(:) = x ! ! STEP 2 ! ! Apply the base preconditioner at the finest level ! call mld_baseprec_aply(sone,precv(1)%prec,mlprec_wrk(1)%x2l,& & szero,mlprec_wrk(1)%y2l,precv(1)%base_desc,& & trans,work,info) if (info /=0) then call psb_errpush(4010,name,a_err=' baseprec_aply') goto 9999 end if ! ! STEP 3 ! ! Compute the residual at the finest level ! mlprec_wrk(1)%tx = mlprec_wrk(1)%x2l call psb_spmm(-sone,precv(1)%base_a,mlprec_wrk(1)%y2l,& & sone,mlprec_wrk(1)%tx,precv(1)%base_desc,info,& & work=work,trans=trans) if (info /=0) then call psb_errpush(4001,name,a_err=' fine level residual') goto 9999 end if ! ! STEP 4 ! ! For each level but the finest one ... ! do ilev = 2, nlev nc2l = psb_cd_get_local_cols(precv(ilev)%base_desc) nr2l = psb_cd_get_local_rows(precv(ilev)%base_desc) allocate(mlprec_wrk(ilev)%tx(nc2l),mlprec_wrk(ilev)%y2l(nc2l),& & mlprec_wrk(ilev)%x2l(nc2l), stat=info) if (info /= 0) then info=4025 call psb_errpush(info,name,i_err=(/4*nc2l,0,0,0,0/),& & a_err='real(psb_spk_)') goto 9999 end if ! Apply prolongator transpose, i.e. restriction call psb_forward_map(sone,mlprec_wrk(ilev-1)%tx,& & szero,mlprec_wrk(ilev)%x2l,& & precv(ilev)%map_desc,info,work=work) if (info /=0) then call psb_errpush(4001,name,a_err='Error during restriction') goto 9999 end if ! ! Apply the base preconditioner ! call mld_baseprec_aply(sone,precv(ilev)%prec,mlprec_wrk(ilev)%x2l,& & szero,mlprec_wrk(ilev)%y2l,precv(ilev)%base_desc,trans,work,info) ! ! Compute the residual (at all levels but the coarsest one) ! if (ilev < nlev) then mlprec_wrk(ilev)%tx = mlprec_wrk(ilev)%x2l if (info == 0) call psb_spmm(-sone,precv(ilev)%base_a,& & mlprec_wrk(ilev)%y2l,sone,mlprec_wrk(ilev)%tx,& & precv(ilev)%base_desc,info,work=work,trans=trans) endif if (info /=0) then call psb_errpush(4001,name,a_err='Error on up sweep residual') goto 9999 end if enddo ! ! STEP 5 ! ! For each level but the coarsest one ... ! do ilev = nlev-1, 1, -1 ! ! Apply prolongator ! call psb_backward_map(sone,mlprec_wrk(ilev+1)%y2l,& & sone,mlprec_wrk(ilev)%y2l,& & precv(ilev+1)%map_desc,info,work=work) if (info /=0) then call psb_errpush(4001,name,a_err='Error during prolongation') goto 9999 end if enddo ! ! STEP 6 ! ! Compute the output vector Y ! call psb_geaxpby(alpha,mlprec_wrk(1)%y2l,beta,y,& & precv(1)%base_desc,info) if (info /=0) then call psb_errpush(4001,name,a_err='Error on final update') goto 9999 end if deallocate(mlprec_wrk,stat=info) if (info /= 0) then call psb_errpush(4000,name) goto 9999 end if 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 mlt_pre_ml_aply ! ! Subroutine: mlt_post_ml_aply ! Version: real ! Note: internal subroutine of mld_smlprec_aply. ! ! This routine computes ! ! Y = beta*Y + alpha*op(M^(-1))*X, ! where ! - M is a hybrid multilevel domain decomposition (Schwarz) preconditioner ! associated to a certain matrix A and stored in the array precv, ! - op(M^(-1)) is M^(-1) or its transpose, according to the value of trans, ! - X and Y are vectors, ! - alpha and beta are scalars. ! ! The preconditioner M is hybrid in the sense that it is multiplicative through the ! levels and additive inside a level; post-smoothing only is applied at each level. ! ! For each level we have as many submatrices as processes (except for the coarsest ! level where we might have a replicated index space) and each process takes care ! of one submatrix. ! ! The multilevel preconditioner M is regarded as an array of 'one-level preconditioners', ! each representing the part of the preconditioner associated to a certain level. ! For