Merge branch 'dev-openmp' into development

.gitignore

@@ -12,9 +12,9 @@ config.log
 config.status
 
 # generated folder
-include/
-modules/
-docs/src/tmp
+/include/
+/modules/
+/docs/src/tmp
 autom4te.cache
 
 # the executable from tests

amgprec/amg_base_prec_type.F90

@@ -94,6 +94,7 @@ module amg_base_prec_type
     integer(psb_ipk_) :: aggr_omega_alg, aggr_eig, aggr_filter
     integer(psb_ipk_) :: coarse_mat, coarse_solve
   contains
+    procedure, pass(pm) :: get_coarse_mat  => ml_parms_get_coarse_mat
     procedure, pass(pm) :: get_coarse  => ml_parms_get_coarse
     procedure, pass(pm) :: clone       => ml_parms_clone
     procedure, pass(pm) :: descr       => ml_parms_descr
@@ -577,6 +578,33 @@ contains
     end select
   end function amg_stringval
 
+  function amg_get_coarse_mat_name(val) result(res)
+    character(len=15) :: res
+    integer :: val
+    select case(val)
+    case (0,1)
+      res = matrix_names(val)
+    case default
+      res = 'Unknown '
+    end select
+  end function amg_get_coarse_mat_name
+
+  subroutine amg_warn_coarse_mat(val,expected)
+    integer(psb_ipk_) :: val, expected
+    if (val /= expected) then
+      write(0,*) 'Warning: resetting COARSE_MAT on an existing hierarchy from ',&
+           & amg_get_coarse_mat_name(val), ' to ',amg_get_coarse_mat_name(expected)
+    end if
+  end subroutine amg_warn_coarse_mat
+
+  
+  function  ml_parms_get_coarse_mat(pm) result(res)
+    implicit none
+    class(amg_ml_parms), intent(in)    :: pm
+    integer(psb_ipk_) :: res
+    res   = pm%coarse_mat
+  end function ml_parms_get_coarse_mat
+
   subroutine  ml_parms_get_coarse(pm,pmin)
     implicit none
     class(amg_ml_parms), intent(inout) :: pm

amgprec/amg_c_ainv_solver.F90

@@ -58,6 +58,7 @@ module amg_c_ainv_solver
     procedure, pass(sv) :: check   => amg_c_ainv_solver_check
     procedure, pass(sv) :: build   => amg_c_ainv_solver_bld
     procedure, pass(sv) :: clone   => amg_c_ainv_solver_clone
+    procedure, pass(sv) :: clone_settings   => amg_c_ainv_solver_clone_settings
     procedure, pass(sv) :: cseti   => amg_c_ainv_solver_cseti
     procedure, pass(sv) :: csetc   => amg_c_ainv_solver_csetc
     procedure, pass(sv) :: csetr   => amg_c_ainv_solver_csetr
@@ -85,6 +86,16 @@ module amg_c_ainv_solver
     end subroutine amg_c_ainv_solver_clone
   end interface
 
+  interface
+    subroutine amg_c_ainv_solver_clone_settings(sv,svout,info)
+      import :: psb_desc_type, psb_cspmat_type,  psb_c_base_sparse_mat, &
+       & amg_c_base_solver_type, psb_dpk_, amg_c_ainv_solver_type, psb_ipk_
+      Implicit None
+      class(amg_c_ainv_solver_type), intent(inout)              :: sv
+      class(amg_c_base_solver_type), allocatable, intent(inout) :: svout
+      integer(psb_ipk_), intent(out)                            :: info
+    end subroutine amg_c_ainv_solver_clone_settings
+  end interface
 
   interface
     subroutine amg_c_ainv_solver_bld(a,desc_a,sv,info,b,amold,vmold,imold)

amgprec/amg_c_invk_solver.f90

@@ -52,6 +52,7 @@ module amg_c_invk_solver
   contains
     procedure, pass(sv) :: check   => amg_c_invk_solver_check
     procedure, pass(sv) :: clone   => amg_c_invk_solver_clone
+    procedure, pass(sv) :: clone_settings   => amg_c_invk_solver_clone_settings
     procedure, pass(sv) :: build   => amg_c_invk_solver_bld
     procedure, pass(sv) :: cseti   => amg_c_invk_solver_cseti
     procedure, pass(sv) :: descr   => amg_c_invk_solver_descr
@@ -73,6 +74,17 @@ module amg_c_invk_solver
     end subroutine amg_c_invk_solver_clone
   end interface
 
+  interface
+    subroutine amg_c_invk_solver_clone_settings(sv,svout,info)
+      import :: psb_desc_type, psb_cspmat_type,  psb_c_base_sparse_mat, &
+       & amg_c_base_solver_type, psb_spk_, amg_c_invk_solver_type, psb_ipk_
+      Implicit None
+      class(amg_c_invk_solver_type), intent(inout)              :: sv
+      class(amg_c_base_solver_type), allocatable, intent(inout) :: svout
+      integer(psb_ipk_), intent(out)                            :: info
+    end subroutine amg_c_invk_solver_clone_settings
+  end interface
+
   interface
     subroutine amg_c_invk_solver_bld(a,desc_a,sv,info,b,amold,vmold,imold)
       import :: psb_desc_type, psb_cspmat_type,  psb_c_base_sparse_mat, &

amgprec/amg_c_invt_solver.f90

@@ -52,6 +52,7 @@ module amg_c_invt_solver
   contains
     procedure, pass(sv) :: check   => amg_c_invt_solver_check
     procedure, pass(sv) :: clone   => amg_c_invt_solver_clone
+    procedure, pass(sv) :: clone_settings   => amg_c_invt_solver_clone_settings
     procedure, pass(sv) :: build   => amg_c_invt_solver_bld
     procedure, pass(sv) :: cseti   => amg_c_invt_solver_cseti
     procedure, pass(sv) :: csetr   => amg_c_invt_solver_csetr
@@ -73,6 +74,17 @@ module amg_c_invt_solver
     end subroutine amg_c_invt_solver_clone
   end interface
 
+  interface
+    subroutine amg_c_invt_solver_clone_settings(sv,svout,info)
+      import :: psb_desc_type, psb_cspmat_type,  psb_c_base_sparse_mat, &
+       & amg_c_base_solver_type, psb_spk_, amg_c_invt_solver_type, psb_ipk_
+      Implicit None
+      class(amg_c_invt_solver_type), intent(inout)              :: sv
+      class(amg_c_base_solver_type), allocatable, intent(inout) :: svout
+      integer(psb_ipk_), intent(out)                            :: info
+    end subroutine amg_c_invt_solver_clone_settings
+  end interface
+
   interface
     subroutine amg_c_invt_solver_bld(a,desc_a,sv,info,b,amold,vmold,imold)
       import :: psb_desc_type, psb_cspmat_type,  psb_c_base_sparse_mat, &

amgprec/amg_c_onelev_mod.f90

@@ -189,6 +189,7 @@ module amg_c_onelev_mod
     procedure, pass(lv) :: descr   => amg_c_base_onelev_descr
     procedure, pass(lv) :: default => c_base_onelev_default
     procedure, pass(lv) :: free    => amg_c_base_onelev_free
+    procedure, pass(lv) :: free_smoothers => amg_c_base_onelev_free_smoothers
     procedure, pass(lv) :: nullify => c_base_onelev_nullify
     procedure, pass(lv) :: check => amg_c_base_onelev_check
     procedure, pass(lv) :: dump  => amg_c_base_onelev_dump
@@ -285,7 +286,7 @@ module amg_c_onelev_mod
     end subroutine amg_c_base_onelev_cnv
   end interface
 
-interface
+  interface
     subroutine amg_c_base_onelev_free(lv,info)
       import :: psb_cspmat_type, psb_c_vect_type, psb_c_base_vect_type, &
            & psb_clinmap_type, psb_spk_, amg_c_onelev_type, &
@@ -297,6 +298,18 @@ interface
     end subroutine amg_c_base_onelev_free
   end interface
 
+  interface
+    subroutine amg_c_base_onelev_free_smoothers(lv,info)
+      import :: psb_cspmat_type, psb_c_vect_type, psb_c_base_vect_type, &
+           & psb_clinmap_type, psb_spk_, amg_c_onelev_type, &
+           & psb_ipk_, psb_epk_, psb_desc_type
+      implicit none
+
+      class(amg_c_onelev_type), intent(inout) :: lv
+      integer(psb_ipk_), intent(out)                :: info
+    end subroutine amg_c_base_onelev_free_smoothers
+  end interface
+
   interface
     subroutine amg_c_base_onelev_check(lv,info)
       import :: psb_cspmat_type, psb_c_vect_type, psb_c_base_vect_type, &

amgprec/amg_c_prec_type.f90

@@ -135,7 +135,9 @@ module amg_c_prec_type
     procedure, pass(prec)               :: build        => amg_cprecbld
     procedure, pass(prec)               :: hierarchy_build   => amg_c_hierarchy_bld
     procedure, pass(prec)               :: hierarchy_rebuild => amg_c_hierarchy_rebld
+    procedure, pass(prec)               :: hierarchy_free    => amg_c_hierarchy_free    
     procedure, pass(prec)               :: smoothers_build   => amg_c_smoothers_bld
+    procedure, pass(prec)               :: smoothers_free    => amg_c_smoothers_free
     procedure, pass(prec)               :: descr        =>  amg_cfile_prec_descr
   end type amg_cprec_type
 
@@ -345,6 +347,14 @@ module amg_c_prec_type
     end subroutine amg_c_smoothers_bld
   end interface amg_smoothers_bld
 