each level ilev, the base preconditioner K(ilev) is stored in precv(ilev) ! and is associated to a matrix A(ilev), obtained by 'tranferring' the original ! matrix A (i.e. the matrix to be preconditioned) to the level ilev, through smoothed ! aggregation. ! ! The levels are numbered in increasing order starting from the finest one, i.e. ! level 1 is the finest level and A(1) is the matrix A. ! ! For details on hybrid multiplicative multilevel Schwarz preconditioners, see ! B.F. Smith, P.E. Bjorstad & W.D. Gropp, ! Domain decomposition: parallel multilevel methods for elliptic partial ! differential equations, Cambridge University Press, 1996. ! ! For a description of the arguments see mld_smlprec_aply. ! ! A sketch of the algorithm implemented in this routine is provided below. ! (AV(ilev; sm_pr_) denotes the smoothed prolongator from level ilev to ! level ilev-1, while AV(ilev; sm_pr_t_) denotes its transpose, i.e. the ! corresponding restriction operator from level ilev-1 to level ilev). ! ! 1. X(1) = Xext ! ! 2. DO ilev=2, nlev ! ! ! Transfer X(ilev-1) to the next coarser level. ! X(ilev) = AV(ilev; sm_pr_t_)*X(ilev-1) ! ! ENDDO ! ! 3.! Apply the preconditioner at the coarsest level. ! Y(nlev) = (K(nlev)^(-1))*X(nlev) ! ! 4. DO ilev=nlev-1,1,-1 ! ! ! Transfer Y(ilev+1) to the next finer level. ! Y(ilev) = AV(ilev+1; sm_pr_)*Y(ilev+1) ! ! ! Compute the residual at the current level and apply to it the ! ! base preconditioner. The sum over the subdomains is carried out ! ! in the application of K(ilev). ! Y(ilev) = Y(ilev) + (K(ilev)^(-1))*(X(ilev)-A(ilev)*Y(ilev)) ! ! ENDDO ! ! 5. Yext = beta*Yext + alpha*Y(1) ! ! subroutine mlt_post_ml_aply(alpha,precv,x,beta,y,desc_data,trans,work,info) implicit none ! Arguments type(psb_desc_type),intent(in) :: desc_data type(mld_s_onelev_prec_type), intent(in) :: precv(:) real(psb_spk_),intent(in) :: alpha,beta real(psb_spk_),intent(in) :: x(:) real(psb_spk_),intent(inout) :: y(:) character, intent(in) :: trans real(psb_spk_),target :: work(:) integer, intent(out) :: info ! Local variables integer :: ictxt,np,me,i, nr2l,nc2l,err_act integer :: debug_level, debug_unit integer :: nlev, ilev character(len=20) :: name type psb_mlprec_wrk_type real(psb_spk_), allocatable :: tx(:), ty(:), x2l(:), y2l(:) end type psb_mlprec_wrk_type type(psb_mlprec_wrk_type), allocatable :: mlprec_wrk(:) name='mlt_post_ml_aply' info = 0 call psb_erractionsave(err_act) debug_unit = psb_get_debug_unit() debug_level = psb_get_debug_level() ictxt = psb_cd_get_context(desc_data) call psb_info(ictxt, me, np) if (debug_level >= psb_debug_inner_) & & write(debug_unit,*) me,' ',trim(name),& & ' Entry ', size(precv) nlev = size(precv) allocate(mlprec_wrk(nlev),stat=info) if (info /= 0) then call psb_errpush(4010,name,a_err='Allocate') goto 9999 end if ! ! STEP 1 ! ! Copy the input vector X ! if (debug_level >= psb_debug_inner_) & & write(debug_unit,*) me,' ',trim(name),& & ' desc_data status',allocated(desc_data%matrix_data) nc2l = psb_cd_get_local_cols(precv(1)%base_desc) allocate(mlprec_wrk(1)%x2l(nc2l),mlprec_wrk(1)%y2l(nc2l), & & mlprec_wrk(1)%tx(nc2l), stat=info) if (info /= 0) then info=4025 call psb_errpush(info,name,i_err=(/4*nc2l,0,0,0,0/),& & a_err='real(psb_spk_)') goto 9999 end if call psb_geaxpby(sone,x,szero,mlprec_wrk(1)%tx,& & precv(1)%base_desc,info) call psb_geaxpby(sone,x,szero,mlprec_wrk(1)%x2l,& & precv(1)%base_desc,info) ! ! STEP 2 ! ! For each level but the finest one ... ! do ilev=2, nlev nc2l = psb_cd_get_local_cols(precv(ilev)%base_desc) nr2l = psb_cd_get_local_rows(precv(ilev)%base_desc) if (debug_level >= psb_debug_inner_) & & write(debug_unit,*) me,' ',trim(name), & & ' starting up sweep ',& & ilev,allocated(precv(ilev)%iprcparm),nc2l, nr2l allocate(mlprec_wrk(ilev)%tx(nc2l),mlprec_wrk(ilev)%y2l(nc2l),& & mlprec_wrk(ilev)%x2l(nc2l), stat=info) if (info /= 0) then info=4025 call psb_errpush(info,name,i_err=(/4*nc2l,0,0,0,0/),& & a_err='real(psb_spk_)') goto 9999 end if ! Apply prolongator transpose, i.e. restriction call psb_forward_map(sone,mlprec_wrk(ilev-1)%x2l,& & szero,mlprec_wrk(ilev)%x2l,& & precv(ilev)%map_desc,info,work=work) if (info /=0) then call psb_errpush(4001,name,a_err='Error during restriction') goto 9999 end if ! ! update x2l ! call psb_geaxpby(sone,mlprec_wrk(ilev)%x2l,szero,mlprec_wrk(ilev)%tx,& & precv(ilev)%base_desc,info) if (info /= 0) then call psb_errpush(4001,name,a_err='Error in update') goto 9999 end if if (debug_level >= psb_debug_inner_) & & write(debug_unit,*) me,' ',trim(name),& & ' done up sweep ', ilev enddo ! ! STEP 3 ! ! Apply the base preconditioner at the coarsest level ! call mld_baseprec_aply(sone,precv(nlev)%prec,mlprec_wrk(nlev)%x2l, & & szero, mlprec_wrk(nlev)%y2l,precv(nlev)%base_desc,trans,work,info) if (info /=0) then call psb_errpush(4010,name,a_err='baseprec_aply') goto 9999 end if if (debug_level >= psb_debug_inner_) write(debug_unit,*) & & me,' ',trim(name), ' done baseprec_aply ', nlev ! ! STEP 4 ! ! For each level but the coarsest one ... ! do ilev=nlev-1, 1, -1 if (debug_level >= psb_debug_inner_) & & write(debug_unit,*) me,' ',trim(name),& & ' starting down sweep',ilev ! ! Apply prolongator ! call psb_backward_map(sone,mlprec_wrk(ilev+1)%y2l,& & szero,mlprec_wrk(ilev)%y2l,& & precv(ilev+1)%map_desc,info,work=work) if (info /=0) then call psb_errpush(4001,name,a_err='Error during prolongation') goto 9999 end if ! ! Compute the residual ! call psb_spmm(-sone,precv(ilev)%base_a,mlprec_wrk(ilev)%y2l,& & sone,mlprec_wrk(ilev)%tx,precv(ilev)%base_desc,info,& & work=work,trans=trans) ! ! Apply the base preconditioner ! if (info == 0) call mld_baseprec_aply(sone,precv(ilev)%prec,& & mlprec_wrk(ilev)%tx,sone,mlprec_wrk(ilev)%y2l,precv(ilev)%base_desc,& & trans,work,info) if (info /=0) then call psb_errpush(4001,name,a_err=' spmm/baseprec_aply') goto 9999 end if if (debug_level >= psb_debug_inner_) & & write(debug_unit,*) me,' ',trim(name),& & ' done down sweep',ilev enddo ! ! STEP 5 ! ! Compute the output vector Y ! call psb_geaxpby(alpha,mlprec_wrk(1)%y2l,beta,y,precv(1)%base_desc,info) if (info /=0) then call psb_errpush(4001,name,a_err=' Final update') goto 9999 end if deallocate(mlprec_wrk,stat=info) if (info /= 0) then call psb_errpush(4000,name) goto 9999 end if 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 mlt_post_ml_aply ! ! Subroutine: mlt_twoside_ml_aply ! Version: real ! Note: internal subroutine of mld_smlprec_aply. ! ! This routine computes ! ! Y = beta*Y + alpha*op(M^(-1))*X, ! where ! - M is a symmetrized hybrid multilevel domain decomposition (Schwarz) ! preconditioner associated to a certain matrix A and stored in the array ! precv, ! - op(M^(-1)) is M^(-1) or its transpose, according to the value of trans, ! - X and Y are vectors, ! - alpha and beta are scalars. ! ! The preconditioner M is hybrid in the sense that it is multiplicative through ! the levels and additive inside a level; it is symmetrized since pre-smoothing ! and post-smoothing are applied at each level. ! ! For each level we have as many submatrices as processes (except for the coarsest ! level