+  interface amg_smoothers_free
+    module procedure  amg_c_smoothers_free
+  end interface amg_smoothers_free
+
+  interface amg_hierarchy_free
+    module procedure amg_c_hierarchy_free
+  end interface amg_hierarchy_free
+
 contains
   !
   ! Function returning a pointer to the smoother
@@ -618,6 +628,68 @@ contains
 
   end subroutine amg_c_prec_free
 
+  subroutine amg_c_smoothers_free(prec,info)
+
+    implicit none
+
+    ! Arguments
+    class(amg_cprec_type), intent(inout) :: prec
+    integer(psb_ipk_), intent(out)        :: info
+
+    ! Local variables
+    integer(psb_ipk_)   :: me,err_act,i
+    character(len=20)   :: name
+
+    info=psb_success_
+    name = 'amg_c_smoothers_free'
+    call psb_erractionsave(err_act)
+    if (psb_errstatus_fatal()) then
+      info = psb_err_internal_error_; goto 9999
+    end if
+
+    if (allocated(prec%precv)) then
+      do i=1,size(prec%precv)
+        call prec%precv(i)%free_smoothers(info)
+      end do
+    end if
+
+    call psb_erractionrestore(err_act)
+    return
+
+9999 call psb_error_handler(err_act)
+    return
+
+  end subroutine amg_c_smoothers_free
+
+  subroutine amg_c_hierarchy_free(prec,info)
+
+    implicit none
+
+    ! Arguments
+    class(amg_cprec_type), intent(inout) :: prec
+    integer(psb_ipk_), intent(out)        :: info
+
+    ! Local variables
+    integer(psb_ipk_)   :: me,err_act,i
+    character(len=20)   :: name
+
+    info=psb_success_
+    name = 'amg_c_hierarchy_free'
+    call psb_erractionsave(err_act)
+    if (psb_errstatus_fatal()) then
+      info = psb_err_internal_error_; goto 9999
+    end if
+
+    me=-1
+    write(0,*) 'Missing implementation '
+
+    call psb_erractionrestore(err_act)
+    return
+
+9999 call psb_error_handler(err_act)
+    return
+
+  end subroutine amg_c_hierarchy_free
 
 
   !

amgprec/amg_d_ainv_solver.F90

@@ -58,6 +58,7 @@ module amg_d_ainv_solver
     procedure, pass(sv) :: check   => amg_d_ainv_solver_check
     procedure, pass(sv) :: build   => amg_d_ainv_solver_bld
     procedure, pass(sv) :: clone   => amg_d_ainv_solver_clone
+    procedure, pass(sv) :: clone_settings   => amg_d_ainv_solver_clone_settings
     procedure, pass(sv) :: cseti   => amg_d_ainv_solver_cseti
     procedure, pass(sv) :: csetc   => amg_d_ainv_solver_csetc
     procedure, pass(sv) :: csetr   => amg_d_ainv_solver_csetr
@@ -85,6 +86,16 @@ module amg_d_ainv_solver
     end subroutine amg_d_ainv_solver_clone
   end interface
 
+  interface
+    subroutine amg_d_ainv_solver_clone_settings(sv,svout,info)
+      import :: psb_desc_type, psb_dspmat_type,  psb_d_base_sparse_mat, &
+       & amg_d_base_solver_type, psb_dpk_, amg_d_ainv_solver_type, psb_ipk_
+      Implicit None
+      class(amg_d_ainv_solver_type), intent(inout)              :: sv
+      class(amg_d_base_solver_type), allocatable, intent(inout) :: svout
+      integer(psb_ipk_), intent(out)                            :: info
+    end subroutine amg_d_ainv_solver_clone_settings
+  end interface
 
   interface
     subroutine amg_d_ainv_solver_bld(a,desc_a,sv,info,b,amold,vmold,imold)

amgprec/amg_d_invk_solver.f90

@@ -52,6 +52,7 @@ module amg_d_invk_solver
   contains
     procedure, pass(sv) :: check   => amg_d_invk_solver_check
     procedure, pass(sv) :: clone   => amg_d_invk_solver_clone
+    procedure, pass(sv) :: clone_settings   => amg_d_invk_solver_clone_settings
     procedure, pass(sv) :: build   => amg_d_invk_solver_bld
     procedure, pass(sv) :: cseti   => amg_d_invk_solver_cseti
     procedure, pass(sv) :: descr   => amg_d_invk_solver_descr
@@ -73,6 +74,17 @@ module amg_d_invk_solver
     end subroutine amg_d_invk_solver_clone
   end interface
 
+  interface
+    subroutine amg_d_invk_solver_clone_settings(sv,svout,info)
+      import :: psb_desc_type, psb_dspmat_type,  psb_d_base_sparse_mat, &
+       & amg_d_base_solver_type, psb_dpk_, amg_d_invk_solver_type, psb_ipk_
+      Implicit None
+      class(amg_d_invk_solver_type), intent(inout)              :: sv
+      class(amg_d_base_solver_type), allocatable, intent(inout) :: svout
+      integer(psb_ipk_), intent(out)                            :: info
+    end subroutine amg_d_invk_solver_clone_settings
+  end interface
+
   interface
     subroutine amg_d_invk_solver_bld(a,desc_a,sv,info,b,amold,vmold,imold)
       import :: psb_desc_type, psb_dspmat_type,  psb_d_base_sparse_mat, &

amgprec/amg_d_invt_solver.f90

@@ -52,6 +52,7 @@ module amg_d_invt_solver
   contains
     procedure, pass(sv) :: check   => amg_d_invt_solver_check
     procedure, pass(sv) :: clone   => amg_d_invt_solver_clone
+    procedure, pass(sv) :: clone_settings   => amg_d_invt_solver_clone_settings
     procedure, pass(sv) :: build   => amg_d_invt_solver_bld
     procedure, pass(sv) :: cseti   => amg_d_invt_solver_cseti
     procedure, pass(sv) :: csetr   => amg_d_invt_solver_csetr
@@ -73,6 +74,17 @@ module amg_d_invt_solver
     end subroutine amg_d_invt_solver_clone
   end interface
 
+  interface
+    subroutine amg_d_invt_solver_clone_settings(sv,svout,info)
+      import :: psb_desc_type, psb_dspmat_type,  psb_d_base_sparse_mat, &
+       & amg_d_base_solver_type, psb_dpk_, amg_d_invt_solver_type, psb_ipk_
+      Implicit None
+      class(amg_d_invt_solver_type), intent(inout)              :: sv
+      class(amg_d_base_solver_type), allocatable, intent(inout) :: svout
+      integer(psb_ipk_), intent(out)                            :: info
+    end subroutine amg_d_invt_solver_clone_settings
+  end interface
+
   interface
     subroutine amg_d_invt_solver_bld(a,desc_a,sv,info,b,amold,vmold,imold)
       import :: psb_desc_type, psb_dspmat_type,  psb_d_base_sparse_mat, &

amgprec/amg_d_onelev_mod.f90

@@ -190,6 +190,7 @@ module amg_d_onelev_mod
     procedure, pass(lv) :: descr   => amg_d_base_onelev_descr
     procedure, pass(lv) :: default => d_base_onelev_default
     procedure, pass(lv) :: free    => amg_d_base_onelev_free
+    procedure, pass(lv) :: free_smoothers => amg_d_base_onelev_free_smoothers
     procedure, pass(lv) :: nullify => d_base_onelev_nullify
     procedure, pass(lv) :: check => amg_d_base_onelev_check
     procedure, pass(lv) :: dump  => amg_d_base_onelev_dump
@@ -286,7 +287,7 @@ module amg_d_onelev_mod
     end subroutine amg_d_base_onelev_cnv
   end interface
 
-interface
+  interface
     subroutine amg_d_base_onelev_free(lv,info)
       import :: psb_dspmat_type, psb_d_vect_type, psb_d_base_vect_type, &
            & psb_dlinmap_type, psb_dpk_, amg_d_onelev_type, &
@@ -298,6 +299,18 @@ interface
     end subroutine amg_d_base_onelev_free
   end interface
 
+  interface
+    subroutine amg_d_base_onelev_free_smoothers(lv,info)
+      import :: psb_dspmat_type, psb_d_vect_type, psb_d_base_vect_type, &
+           & psb_dlinmap_type, psb_dpk_, amg_d_onelev_type, &
+           & psb_ipk_, psb_epk_, psb_desc_type
+      implicit none
+
+      class(amg_d_onelev_type), intent(inout) :: lv
+      integer(psb_ipk_), intent(out)                :: info
+    end subroutine amg_d_base_onelev_free_smoothers
+  end interface
+
   interface
     subroutine amg_d_base_onelev_check(lv,info)
       import :: psb_dspmat_type, psb_d_vect_type, psb_d_base_vect_type, &

amgprec/amg_d_prec_type.f90

@@ -135,7 +135,9 @@ module amg_d_prec_type
     procedure, pass(prec)               :: build        => amg_dprecbld
     procedure, pass(prec)               :: hierarchy_build   => amg_d_hierarchy_bld
     procedure, pass(prec)               :: hierarchy_rebuild => amg_d_hierarchy_rebld
+    procedure, pass(prec)               :: hierarchy_free    => amg_d_hierarchy_free    
     procedure, pass(prec)               :: smoothers_build   => amg_d_smoothers_bld
+    procedure, pass(prec)               :: smoothers_free    => amg_d_smoothers_free
     procedure, pass(prec)               :: descr        =>  amg_dfile_prec_descr
   end type amg_dprec_type
 
@@ -345,6 +347,14 @@ module amg_d_prec_type
     end subroutine amg_d_smoothers_bld
   end interface amg_smoothers_bld
 
+  interface amg_smoothers_free
+    module procedure  amg_d_smoothers_free
+  end interface amg_smoothers_free
+
+  interface amg_hierarchy_free
+    module procedure amg_d_hierarchy_free
+  end interface amg_hierarchy_free
+
 contains
   !
   ! Function returning a pointer to the smoother
@@ -618,6 +628,68 @@ contains
 
   end subroutine amg_d_prec_free
 
+  subroutine amg_d_smoothers_free(prec,info)
+
+    implicit none
+
+    ! Arguments
+    class(amg_dprec_type), intent(inout) :: prec
+    integer(psb_ipk_), intent(out)        :: info
+
+    ! Local variables
+    integer(psb_ipk_)   :: me,err_act,i
+    character(len=20)   :: name
+
+    info=psb_success_
+    name = 'amg_d_smoothers_free'
+    call psb_erractionsave(err_act)
+    if (psb_errstatus_fatal()) then
+      info = psb_err_internal_error_; goto 9999
+    end if
+
+    if (allocated(prec%precv)) then
+      do i=1,size(prec%precv)
+        call prec%precv(i)%free_smoothers(info)
+      end do
+    end if
+
+    call psb_erractionrestore(err_act)
+    return
+
+9999 call psb_error_handler(err_act)
+    return
+
+  end subroutine amg_d_smoothers_free
+
+  subroutine amg_d_hierarchy_free(prec,info)
+
+    implicit none
+
+    ! Arguments
+    class(amg_dprec_type), intent(inout) :: prec
+    integer(psb_ipk_), intent(out)        :: info
+
+    ! Local variables
+    integer(psb_ipk_)   :: me,err_act,i
+    character(len=20)   :: name
+
+    info=psb_success_
+    name = 'amg_d_hierarchy_free'
+    call psb_erractionsave(err_act)
+    if (psb_errstatus_fatal()) then
+      info = psb_err_internal_error_; goto 9999
+    end if
+
+    me=-1
+    write(0,*) 'Missing implementation '
+
+    call psb_erractionrestore(err_act)
+    return
+
+9999 call psb_error_handler(err_act)
+    return
+
+  end subroutine amg_d_hierarchy_free
 
 
   !

amgprec/amg_s_ainv_solver.F90

@@ -58,6 +58,7 @@ module amg_s_ainv_solver
     procedure, pass(sv) :: check   => amg_s_ainv_solver_check
     procedure, pass(sv) :: build   => amg_s_ainv_solver_bld
     procedure, pass(sv) :: clone   => amg_s_ainv_solver_clone
+    procedure, pass(sv) :: clone_settings   => amg_s_ainv_solver_clone_settings
     procedure, pass(sv) :: cseti   => amg_s_ainv_solver_cseti
     procedure, pass(sv) :: csetc   => amg_s_ainv_solver_csetc
     procedure, pass(sv) :: csetr   => amg_s_ainv_solver_csetr
@@ -85,6 +86,16 @@ module amg_s_ainv_solver
     end subroutine amg_s_ainv_solver_clone
   end interface
 
+  interface
+    subroutine amg_s_ainv_solver_clone_settings(sv,svout,info)
+      import :: psb_desc_type, psb_sspmat_type,  psb_s_base_sparse_mat, &
+       & amg_s_base_solver_type, psb_dpk_, amg_s_ainv_solver_type, psb_ipk_
+      Implicit None
+      class(amg_s_ainv_solver_type), intent(inout)              :: sv
+      class(amg_s_base_solver_type), allocatable, intent(inout) :: svout
+      integer(psb_ipk_), intent(out)                            :: info
+    end subroutine amg_s_ainv_solver_clone_settings
+  end interface
 
   interface
     subroutine amg_s_ainv_solver_bld(a,desc_a,sv,info,b,amold,vmold,imold)

amgprec/amg_s_invk_solver.f90

@@ -52,6 +52,7 @@ module amg_s_invk_solver
   contains
     procedure, pass(sv) :: check   => amg_s_invk_solver_check
     procedure, pass(sv) :: clone   => amg_s_invk_solver_clone
+    procedure, pass(sv) :: clone_settings   => amg_s_invk_solver_clone_settings
     procedure, pass(sv) :: build   => amg_s_invk_solver_bld
     procedure, pass(sv) :: cseti   => amg_s_invk_solver_cseti
     procedure, pass(sv) :: descr   => amg_s_invk_solver_descr
@@ -73,6 +74,17 @@ module amg_s_invk_solver
     end subroutine amg_s_invk_solver_clone
   end interface
 
+  interface
+    subroutine amg_s_invk_solver_clone_settings(sv,svout,info)
+      import :: psb_desc_type, psb_sspmat_type,  psb_s_base_sparse_mat, &
+       & amg_s_base_solver_type, psb_spk_, amg_s_invk_solver_type, psb_ipk_
+      Implicit None
+      class(amg_s_invk_solver_type), intent(inout)              :: sv
+      class(amg_s_base_solver_type), allocatable, intent(inout) :: svout
+      integer(psb_ipk_), intent(out)                            :: info
+    end subroutine amg_s_invk_solver_clone_settings
+  end interface
+
   interface
     subroutine amg_s_invk_solver_bld(a,desc_a,sv,info,b,amold,vmold,imold)
       import :: psb_desc_type, psb_sspmat_type,  psb_s_base_sparse_mat, &

amgprec/amg_s_invt_solver.f90

@@ -52,6 +52,7 @@ module amg_s_invt_solver
   contains
     procedure, pass(sv) :: check   => amg_s_invt_solver_check
     procedure, pass(sv) :: clone   => amg_s_invt_solver_clone
+    procedure, pass(sv) :: clone_settings   => amg_s_invt_solver_clone_settings
     procedure, pass(sv) :: build   => amg_s_invt_solver_bld
     procedure, pass(sv) :: cseti   => amg_s_invt_solver_cseti
     procedure, pass(sv) :: csetr   => amg_s_invt_solver_csetr
@@ -73,6 +74,17 @@ module amg_s_invt_solver
     end subroutine amg_s_invt_solver_clone
   end interface
 
+  interface
+    subroutine amg_s_invt_solver_clone_settings(sv,svout,info)
+      import :: psb_desc_type, psb_sspmat_type,  psb_s_base_sparse_mat, &
+       & amg_s_base_solver_type, psb_spk_, amg_s_invt_solver_type, psb_ipk_
+      Implicit None
+      class(amg_s_invt_solver_type), intent(inout)              :: sv
+      class(amg_s_base_solver_type), allocatable, intent(inout) :: svout
+      integer(psb_ipk_), intent(out)                            :: info
+    end subroutine amg_s_invt_solver_clone_settings
+  end interface
+
   interface
     subroutine amg_s_invt_solver_bld(a,desc_a,sv,info,b,amold,vmold,imold)
       import :: psb_desc_type, psb_sspmat_type,  psb_s_base_sparse_mat, &

amgprec/amg_s_onelev_mod.f90

@@ -190,6 +190,7 @@ module amg_s_onelev_mod
     procedure, pass(lv) :: descr   => amg_s_base_onelev_descr
     procedure, pass(lv) :: default => s_base_onelev_default
     procedure, pass(lv) :: free    => amg_s_base_onelev_free
+    procedure, pass(lv) :: free_smoothers => amg_s_base_onelev_free_smoothers
     procedure, pass(lv) :: nullify => s_base_onelev_nullify
     procedure, pass(lv) :: check => amg_s_base_onelev_check
     procedure, pass(lv) :: dump  => amg_s_base_onelev_dump
@@ -286,7 +287,7 @@ module amg_s_onelev_mod
     end subroutine amg_s_base_onelev_cnv
   end interface
 
-interface
+  interface
     subroutine amg_s_base_onelev_free(lv,info)
       import :: psb_sspmat_type, psb_s_vect_type, psb_s_base_vect_type, &
            & psb_slinmap_type, psb_spk_, amg_s_onelev_type, &
@@ -298,6 +299,18 @@ interface
     end subroutine amg_s_base_onelev_free
   end interface
 
+  interface
+    subroutine amg_s_base_onelev_free_smoothers(lv,info)
+      import :: psb_sspmat_type, psb_s_vect_type, psb_s_base_vect_type, &
+           & psb_slinmap_type, psb_spk_, amg_s_onelev_type, &
+           & psb_ipk_, psb_epk_, psb_desc_type
+      implicit none
+
+      class(amg_s_onelev_type), intent(inout) :: lv
+      integer(psb_ipk_), intent(out)                :: info
+    end subroutine amg_s_base_onelev_free_smoothers
+  end interface
+
   interface
     subroutine amg_s_base_onelev_check(lv,info)
       import :: psb_sspmat_type, psb_s_vect_type, psb_s_base_vect_type, &

amgprec/amg_s_prec_type.f90

@@ -135,7 +135,9 @@ module amg_s_prec_type
     procedure, pass(prec)               :: build        => amg_sprecbld
     procedure, pass(prec)               :: hierarchy_build   => amg_s_hierarchy_bld
     procedure, pass(prec)               :: hierarchy_rebuild => amg_s_hierarchy_rebld
+    procedure, pass(prec)               :: hierarchy_free    => amg_s_hierarchy_free    
     procedure, pass(prec)               :: smoothers_build   => amg_s_smoothers_bld
+    procedure, pass(prec)               :: smoothers_free    => amg_s_smoothers_free
     procedure, pass(prec)               :: descr        =>  amg_sfile_prec_descr
   end type amg_sprec_type
 
@@ -345,6 +347,14 @@ module amg_s_prec_type
     end subroutine amg_s_smoothers_bld
   end interface amg_smoothers_bld
 
+  interface amg_smoothers_free
+    module procedure  amg_s_smoothers_free
+  end interface amg_smoothers_free
+
+  interface amg_hierarchy_free
+    module procedure amg_s_hierarchy_free
+  end interface amg_hierarchy_free
+
 contains
   !
   ! Function returning a pointer to the smoother
@@ -618,6 +628,68 @@ contains
 
   end subroutine amg_s_prec_free
 
+  subroutine amg_s_smoothers_free(prec,info)
+
+    implicit none
+
+    ! Arguments
+    class(amg_sprec_type), intent(inout) :: prec
+    integer(psb_ipk_), intent(out)        :: info
+
+    ! Local variables
+    integer(psb_ipk_)   :: me,err_act,i
+    character(len=20)   :: name
+
+    info=psb_success_
+    name = 'amg_s_smoothers_free'
+    call psb_erractionsave(err_act)
+    if (psb_errstatus_fatal()) then
+      info = psb_err_internal_error_; goto 9999
+    end if
+
+    if (allocated(prec%precv)) then
+      do i=1,size(prec%precv)
+        call prec%precv(i)%free_smoothers(info)
+      end do
+    end if
+
+    call psb_erractionrestore(err_act)
+    return
+
+9999 call psb_error_handler(err_act)
+    return
+
+  end subroutine amg_s_smoothers_free
+
+  subroutine amg_s_hierarchy_free(prec,info)
+
+    implicit none
+
+    ! Arguments
+    class(amg_sprec_type), intent(inout) :: prec
+    integer(psb_ipk_), intent(out)        :: info
+
+    ! Local variables
+    integer(psb_ipk_)   :: me,err_act,i
+    character(len=20)   :: name
+
+    info=psb_success_
+    name = 'amg_s_hierarchy_free'
+    call psb_erractionsave(err_act)
+    if (psb_errstatus_fatal()) then
+      info = psb_err_internal_error_; goto 9999
+    end if
+
+    me=-1
+    write(0,*) 'Missing implementation '
+
+    call psb_erractionrestore(err_act)
+    return
+
+9999 call psb_error_handler(err_act)
+    return
+
+  end subroutine amg_s_hierarchy_free
 
 
   !

amgprec/amg_z_ainv_solver.F90

@@ -58,6 +58,7 @@ module amg_z_ainv_solver
     procedure, pass(sv) :: check   => amg_z_ainv_solver_check
     procedure, pass(sv) :: build   => amg_z_ainv_solver_bld
     procedure, pass(sv) :: clone   => amg_z_ainv_solver_clone
+    procedure, pass(sv) :: clone_settings   => amg_z_ainv_solver_clone_settings
     procedure, pass(sv) :: cseti   => amg_z_ainv_solver_cseti
     procedure, pass(sv) :: csetc   => amg_z_ainv_solver_csetc
     procedure, pass(sv) :: csetr   => amg_z_ainv_solver_csetr
@@ -85,6 +86,16 @@ module amg_z_ainv_solver
     end subroutine amg_z_ainv_solver_clone
   end interface
 