where we might have a replicated index space) and each process takes care ! of one submatrix. ! ! The multilevel preconditioner M is regarded as an array of 'one-level preconditioners', ! each representing the part of the preconditioner associated to a certain level. ! For each level ilev, the base preconditioner K(ilev) is stored in precv(ilev) ! and is associated to a matrix A(ilev), obtained by 'tranferring' the original ! matrix A (i.e. the matrix to be preconditioned) to the level ilev, through smoothed ! aggregation. ! ! The levels are numbered in increasing order starting from the finest one, i.e. ! level 1 is the finest level and A(1) is the matrix A. ! ! For details on the symmetrized hybrid multiplicative multilevel Schwarz ! preconditioner, see the Algorithm 3.2.2 of the book: ! B.F. Smith, P.E. Bjorstad & W.D. Gropp, ! Domain decomposition: parallel multilevel methods for elliptic partial ! differential equations, Cambridge University Press, 1996. ! ! For a description of the arguments see mld_smlprec_aply. ! ! A sketch of the algorithm implemented in this routine is provided below. ! (AV(ilev; sm_pr_) denotes the smoothed prolongator from level ilev to ! level ilev-1, while AV(ilev; sm_pr_t_) denotes its transpose, i.e. the ! corresponding restriction operator from level ilev-1 to level ilev). ! ! 1. X(1) = Xext ! ! 2. ! Apply the base peconditioner at the finest level ! Y(1) = (K(1)^(-1))*X(1) ! ! 3. ! Compute the residual at the finest level ! TX(1) = X(1) - A(1)*Y(1) ! ! 4. DO ilev=2, nlev ! ! ! Transfer the residual to the current (coarser) level ! X(ilev) = AV(ilev; sm_pr_t)*TX(ilev-1) ! ! ! Apply the base preconditioner at the current level. ! ! The sum over the subdomains is carried out in the ! ! application of K(ilev) ! Y(ilev) = (K(ilev)^(-1))*X(ilev) ! ! ! Compute the residual at the current level ! TX(ilev) = (X(ilev)-A(ilev)*Y(ilev)) ! ! ENDDO ! ! 5. DO ilev=NLEV-1,1,-1 ! ! ! Transfer Y(ilev+1) to the next finer level ! Y(ilev) = Y(ilev) + AV(ilev+1; sm_pr_)*Y(ilev+1) ! ! ! Compute the residual at the current level and apply to it the ! ! base preconditioner. The sum over the subdomains is carried out ! ! in the application of K(ilev) ! Y(ilev) = Y(ilev) + (K(ilev)**(-1))*(X(ilev)-A(ilev)*Y(ilev)) ! ! ENDDO ! ! 6. Yext = beta*Yext + alpha*Y(1) ! subroutine mlt_twoside_ml_aply(alpha,precv,x,beta,y,desc_data,trans,work,info) implicit none ! Arguments type(psb_desc_type),intent(in) :: desc_data type(mld_s_onelev_prec_type), intent(in) :: precv(:) real(psb_spk_),intent(in) :: alpha,beta real(psb_spk_),intent(in) :: x(:) real(psb_spk_),intent(inout) :: y(:) character, intent(in) :: trans real(psb_spk_),target :: work(:) integer, intent(out) :: info ! Local variables integer :: ictxt,np,me,i, nr2l,nc2l,err_act integer :: debug_level, debug_unit integer :: nlev, ilev character(len=20) :: name type psb_mlprec_wrk_type real(psb_spk_), allocatable :: tx(:), ty(:), x2l(:), y2l(:) end type psb_mlprec_wrk_type type(psb_mlprec_wrk_type), allocatable :: mlprec_wrk(:) name='mlt_twoside_ml_aply' info = 0 call psb_erractionsave(err_act) debug_unit = psb_get_debug_unit() debug_level = psb_get_debug_level() ictxt = psb_cd_get_context(desc_data) call psb_info(ictxt, me, np) if (debug_level >= psb_debug_inner_) & & write(debug_unit,*) me,' ',trim(name),& & ' Entry ', size(precv) nlev = size(precv) allocate(mlprec_wrk(nlev),stat=info) if (info /= 0) then call psb_errpush(4010,name,a_err='Allocate') goto 9999 end if ! STEP 1 ! ! Copy the input vector X ! nc2l = psb_cd_get_local_cols(precv(1)%base_desc) allocate(mlprec_wrk(1)%x2l(nc2l),mlprec_wrk(1)%y2l(nc2l), & & mlprec_wrk(1)%ty(nc2l), mlprec_wrk(1)%tx(nc2l), stat=info) if (info /= 0) then info=4025 call psb_errpush(info,name,i_err=(/4*nc2l,0,0,0,0/),& & a_err='real(psb_spk_)') goto 9999 end if call psb_geaxpby(sone,x,szero,mlprec_wrk(1)%x2l,& & precv(1)%base_desc,info) call psb_geaxpby(sone,x,szero,mlprec_wrk(1)%tx,& & precv(1)%base_desc,info) ! ! STEP 2 ! ! Apply the base preconditioner at the finest level ! call mld_baseprec_aply(sone,precv(1)%prec,mlprec_wrk(1)%x2l,& & szero,mlprec_wrk(1)%y2l,precv(1)%base_desc,& & trans,work,info) ! ! STEP 3 ! ! Compute the residual at the finest level ! mlprec_wrk(1)%ty = mlprec_wrk(1)%x2l if (info == 0) call psb_spmm(-sone,precv(1)%base_a,mlprec_wrk(1)%y2l,& & sone,mlprec_wrk(1)%ty,precv(1)%base_desc,info,& & work=work,trans=trans) if (info /=0) then call psb_errpush(4010,name,a_err='Fine level baseprec/residual') goto 9999 end if ! ! STEP 4 ! ! For each level but the finest one ... ! do ilev = 2, nlev nc2l = psb_cd_get_local_cols(precv(ilev)%base_desc) nr2l = psb_cd_get_local_rows(precv(ilev)%base_desc) allocate(mlprec_wrk(ilev)%tx(nc2l),mlprec_wrk(ilev)%ty(nc2l),& & mlprec_wrk(ilev)%y2l(nc2l),mlprec_wrk(ilev)%x2l(nc2l), stat=info) if (info /= 0) then info=4025 call psb_errpush(info,name,i_err=(/4*nc2l,0,0,0,0/),& & a_err='real(psb_spk_)') goto 9999 end if ! Apply prolongator transpose, i.e. restriction call psb_forward_map(sone,mlprec_wrk(ilev-1)%ty,& & szero,mlprec_wrk(ilev)%x2l,& & precv(ilev)%map_desc,info,work=work) if (info /=0) then call psb_errpush(4001,name,a_err='Error during restriction') goto 9999 end if call psb_geaxpby(sone,mlprec_wrk(ilev)%x2l,szero,mlprec_wrk(ilev)%tx,& & precv(ilev)%base_desc,info) ! ! Apply the base preconditioner ! if (info == 0) call mld_baseprec_aply(sone,precv(ilev)%prec,& & mlprec_wrk(ilev)%x2l,szero,mlprec_wrk(ilev)%y2l,& & precv(ilev)%base_desc,trans,work,info) ! ! Compute the residual (at all levels but the coarsest one) ! if(ilev < nlev) then mlprec_wrk(ilev)%ty = mlprec_wrk(ilev)%x2l if (info == 0) call psb_spmm(-sone,precv(ilev)%base_a,& & mlprec_wrk(ilev)%y2l,sone,mlprec_wrk(ilev)%ty,& & precv(ilev)%base_desc,info,work=work,trans=trans) endif if (info /=0) then call psb_errpush(4001,name,a_err='baseprec_aply/residual') goto 9999 end if enddo ! ! STEP 5 ! ! For each level but the coarsest one ... ! do ilev=nlev-1, 1, -1 ! ! Apply prolongator ! call psb_backward_map(sone,mlprec_wrk(ilev+1)%y2l,& & sone,mlprec_wrk(ilev)%y2l,& & precv(ilev+1)%map_desc,info,work=work) if (info /=0 ) then call psb_errpush(4001,name,a_err='Error during restriction') goto 9999 end if ! ! Compute the residual ! call psb_spmm(-sone,precv(ilev)%base_a,mlprec_wrk(ilev)%y2l,& & sone,mlprec_wrk(ilev)%tx,precv(ilev)%base_desc,info,& & work=work,trans=trans) ! ! Apply the base preconditioner ! if (info == 0) call mld_baseprec_aply(sone,precv(ilev)%prec,mlprec_wrk(ilev)%tx,& & sone,mlprec_wrk(ilev)%y2l,precv(ilev)%base_desc, trans, work,info) if (info /= 0) then call psb_errpush(4001,name,a_err='Error: residual/baseprec_aply') goto 9999 end if enddo ! ! STEP 6 ! ! Compute the output vector Y ! call psb_geaxpby(alpha,mlprec_wrk(1)%y2l,beta,y,& & precv(1)%base_desc,info) if (info /= 0) then call psb_errpush(4001,name,a_err='Error final update') goto 9999 end if deallocate(mlprec_wrk,stat=info) if (info /= 0) then call psb_errpush(4000,name) goto 9999 end if 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 mlt_twoside_ml_aply end subroutine mld_smlprec_aply