+  interface
+    subroutine amg_z_ainv_solver_clone_settings(sv,svout,info)
+      import :: psb_desc_type, psb_zspmat_type,  psb_z_base_sparse_mat, &
+       & amg_z_base_solver_type, psb_dpk_, amg_z_ainv_solver_type, psb_ipk_
+      Implicit None
+      class(amg_z_ainv_solver_type), intent(inout)              :: sv
+      class(amg_z_base_solver_type), allocatable, intent(inout) :: svout
+      integer(psb_ipk_), intent(out)                            :: info
+    end subroutine amg_z_ainv_solver_clone_settings
+  end interface
 
   interface
     subroutine amg_z_ainv_solver_bld(a,desc_a,sv,info,b,amold,vmold,imold)

amgprec/amg_z_invk_solver.f90

@@ -52,6 +52,7 @@ module amg_z_invk_solver
   contains
     procedure, pass(sv) :: check   => amg_z_invk_solver_check
     procedure, pass(sv) :: clone   => amg_z_invk_solver_clone
+    procedure, pass(sv) :: clone_settings   => amg_z_invk_solver_clone_settings
     procedure, pass(sv) :: build   => amg_z_invk_solver_bld
     procedure, pass(sv) :: cseti   => amg_z_invk_solver_cseti
     procedure, pass(sv) :: descr   => amg_z_invk_solver_descr
@@ -73,6 +74,17 @@ module amg_z_invk_solver
     end subroutine amg_z_invk_solver_clone
   end interface
 
+  interface
+    subroutine amg_z_invk_solver_clone_settings(sv,svout,info)
+      import :: psb_desc_type, psb_zspmat_type,  psb_z_base_sparse_mat, &
+       & amg_z_base_solver_type, psb_dpk_, amg_z_invk_solver_type, psb_ipk_
+      Implicit None
+      class(amg_z_invk_solver_type), intent(inout)              :: sv
+      class(amg_z_base_solver_type), allocatable, intent(inout) :: svout
+      integer(psb_ipk_), intent(out)                            :: info
+    end subroutine amg_z_invk_solver_clone_settings
+  end interface
+
   interface
     subroutine amg_z_invk_solver_bld(a,desc_a,sv,info,b,amold,vmold,imold)
       import :: psb_desc_type, psb_zspmat_type,  psb_z_base_sparse_mat, &

amgprec/amg_z_invt_solver.f90

@@ -52,6 +52,7 @@ module amg_z_invt_solver
   contains
     procedure, pass(sv) :: check   => amg_z_invt_solver_check
     procedure, pass(sv) :: clone   => amg_z_invt_solver_clone
+    procedure, pass(sv) :: clone_settings   => amg_z_invt_solver_clone_settings
     procedure, pass(sv) :: build   => amg_z_invt_solver_bld
     procedure, pass(sv) :: cseti   => amg_z_invt_solver_cseti
     procedure, pass(sv) :: csetr   => amg_z_invt_solver_csetr
@@ -73,6 +74,17 @@ module amg_z_invt_solver
     end subroutine amg_z_invt_solver_clone
   end interface
 
+  interface
+    subroutine amg_z_invt_solver_clone_settings(sv,svout,info)
+      import :: psb_desc_type, psb_zspmat_type,  psb_z_base_sparse_mat, &
+       & amg_z_base_solver_type, psb_dpk_, amg_z_invt_solver_type, psb_ipk_
+      Implicit None
+      class(amg_z_invt_solver_type), intent(inout)              :: sv
+      class(amg_z_base_solver_type), allocatable, intent(inout) :: svout
+      integer(psb_ipk_), intent(out)                            :: info
+    end subroutine amg_z_invt_solver_clone_settings
+  end interface
+
   interface
     subroutine amg_z_invt_solver_bld(a,desc_a,sv,info,b,amold,vmold,imold)
       import :: psb_desc_type, psb_zspmat_type,  psb_z_base_sparse_mat, &

amgprec/amg_z_onelev_mod.f90

@@ -189,6 +189,7 @@ module amg_z_onelev_mod
     procedure, pass(lv) :: descr   => amg_z_base_onelev_descr
     procedure, pass(lv) :: default => z_base_onelev_default
     procedure, pass(lv) :: free    => amg_z_base_onelev_free
+    procedure, pass(lv) :: free_smoothers => amg_z_base_onelev_free_smoothers
     procedure, pass(lv) :: nullify => z_base_onelev_nullify
     procedure, pass(lv) :: check => amg_z_base_onelev_check
     procedure, pass(lv) :: dump  => amg_z_base_onelev_dump
@@ -285,7 +286,7 @@ module amg_z_onelev_mod
     end subroutine amg_z_base_onelev_cnv
   end interface
 
-interface
+  interface
     subroutine amg_z_base_onelev_free(lv,info)
       import :: psb_zspmat_type, psb_z_vect_type, psb_z_base_vect_type, &
            & psb_zlinmap_type, psb_dpk_, amg_z_onelev_type, &
@@ -297,6 +298,18 @@ interface
     end subroutine amg_z_base_onelev_free
   end interface
 
+  interface
+    subroutine amg_z_base_onelev_free_smoothers(lv,info)
+      import :: psb_zspmat_type, psb_z_vect_type, psb_z_base_vect_type, &
+           & psb_zlinmap_type, psb_dpk_, amg_z_onelev_type, &
+           & psb_ipk_, psb_epk_, psb_desc_type
+      implicit none
+
+      class(amg_z_onelev_type), intent(inout) :: lv
+      integer(psb_ipk_), intent(out)                :: info
+    end subroutine amg_z_base_onelev_free_smoothers
+  end interface
+
   interface
     subroutine amg_z_base_onelev_check(lv,info)
       import :: psb_zspmat_type, psb_z_vect_type, psb_z_base_vect_type, &

amgprec/amg_z_prec_type.f90

@@ -135,7 +135,9 @@ module amg_z_prec_type
     procedure, pass(prec)               :: build        => amg_zprecbld
     procedure, pass(prec)               :: hierarchy_build   => amg_z_hierarchy_bld
     procedure, pass(prec)               :: hierarchy_rebuild => amg_z_hierarchy_rebld
+    procedure, pass(prec)               :: hierarchy_free    => amg_z_hierarchy_free    
     procedure, pass(prec)               :: smoothers_build   => amg_z_smoothers_bld
+    procedure, pass(prec)               :: smoothers_free    => amg_z_smoothers_free
     procedure, pass(prec)               :: descr        =>  amg_zfile_prec_descr
   end type amg_zprec_type
 
@@ -345,6 +347,14 @@ module amg_z_prec_type
     end subroutine amg_z_smoothers_bld
   end interface amg_smoothers_bld
 
+  interface amg_smoothers_free
+    module procedure  amg_z_smoothers_free
+  end interface amg_smoothers_free
+
+  interface amg_hierarchy_free
+    module procedure amg_z_hierarchy_free
+  end interface amg_hierarchy_free
+
 contains
   !
   ! Function returning a pointer to the smoother
@@ -618,6 +628,68 @@ contains
 
   end subroutine amg_z_prec_free
 
+  subroutine amg_z_smoothers_free(prec,info)
+
+    implicit none
+
+    ! Arguments
+    class(amg_zprec_type), intent(inout) :: prec
+    integer(psb_ipk_), intent(out)        :: info
+
+    ! Local variables
+    integer(psb_ipk_)   :: me,err_act,i
+    character(len=20)   :: name
+
+    info=psb_success_
+    name = 'amg_z_smoothers_free'
+    call psb_erractionsave(err_act)
+    if (psb_errstatus_fatal()) then
+      info = psb_err_internal_error_; goto 9999
+    end if
+
+    if (allocated(prec%precv)) then
+      do i=1,size(prec%precv)
+        call prec%precv(i)%free_smoothers(info)
+      end do
+    end if
+
+    call psb_erractionrestore(err_act)
+    return
+
+9999 call psb_error_handler(err_act)
+    return
+
+  end subroutine amg_z_smoothers_free
+
+  subroutine amg_z_hierarchy_free(prec,info)
+
+    implicit none
+
+    ! Arguments
+    class(amg_zprec_type), intent(inout) :: prec
+    integer(psb_ipk_), intent(out)        :: info
+
+    ! Local variables
+    integer(psb_ipk_)   :: me,err_act,i
+    character(len=20)   :: name
+
+    info=psb_success_
+    name = 'amg_z_hierarchy_free'
+    call psb_erractionsave(err_act)
+    if (psb_errstatus_fatal()) then
+      info = psb_err_internal_error_; goto 9999
+    end if
+
+    me=-1
+    write(0,*) 'Missing implementation '
+
+    call psb_erractionrestore(err_act)
+    return
+
+9999 call psb_error_handler(err_act)
+    return
+
+  end subroutine amg_z_hierarchy_free
 
 
   !

amgprec/impl/aggregator/amg_c_ptap_bld.f90

@@ -76,7 +76,7 @@ subroutine amg_c_ptap_bld(a_csr,desc_a,nlaggr,parms,ac,&
   integer(psb_ipk_) :: nrow, ncol, nrl, nzl, ip, nzt, i, k
   integer(psb_lpk_) ::  nrsave, ncsave, nzsave, nza
   logical, parameter :: do_timings=.false., oldstyle=.false., debug=.false.  
-  integer(psb_ipk_), save :: idx_spspmm=-1
+  integer(psb_ipk_), save :: idx_spspmm=-1, idx_cpytrans1=-1, idx_cpytrans2=-1
 
   name='amg_ptap_bld'
   if(psb_get_errstatus().ne.0) return 
@@ -93,7 +93,11 @@ subroutine amg_c_ptap_bld(a_csr,desc_a,nlaggr,parms,ac,&
   ncol  = desc_a%get_local_cols()
 
   if ((do_timings).and.(idx_spspmm==-1)) &
-       & idx_spspmm = psb_get_timer_idx("SPMM_BLD: par_spspmm")
+       & idx_spspmm = psb_get_timer_idx("PTAP_BLD: par_spspmm")
+  if ((do_timings).and.(idx_cpytrans1==-1)) &
+       & idx_cpytrans1 = psb_get_timer_idx("PTAP_BLD: cpy&trans1")
+  if ((do_timings).and.(idx_cpytrans2==-1)) &
+       & idx_cpytrans2 = psb_get_timer_idx("PTAP_BLD: cpy&trans2")
 
   naggr   = nlaggr(me+1)
   ntaggr  = sum(nlaggr)
@@ -128,6 +132,7 @@ subroutine amg_c_ptap_bld(a_csr,desc_a,nlaggr,parms,ac,&
   ! Ok first product done.
 
   if (present(desc_ax)) then
+    if (do_timings) call psb_tic(idx_cpytrans1)
     block 
       call coo_prol%cp_to_coo(coo_restr,info)
       call coo_restr%set_ncols(desc_ac%get_local_cols())
@@ -137,7 +142,7 @@ subroutine amg_c_ptap_bld(a_csr,desc_a,nlaggr,parms,ac,&
       call coo_restr%set_ncols(desc_ax%get_local_cols())
     end block
     call csr_restr%cp_from_coo(coo_restr,info)
-
+    if (do_timings) call psb_toc(idx_cpytrans1)
     if (info /= psb_success_) then 
       call psb_errpush(psb_err_from_subroutine_,name,a_err='spcnv coo_restr')
       goto 9999
@@ -167,27 +172,28 @@ subroutine amg_c_ptap_bld(a_csr,desc_a,nlaggr,parms,ac,&
 
     call coo_restr%transp()
     nzl = coo_restr%get_nzeros()
-    nrl = desc_ac%get_local_rows() 
-    i=0
+    nrl = desc_ac%get_local_rows()
+    call coo_restr%fix(info)
+    i=coo_restr%get_nzeros()
     !
     ! Only keep local rows
     !
-    do k=1, nzl
-      if ((1 <= coo_restr%ia(k)) .and.(coo_restr%ia(k) <= nrl)) then
-        i = i+1
-        coo_restr%val(i) = coo_restr%val(k)
-        coo_restr%ia(i)  = coo_restr%ia(k)
-        coo_restr%ja(i)  = coo_restr%ja(k)
+    search: do k=i,1,-1
+      if (coo_restr%ia(k) <= nrl) then
+        call coo_restr%set_nzeros(k)
+        exit search
       end if
-    end do
-    call coo_restr%set_nzeros(i)
-    call coo_restr%fix(info) 
+    end do search
+
     nzl  = coo_restr%get_nzeros()
     call coo_restr%set_nrows(desc_ac%get_local_rows())
     call coo_restr%set_ncols(desc_a%get_local_cols())
     if (debug) call check_coo(me,trim(name)//' Check 2 on coo_restr:',coo_restr)
+    if (do_timings) call psb_tic(idx_cpytrans2)
+        
     call csr_restr%cp_from_coo(coo_restr,info)
 
+    if (do_timings) call psb_toc(idx_cpytrans2)
     if (info /= psb_success_) then 
       call psb_errpush(psb_err_from_subroutine_,name,a_err='spcnv coo_restr')
       goto 9999

amgprec/impl/aggregator/amg_c_soc1_map_bld.F90

@@ -72,7 +72,9 @@ subroutine amg_c_soc1_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
   use psb_base_mod
   use amg_base_prec_type
   use amg_c_inner_mod
-
+#if defined(OPENMP)
+  use omp_lib
+#endif
   implicit none
 
   ! Arguments
@@ -85,7 +87,7 @@ subroutine amg_c_soc1_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
   integer(psb_ipk_), intent(out)               :: info
 
   ! Local variables
-  integer(psb_ipk_), allocatable :: ils(:), neigh(:), irow(:), icol(:),&
+  integer(psb_ipk_), allocatable :: neigh(:), irow(:), icol(:),&
        & ideg(:), idxs(:)
   integer(psb_lpk_), allocatable :: tmpaggr(:)
   complex(psb_spk_), allocatable  :: val(:), diag(:)
@@ -99,6 +101,9 @@ subroutine amg_c_soc1_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
   integer(psb_ipk_)   :: nrow, ncol, n_ne
   integer(psb_lpk_)   :: nrglob
   character(len=20)   :: name, ch_err
+  integer(psb_ipk_), save :: idx_soc1_p1=-1, idx_soc1_p2=-1, idx_soc1_p3=-1
+  integer(psb_ipk_), save :: idx_soc1_p0=-1
+  logical, parameter      :: do_timings=.true.
 
   info=psb_success_
   name = 'amg_soc1_map_bld'
@@ -114,6 +119,14 @@ subroutine amg_c_soc1_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
   nrow   = desc_a%get_local_rows()
   ncol   = desc_a%get_local_cols()
   nrglob = desc_a%get_global_rows()
+  if ((do_timings).and.(idx_soc1_p0==-1))       &
+       & idx_soc1_p0 = psb_get_timer_idx("SOC1_MAP: phase0")
+  if ((do_timings).and.(idx_soc1_p1==-1))       &
+       & idx_soc1_p1 = psb_get_timer_idx("SOC1_MAP: phase1")
+  if ((do_timings).and.(idx_soc1_p2==-1))       &
+       & idx_soc1_p2 = psb_get_timer_idx("SOC1_MAP: phase2")
+  if ((do_timings).and.(idx_soc1_p3==-1))       &
+       & idx_soc1_p3 = psb_get_timer_idx("SOC1_MAP: phase3")
 
   nr = a%get_nrows()
   nc = a%get_ncols()
@@ -133,41 +146,204 @@ subroutine amg_c_soc1_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
     goto 9999
   end if
 
+  if (do_timings) call psb_tic(idx_soc1_p0)
   call a%cp_to(acsr)
+  if (do_timings) call psb_toc(idx_soc1_p0)
   if (clean_zeros) call acsr%clean_zeros(info)
   if (iorder == amg_aggr_ord_nat_) then 
+    !$omp parallel do private(i) schedule(static)
     do i=1, nr
       ilaggr(i) = -(nr+1)
-      idxs(i)   = i 
+      idxs(i)   = i
     end do
-  else 
+    !$omp end parallel do 
+  else
+    !$omp parallel do private(i) schedule(static)
     do i=1, nr
       ilaggr(i) = -(nr+1)
       ideg(i)   = acsr%irp(i+1) - acsr%irp(i)
     end do
+    !$omp end parallel do 
     call psb_msort(ideg,ix=idxs,dir=psb_sort_down_)
   end if
-
+  if (do_timings) call psb_tic(idx_soc1_p1)
 
   !
   ! Phase one: Start with disjoint groups.
   ! 
   naggr = 0
-  icnt = 0
+#if defined(OPENMP)
+  block
+    integer(psb_ipk_), allocatable  :: bnds(:), locnaggr(:)
+    integer(psb_ipk_) :: myth,nths, kk
+    ! The parallelization makes use of a locaggr(:) array; each thread
+    ! keeps its own version of naggr, and when the loop ends, a prefix is applied
+    ! to locnaggr to determine:
+    ! 1. The total number of aggregaters NAGGR;
+    ! 2. How much should each thread shift its own aggregates
+    ! Part 2 requires to keep track of which thread defined each entry
+    ! of ilaggr(), so that each entry can be adjusted correctly: even
+    ! if an entry I belongs to the range BNDS(TH)>BNDS(TH+1)-1, it may have
+    ! been set because it is strongly connected to an entry J belonging to a
+    ! different thread. 
+
+    !$omp parallel shared(bnds,idxs,locnaggr,ilaggr,nr,naggr,diag,theta,nths,info) &
+    !$omp private(icol,val,myth,kk) 
+    block
+      integer(psb_ipk_) :: ii,nlp,k,kp,n,ia,isz, nc, i,j,m, nz, ilg,  ip, rsz
+      integer(psb_lpk_) :: itmp
+      !$omp master
+      nths = omp_get_num_threads()
+      allocate(bnds(0:nths),locnaggr(0:nths+1))
+      locnaggr(:) = 0
+      bnds(0) = 1
+      !$omp end master      
+      !$omp barrier
+      myth = omp_get_thread_num()
+      rsz = nr/nths
+      if (myth < mod(nr,nths)) rsz = rsz + 1
+      bnds(myth+1) = rsz
+      !$omp barrier
+      !$omp master
+      do i=1,nths
+        bnds(i) = bnds(i) + bnds(i-1)
+      end do
+      info = 0
+      !$omp end master
+      !$omp barrier
+      
+      !$omp do schedule(static) private(disjoint) 
+      do kk=0, nths-1
+        step1: do ii=bnds(kk), bnds(kk+1)-1
+          i = idxs(ii)          
+          if (info /= 0) cycle step1
+          if ((i<1).or.(i>nr)) then
+            !$omp atomic write
+            info=psb_err_internal_error_
+            !$omp end atomic 
+            call psb_errpush(info,name)
+            cycle step1
+            !goto 9999
+          end if
+            
+          if (ilaggr(i) == -(nr+1)) then
+            nz         = (acsr%irp(i+1)-acsr%irp(i))
+            if ((nz<0).or.(nz>size(icol))) then
+              !$omp atomic write
+              info=psb_err_internal_error_
+              !$omp end atomic 
+              call psb_errpush(info,name)
+              cycle step1
+              !goto 9999
+            end if
+
+            icol(1:nz) = acsr%ja(acsr%irp(i):acsr%irp(i+1)-1)
+            val(1:nz)  = acsr%val(acsr%irp(i):acsr%irp(i+1)-1) 
+
+            !
+            ! Build the set of all strongly coupled nodes 
+            !
+            ip = 0
+            do k=1, nz
+              j   = icol(k)
+              ! If any of the neighbours is already assigned,
+              ! we will not reset. 
+              if (j>nr) cycle step1
+              if (ilaggr(j) > 0) cycle step1
+              if (abs(val(k)) > theta*sqrt(abs(diag(i)*diag(j)))) then
+                ip = ip + 1
+                icol(ip) = icol(k)
+              end if
+            enddo
+            
+            !
+            ! If the whole strongly coupled neighborhood of I is
+            ! as yet unconnected, turn it into the next aggregate.
+            ! Same if ip==0 (in which case, neighborhood only
+            ! contains I even if it does not look like it from matrix)
+            ! The fact that DISJOINT is private and not under lock
+            ! generates a certain un-repeatability, in that between
+            ! computing DISJOINT and assigning, another thread might
+            ! alter the values of ILAGGR.
+            ! However, a certain unrepeatability is already present
+            ! because the sequence of aggregates is computed with a
+            ! different order than in serial mode.
+            ! In any case, even if the enteries of ILAGGR may be
+            ! overwritten, the important thing is that each entry is
+            ! consistent and they generate a correct aggregation map.
+            !
+            disjoint = all(ilaggr(icol(1:ip)) == -(nr+1)).or.(ip==0)
+            if (disjoint) then
+              locnaggr(kk)     = locnaggr(kk) + 1
+              itmp = (bnds(kk)-1+locnaggr(kk))*nths+kk
+              if (itmp < (bnds(kk)-1+locnaggr(kk))) then
+                !$omp atomic update
+                info = max(12345678,info)
+                !$omp end atomic
+                cycle step1
+              end if
+              !$omp atomic write
+              ilaggr(i) = itmp
+              !$omp end atomic
+              do k=1, ip
+                !$omp atomic write
+                ilaggr(icol(k)) = itmp
+                !$omp end atomic
+              end do
+            end if
+          end if
+        enddo step1
+      end do
+      !$omp end do
+
+      !$omp master
+      naggr = sum(locnaggr(0:nths-1))
+      do i=1,nths
+        locnaggr(i) = locnaggr(i) + locnaggr(i-1)
+      end do
+      do i=nths+1,1,-1
+        locnaggr(i) = locnaggr(i-1)
+      end do
+      locnaggr(0) = 0
+      !$omp end master 
+      !$omp barrier
+      !$omp  do schedule(static) 
+      do kk=0, nths-1
+        do ii=bnds(kk), bnds(kk+1)-1
+          if (ilaggr(ii) > 0) then 
+            kp = mod(ilaggr(ii),nths)
+            ilaggr(ii) = (ilaggr(ii)/nths)- (bnds(kp)-1) + locnaggr(kp)
+          end if
+        end do
+      end do
+      !$omp end do
+    end block
+    !$omp end parallel
+  end block
+  if (info /= 0) then
+    if (info == 12345678) write(0,*) 'Overflow in encoding ILAGGR'
+    info=psb_err_internal_error_
+    call psb_errpush(info,name)
+    goto 9999
+  end if
+#else
   step1: do ii=1, nr
+    if (info /= 0) cycle
     i = idxs(ii)
     if ((i<1).or.(i>nr)) then
       info=psb_err_internal_error_
       call psb_errpush(info,name)
-      goto 9999
+      cycle step1
+      !goto 9999
     end if
-    
+
     if (ilaggr(i) == -(nr+1)) then
       nz         = (acsr%irp(i+1)-acsr%irp(i))
       if ((nz<0).or.(nz>size(icol))) then
         info=psb_err_internal_error_
         call psb_errpush(info,name)
-        goto 9999
+        cycle step1
+        !goto 9999
       end if
 
       icol(1:nz) = acsr%ja(acsr%irp(i):acsr%irp(i+1)-1)
@@ -176,7 +352,7 @@ subroutine amg_c_soc1_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
       !
       ! Build the set of all strongly coupled nodes 
       !
-      ip = 0 
+      ip = 0
       do k=1, nz
         j   = icol(k)
         if ((1<=j).and.(j<=nr)) then 
@@ -194,8 +370,7 @@ subroutine amg_c_soc1_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
       ! contains I even if it does not look like it from matrix)
       !
       disjoint = all(ilaggr(icol(1:ip)) == -(nr+1)).or.(ip==0)
-      if (disjoint) then 
-        icnt      = icnt + 1 
+      if (disjoint) then       
         naggr     = naggr + 1
         do k=1, ip
           ilaggr(icol(k)) = naggr
@@ -204,16 +379,22 @@ subroutine amg_c_soc1_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
       end if
     endif
   enddo step1
-  
+#endif
   if (debug_level >= psb_debug_outer_) then 
     write(debug_unit,*) me,' ',trim(name),&
-         & ' Check 1:',count(ilaggr == -(nr+1))
+         & ' Check   1:',naggr,count(ilaggr(1:nr) == -(nr+1)), count(ilaggr(1:nr)>0),&
+         & count(ilaggr(1:nr) == -(nr+1))+count(ilaggr(1:nr)>0),nr
   end if
-
+  if (do_timings) call psb_toc(idx_soc1_p1)
+  if (do_timings) call psb_tic(idx_soc1_p2)
   !
   ! Phase two: join the neighbours
   !
+  !$omp workshare
   tmpaggr = ilaggr
+  !$omp end workshare
+  !$omp parallel do schedule(static) shared(tmpaggr,ilaggr,nr,naggr,diag,theta)& 
+  !$omp     private(ii,i,j,k,nz,icol,val,ip,cpling)
   step2: do ii=1,nr
     i = idxs(ii)
 
@@ -244,8 +425,15 @@ subroutine amg_c_soc1_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
       end if
     end if
   end do step2
+  !$omp end parallel do
+  if (do_timings) call psb_toc(idx_soc1_p2)
+  if (debug_level >= psb_debug_outer_) then 
+    write(debug_unit,*) me,' ',trim(name),&
+         & ' Check 1.5:',naggr,count(ilaggr(1:nr) == -(nr+1)), count(ilaggr(1:nr)>0),&
+         & count(ilaggr(1:nr) == -(nr+1))+count(ilaggr(1:nr)>0),nr
+  end if
 
-
+  if (do_timings) call psb_tic(idx_soc1_p3)
   !
   ! Phase three: sweep over leftovers, if any 
   !
@@ -274,7 +462,6 @@ subroutine amg_c_soc1_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
         end if
       enddo
       if (ip > 0) then
-        icnt      = icnt + 1 
         naggr     = naggr + 1
         ilaggr(i) = naggr
         do k=1, ip
@@ -292,7 +479,10 @@ subroutine amg_c_soc1_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
   end do step3
 
   ! Any leftovers?
+  !$omp parallel do schedule(static) shared(ilaggr,info)& 
+  !$omp     private(ii,i,j,k,nz,icol,val,ip)
   do i=1, nr
+    if (info /= 0) cycle
     if (ilaggr(i) < 0) then
       nz = (acsr%irp(i+1)-acsr%irp(i))
       if (nz == 1) then
@@ -303,15 +493,18 @@ subroutine amg_c_soc1_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
         ! other processes. 
         ilaggr(i) = -(nrglob+nr)
       else
+        !$omp atomic write
         info=psb_err_internal_error_
+        !$omp end atomic
         call psb_errpush(info,name,a_err='Fatal error: non-singleton leftovers')
-        goto 9999
+        cycle 
       endif
     end if
   end do
-
+  !$omp end parallel do
+  if (info /= 0) goto 9999
+  if (do_timings) call psb_toc(idx_soc1_p3)
   if (naggr > ncol) then 
-    !write(0,*) name,'Error : naggr > ncol',naggr,ncol
     info=psb_err_internal_error_
     call psb_errpush(info,name,a_err='Fatal error: naggr>ncol')
     goto 9999
@@ -336,9 +529,13 @@ subroutine amg_c_soc1_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
   nlaggr(:) = 0
   nlaggr(me+1) = naggr
   call psb_sum(ctxt,nlaggr(1:np))
+  if (debug_level >= psb_debug_outer_) then 
+    write(debug_unit,*) me,' ',trim(name),&
+         & ' Check   2:',naggr,count(ilaggr(1:nr) == -(nr+1)), count(ilaggr(1:nr)>0),&
+         & count(ilaggr(1:nr) == -(nr+1))+count(ilaggr(1:nr)>0),nr
+  end if
 
   call acsr%free()
-
   call psb_erractionrestore(err_act)
   return
 

amgprec/impl/aggregator/amg_c_soc2_map_bld.F90

@@ -68,9 +68,12 @@
 !
 subroutine amg_c_soc2_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,info)
 
-  use psb_base_mod
+  use psb_base_mod 
   use amg_base_prec_type
   use amg_c_inner_mod
+#if defined(OPENMP)
+  use omp_lib
+#endif
 
   implicit none
 
@@ -99,6 +102,9 @@ subroutine amg_c_soc2_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
   integer(psb_ipk_) :: np, me
   integer(psb_ipk_) :: nrow, ncol, n_ne
   character(len=20)  :: name, ch_err
+  integer(psb_ipk_), save :: idx_soc2_p1=-1, idx_soc2_p2=-1, idx_soc2_p3=-1
+  integer(psb_ipk_), save :: idx_soc2_p0=-1
+  logical, parameter      :: do_timings=.true.
 
   info=psb_success_
   name = 'amg_soc2_map_bld'
@@ -114,6 +120,14 @@ subroutine amg_c_soc2_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
   nrow   = desc_a%get_local_rows()
   ncol   = desc_a%get_local_cols()
   nrglob = desc_a%get_global_rows()
+  if ((do_timings).and.(idx_soc2_p0==-1))       &
+       & idx_soc2_p0 = psb_get_timer_idx("SOC2_MAP: phase0")
+  if ((do_timings).and.(idx_soc2_p1==-1))       &
+       & idx_soc2_p1 = psb_get_timer_idx("SOC2_MAP: phase1")
+  if ((do_timings).and.(idx_soc2_p2==-1))       &
+       & idx_soc2_p2 = psb_get_timer_idx("SOC2_MAP: phase2")
+  if ((do_timings).and.(idx_soc2_p3==-1))       &
+       & idx_soc2_p3 = psb_get_timer_idx("SOC2_MAP: phase3")
 
   nr = a%get_nrows()
   nc = a%get_ncols()
@@ -125,6 +139,7 @@ subroutine amg_c_soc2_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
     goto 9999
   end if
 
+  if (do_timings) call psb_tic(idx_soc2_p0)
   diag = a%get_diag(info)
   if(info /= psb_success_) then
     info=psb_err_from_subroutine_
@@ -137,55 +152,217 @@ subroutine amg_c_soc2_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
   ! 
   call a%cp_to(muij)
   if (clean_zeros) call muij%clean_zeros(info)
+  !$omp parallel do private(i,j,k) shared(nr,diag,muij) schedule(static)
   do i=1, nr
     do k=muij%irp(i),muij%irp(i+1)-1
       j = muij%ja(k)
       if (j<= nr) muij%val(k) = abs(muij%val(k))/sqrt(abs(diag(i)*diag(j)))
     end do
   end do
-
+  !$omp end parallel do 
   !
   ! Compute the 1-neigbour; mark strong links with +1, weak links with -1
   !
   call s_neigh_coo%allocate(nr,nr,muij%get_nzeros())
-  ip = 0 
+  !$omp parallel do private(i,j,k) shared(nr,diag,muij) schedule(static)
   do i=1, nr
     do k=muij%irp(i),muij%irp(i+1)-1
       j = muij%ja(k)
+      s_neigh_coo%ia(k)  = i
+      s_neigh_coo%ja(k)  = j
       if (j<=nr) then 
-        ip = ip + 1
-        s_neigh_coo%ia(ip)  = i
-        s_neigh_coo%ja(ip)  = j
         if (real(muij%val(k)) >= theta) then 
-          s_neigh_coo%val(ip) = sone
+          s_neigh_coo%val(k) = sone
         else
-          s_neigh_coo%val(ip) = -sone
+          s_neigh_coo%val(k) = -sone
         end if
+      else
+        s_neigh_coo%val(k) = -sone        
       end if
     end do
   end do
+  !$omp end parallel do
   !write(*,*) 'S_NEIGH: ',nr,ip
-  call s_neigh_coo%set_nzeros(ip)
+  call s_neigh_coo%set_nzeros(muij%get_nzeros())
   call s_neigh%mv_from_coo(s_neigh_coo,info)
 
-  if (iorder == amg_aggr_ord_nat_) then 
+  if (iorder == amg_aggr_ord_nat_) then
+    
+    !$omp parallel do private(i) shared(ilaggr,idxs) schedule(static)
     do i=1, nr
       ilaggr(i) = -(nr+1)
       idxs(i)   = i 
     end do
+    !$omp end parallel do 
   else 
+    !$omp parallel do private(i) shared(ilaggr,idxs,muij)  schedule(static)
     do i=1, nr
       ilaggr(i) = -(nr+1)
       ideg(i)   = muij%irp(i+1) - muij%irp(i)
     end do
+    !$omp end parallel do
     call psb_msort(ideg,ix=idxs,dir=psb_sort_down_)
   end if
 
+  if (do_timings) call psb_toc(idx_soc2_p0)
+  if (do_timings) call psb_tic(idx_soc2_p1)
 
   !
   ! Phase one: Start with disjoint groups.
   ! 
   naggr = 0
+#if defined(OPENMP)
+  block
+    integer(psb_ipk_), allocatable  :: bnds(:), locnaggr(:)
+    integer(psb_ipk_) :: myth,nths, kk
+    ! The parallelization makes use of a locaggr(:) array; each thread
+    ! keeps its own version of naggr, and when the loop ends, a prefix is applied
+    ! to locnaggr to determine:
+    ! 1. The total number of aggregaters NAGGR;
+    ! 2. How much should each thread shift its own aggregates
+    ! Part 2 requires to keep track of which thread defined each entry
+    ! of ilaggr(), so that each entry can be adjusted correctly: even
+    ! if an entry I belongs to the range BNDS(TH)>BNDS(TH+1)-1, it may have
+    ! been set because it is strongly connected to an entry J belonging to a
+    ! different thread. 
+
+    !$omp parallel shared(s_neigh,bnds,idxs,locnaggr,ilaggr,nr,naggr,diag,theta,nths,info) &
+    !$omp private(icol,val,myth,kk) 
+    block
+      integer(psb_ipk_) :: ii,nlp,k,kp,n,ia,isz,nc,i,j,m,nz,ilg,ip,rsz,ip1,nzcnt
+      integer(psb_lpk_) :: itmp
+      !$omp master
+      nths = omp_get_num_threads()
+      allocate(bnds(0:nths),locnaggr(0:nths+1))
+      locnaggr(:) = 0
+      bnds(0) = 1
+      !$omp end master      
+      !$omp barrier
+      myth = omp_get_thread_num()
+      rsz = nr/nths
+      if (myth < mod(nr,nths)) rsz = rsz + 1
+      bnds(myth+1) = rsz
+      !$omp barrier
+      !$omp master
+      do i=1,nths
+        bnds(i) = bnds(i) + bnds(i-1)
+      end do
+      info = 0
+      !$omp end master
+      !$omp barrier
+
+      !$omp do schedule(static) private(disjoint) 
+      do kk=0, nths-1
+        step1: do ii=bnds(kk), bnds(kk+1)-1
+          i = idxs(ii)
+          if (info /= 0) then
+            write(0,*) ' Step1:',kk,ii,i,info
+            cycle step1
+          end if
+          if ((i<1).or.(i>nr)) then
+            !$omp atomic write
+            info=psb_err_internal_error_
+            !$omp end atomic 
+            call psb_errpush(info,name)
+            cycle step1
+            !goto 9999
+          end if
+
+
+          if (ilaggr(i) == -(nr+1)) then 
+            !
+            ! Get the 1-neighbourhood of I 
+            !
+            ip1 = s_neigh%irp(i)
+            nz  = s_neigh%irp(i+1)-ip1
+            !
+            ! If the neighbourhood only contains I, skip it
+            !
+            if (nz ==0) then
+              ilaggr(i) = 0
+              cycle step1
+            end if
+            if ((nz==1).and.(s_neigh%ja(ip1)==i)) then
+              ilaggr(i) = 0
+              cycle step1
+            end if
+
+            nzcnt = count(real(s_neigh%val(ip1:ip1+nz-1)) > 0)
+            icol(1:nzcnt) = pack(s_neigh%ja(ip1:ip1+nz-1),(real(s_neigh%val(ip1:ip1+nz-1)) > 0))
+            disjoint = all(ilaggr(icol(1:nzcnt)) == -(nr+1)) 
+
+            !
+            ! If the whole strongly coupled neighborhood of I is
+            ! as yet unconnected, turn it into the next aggregate.
+            ! Same if ip==0 (in which case, neighborhood only
+            ! contains I even if it does not look like it from matrix)
+            ! The fact that DISJOINT is private and not under lock
+            ! generates a certain un-repeatability, in that between
+            ! computing DISJOINT and assigning, another thread might
+            ! alter the values of ILAGGR.
+            ! However, a certain unrepeatability is already present
+            ! because the sequence of aggregates is computed with a
+            ! different order than in serial mode.
+            ! In any case, even if the enteries of ILAGGR may be
+            ! overwritten, the important thing is that each entry is
+            ! consistent and they generate a correct aggregation map.
+            !
+            if (disjoint) then
+              locnaggr(kk)     = locnaggr(kk) + 1
+              itmp = (bnds(kk)-1+locnaggr(kk))*nths+kk
+              if (itmp < (bnds(kk)-1+locnaggr(kk))) then
+                !$omp atomic update
+                info = max(12345678,info)
+                !$omp end atomic
+                cycle step1
+              end if
+              !$omp atomic write
+              ilaggr(i) = itmp
+              !$omp end atomic
+              do k=1, nzcnt
+                !$omp atomic write
+                ilaggr(icol(k)) = itmp
+                !$omp end atomic
+              end do
+            end if
+          end if
+        enddo step1
+      end do
+      !$omp end do
+
+      !$omp master
+      naggr = sum(locnaggr(0:nths-1))
+      do i=1,nths
+        locnaggr(i) = locnaggr(i) + locnaggr(i-1)
+      end do
+      do i=nths+1,1,-1
+        locnaggr(i) = locnaggr(i-1)
+      end do
+      locnaggr(0) = 0
+      !write(0,*) 'LNAG ',locnaggr(nths+1)
+      !$omp end master 
+      !$omp barrier
+      !$omp  do schedule(static) 
+      do kk=0, nths-1
+        do ii=bnds(kk), bnds(kk+1)-1
+          if (ilaggr(ii) > 0) then 
+            kp = mod(ilaggr(ii),nths)
+            ilaggr(ii) = (ilaggr(ii)/nths)- (bnds(kp)-1) + locnaggr(kp)
+          end if
+        end do
+      end do
+      !$omp end do
+    end block
+    !$omp end parallel
+  end block
+  if (info /= 0) then
+    if (info == 12345678) write(0,*) 'Overflow in encoding ILAGGR'
+    info=psb_err_internal_error_
+    call psb_errpush(info,name)
+    goto 9999
+  end if
+
+#else
   icnt = 0
   step1: do ii=1, nr
     i = idxs(ii)
@@ -224,16 +401,21 @@ subroutine amg_c_soc2_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
       end if
     endif
   enddo step1
-  
+#endif  
   if (debug_level >= psb_debug_outer_) then 
     write(debug_unit,*) me,' ',trim(name),&
          & ' Check 1:',count(ilaggr == -(nr+1))
   end if
-
+  if (do_timings) call psb_toc(idx_soc2_p1)
+  if (do_timings) call psb_tic(idx_soc2_p2)
   !
   ! Phase two: join the neighbours
   !
+  !$omp workshare
   tmpaggr = ilaggr
+  !$omp end workshare
+  !$omp parallel do schedule(static) shared(tmpaggr,ilaggr,nr,naggr,diag,muij,s_neigh)& 
+  !$omp     private(ii,i,j,k,nz,icol,val,ip,cpling)
   step2: do ii=1,nr
     i = idxs(ii)
 
@@ -259,8 +441,9 @@ subroutine amg_c_soc2_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
       end if
     end if
   end do step2
-
-
+  !$omp end parallel do
+  if (do_timings) call psb_toc(idx_soc2_p2)
+  if (do_timings) call psb_tic(idx_soc2_p3)
   !
   ! Phase three: sweep over leftovers, if any 
   !
@@ -294,6 +477,8 @@ subroutine amg_c_soc2_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
   end do step3
 
   ! Any leftovers?
+  !$omp parallel do schedule(static) shared(ilaggr,s_neigh,info)& 
+  !$omp     private(ii,i,j,k)
   do i=1, nr
     if (ilaggr(i) <= 0) then
       nz = (s_neigh%irp(i+1)-s_neigh%irp(i))
@@ -305,13 +490,17 @@ subroutine amg_c_soc2_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
         ! other processes. 
         ilaggr(i) = -(nrglob+nr)
       else
+        !$omp atomic write
         info=psb_err_internal_error_
+        !$omp end atomic
         call psb_errpush(info,name,a_err='Fatal error: non-singleton leftovers')
-        goto 9999
+        cycle
       endif
     end if
   end do
-
+  !$omp end parallel do
+  if (info /= 0) goto 9999
+  if (do_timings) call psb_toc(idx_soc2_p3)  
   if (naggr > ncol) then 
     info=psb_err_internal_error_
     call psb_errpush(info,name,a_err='Fatal error: naggr>ncol')

amgprec/impl/aggregator/amg_d_ptap_bld.f90

@@ -76,7 +76,7 @@ subroutine amg_d_ptap_bld(a_csr,desc_a,nlaggr,parms,ac,&
   integer(psb_ipk_) :: nrow, ncol, nrl, nzl, ip, nzt, i, k
   integer(psb_lpk_) ::  nrsave, ncsave, nzsave, nza
   logical, parameter :: do_timings=.false., oldstyle=.false., debug=.false.  
-  integer(psb_ipk_), save :: idx_spspmm=-1
+  integer(psb_ipk_), save :: idx_spspmm=-1, idx_cpytrans1=-1, idx_cpytrans2=-1
 
   name='amg_ptap_bld'
   if(psb_get_errstatus().ne.0) return 
@@ -93,7 +93,11 @@ subroutine amg_d_ptap_bld(a_csr,desc_a,nlaggr,parms,ac,&
   ncol  = desc_a%get_local_cols()
 
   if ((do_timings).and.(idx_spspmm==-1)) &
-       & idx_spspmm = psb_get_timer_idx("SPMM_BLD: par_spspmm")
+       & idx_spspmm = psb_get_timer_idx("PTAP_BLD: par_spspmm")
+  if ((do_timings).and.(idx_cpytrans1==-1)) &
+       & idx_cpytrans1 = psb_get_timer_idx("PTAP_BLD: cpy&trans1")
+  if ((do_timings).and.(idx_cpytrans2==-1)) &
+       & idx_cpytrans2 = psb_get_timer_idx("PTAP_BLD: cpy&trans2")
 
   naggr   = nlaggr(me+1)
   ntaggr  = sum(nlaggr)
@@ -128,6 +132,7 @@ subroutine amg_d_ptap_bld(a_csr,desc_a,nlaggr,parms,ac,&
   ! Ok first product done.
 
   if (present(desc_ax)) then
+    if (do_timings) call psb_tic(idx_cpytrans1)
     block 
       call coo_prol%cp_to_coo(coo_restr,info)
       call coo_restr%set_ncols(desc_ac%get_local_cols())
@@ -137,7 +142,7 @@ subroutine amg_d_ptap_bld(a_csr,desc_a,nlaggr,parms,ac,&
       call coo_restr%set_ncols(desc_ax%get_local_cols())
     end block
     call csr_restr%cp_from_coo(coo_restr,info)
-
+    if (do_timings) call psb_toc(idx_cpytrans1)
     if (info /= psb_success_) then 
       call psb_errpush(psb_err_from_subroutine_,name,a_err='spcnv coo_restr')
       goto 9999
@@ -167,27 +172,28 @@ subroutine amg_d_ptap_bld(a_csr,desc_a,nlaggr,parms,ac,&
 
     call coo_restr%transp()
     nzl = coo_restr%get_nzeros()
-    nrl = desc_ac%get_local_rows() 
-    i=0
+    nrl = desc_ac%get_local_rows()
+    call coo_restr%fix(info)
+    i=coo_restr%get_nzeros()
     !
     ! Only keep local rows
     !
-    do k=1, nzl
-      if ((1 <= coo_restr%ia(k)) .and.(coo_restr%ia(k) <= nrl)) then
-        i = i+1
-        coo_restr%val(i) = coo_restr%val(k)
-        coo_restr%ia(i)  = coo_restr%ia(k)
-        coo_restr%ja(i)  = coo_restr%ja(k)
+    search: do k=i,1,-1
+      if (coo_restr%ia(k) <= nrl) then
+        call coo_restr%set_nzeros(k)
+        exit search
       end if
-    end do
-    call coo_restr%set_nzeros(i)
-    call coo_restr%fix(info) 
+    end do search
+
     nzl  = coo_restr%get_nzeros()
     call coo_restr%set_nrows(desc_ac%get_local_rows())
     call coo_restr%set_ncols(desc_a%get_local_cols())
     if (debug) call check_coo(me,trim(name)//' Check 2 on coo_restr:',coo_restr)
+    if (do_timings) call psb_tic(idx_cpytrans2)
+        
     call csr_restr%cp_from_coo(coo_restr,info)
 
+    if (do_timings) call psb_toc(idx_cpytrans2)
     if (info /= psb_success_) then 
       call psb_errpush(psb_err_from_subroutine_,name,a_err='spcnv coo_restr')
       goto 9999

amgprec/impl/aggregator/amg_d_soc1_map_bld.F90

@@ -72,7 +72,9 @@ subroutine amg_d_soc1_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
   use psb_base_mod
   use amg_base_prec_type
   use amg_d_inner_mod
-
+#if defined(OPENMP)
+  use omp_lib
+#endif
   implicit none
 
   ! Arguments
@@ -85,7 +87,7 @@ subroutine amg_d_soc1_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
   integer(psb_ipk_), intent(out)               :: info
 
   ! Local variables
-  integer(psb_ipk_), allocatable :: ils(:), neigh(:), irow(:), icol(:),&
+  integer(psb_ipk_), allocatable :: neigh(:), irow(:), icol(:),&
        & ideg(:), idxs(:)
   integer(psb_lpk_), allocatable :: tmpaggr(:)
   real(psb_dpk_), allocatable  :: val(:), diag(:)
@@ -99,6 +101,9 @@ subroutine amg_d_soc1_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
   integer(psb_ipk_)   :: nrow, ncol, n_ne
   integer(psb_lpk_)   :: nrglob
   character(len=20)   :: name, ch_err
+  integer(psb_ipk_), save :: idx_soc1_p1=-1, idx_soc1_p2=-1, idx_soc1_p3=-1
+  integer(psb_ipk_), save :: idx_soc1_p0=-1
+  logical, parameter      :: do_timings=.true.
 
   info=psb_success_
   name = 'amg_soc1_map_bld'
@@ -114,6 +119,14 @@ subroutine amg_d_soc1_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
   nrow   = desc_a%get_local_rows()
   ncol   = desc_a%get_local_cols()
   nrglob = desc_a%get_global_rows()
+  if ((do_timings).and.(idx_soc1_p0==-1))       &
+       & idx_soc1_p0 = psb_get_timer_idx("SOC1_MAP: phase0")
+  if ((do_timings).and.(idx_soc1_p1==-1))       &
+       & idx_soc1_p1 = psb_get_timer_idx("SOC1_MAP: phase1")
+  if ((do_timings).and.(idx_soc1_p2==-1))       &
+       & idx_soc1_p2 = psb_get_timer_idx("SOC1_MAP: phase2")
+  if ((do_timings).and.(idx_soc1_p3==-1))       &
+       & idx_soc1_p3 = psb_get_timer_idx("SOC1_MAP: phase3")
 
   nr = a%get_nrows()
   nc = a%get_ncols()
@@ -133,41 +146,204 @@ subroutine amg_d_soc1_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
     goto 9999
   end if
 
+  if (do_timings) call psb_tic(idx_soc1_p0)
   call a%cp_to(acsr)
+  if (do_timings) call psb_toc(idx_soc1_p0)
   if (clean_zeros) call acsr%clean_zeros(info)
   if (iorder == amg_aggr_ord_nat_) then 
+    !$omp parallel do private(i) schedule(static)
     do i=1, nr
       ilaggr(i) = -(nr+1)
-      idxs(i)   = i 
+      idxs(i)   = i
     end do
-  else 
+    !$omp end parallel do 
+  else
+    !$omp parallel do private(i) schedule(static)
     do i=1, nr
       ilaggr(i) = -(nr+1)
       ideg(i)   = acsr%irp(i+1) - acsr%irp(i)
     end do
+    !$omp end parallel do 
     call psb_msort(ideg,ix=idxs,dir=psb_sort_down_)
   end if
-
+  if (do_timings) call psb_tic(idx_soc1_p1)
 
   !
   ! Phase one: Start with disjoint groups.
   ! 
   naggr = 0
-  icnt = 0
+#if defined(OPENMP)
+  block
+    integer(psb_ipk_), allocatable  :: bnds(:), locnaggr(:)
+    integer(psb_ipk_) :: myth,nths, kk
+    ! The parallelization makes use of a locaggr(:) array; each thread
+    ! keeps its own version of naggr, and when the loop ends, a prefix is applied
+    ! to locnaggr to determine:
+    ! 1. The total number of aggregaters NAGGR;
+    ! 2. How much should each thread shift its own aggregates
+    ! Part 2 requires to keep track of which thread defined each entry
+    ! of ilaggr(), so that each entry can be adjusted correctly: even
+    ! if an entry I belongs to the range BNDS(TH)>BNDS(TH+1)-1, it may have
+    ! been set because it is strongly connected to an entry J belonging to a
+    ! different thread. 
+
+    !$omp parallel shared(bnds,idxs,locnaggr,ilaggr,nr,naggr,diag,theta,nths,info) &
+    !$omp private(icol,val,myth,kk) 
+    block
+      integer(psb_ipk_) :: ii,nlp,k,kp,n,ia,isz, nc, i,j,m, nz, ilg,  ip, rsz
+      integer(psb_lpk_) :: itmp
+      !$omp master
+      nths = omp_get_num_threads()
+      allocate(bnds(0:nths),locnaggr(0:nths+1))
+      locnaggr(:) = 0
+      bnds(0) = 1
+      !$omp end master      
+      !$omp barrier
+      myth = omp_get_thread_num()
+      rsz = nr/nths
+      if (myth < mod(nr,nths)) rsz = rsz + 1
+      bnds(myth+1) = rsz
+      !$omp barrier
+      !$omp master
+      do i=1,nths
+        bnds(i) = bnds(i) + bnds(i-1)
+      end do
+      info = 0
+      !$omp end master
+      !$omp barrier
+      
+      !$omp do schedule(static) private(disjoint) 
+      do kk=0, nths-1
+        step1: do ii=bnds(kk), bnds(kk+1)-1
+          i = idxs(ii)          
+          if (info /= 0) cycle step1
+          if ((i<1).or.(i>nr)) then
+            !$omp atomic write
+            info=psb_err_internal_error_
+            !$omp end atomic 
+            call psb_errpush(info,name)
+            cycle step1
+            !goto 9999
+          end if
+            
+          if (ilaggr(i) == -(nr+1)) then
+            nz         = (acsr%irp(i+1)-acsr%irp(i))
+            if ((nz<0).or.(nz>size(icol))) then
+              !$omp atomic write
+              info=psb_err_internal_error_
+              !$omp end atomic 
+              call psb_errpush(info,name)
+              cycle step1
+              !goto 9999
+            end if
+
+            icol(1:nz) = acsr%ja(acsr%irp(i):acsr%irp(i+1)-1)
+            val(1:nz)  = acsr%val(acsr%irp(i):acsr%irp(i+1)-1) 
+
+            !
+            ! Build the set of all strongly coupled nodes 
+            !
+            ip = 0
+            do k=1, nz
+              j   = icol(k)
+              ! If any of the neighbours is already assigned,
+              ! we will not reset. 
+              if (j>nr) cycle step1
+              if (ilaggr(j) > 0) cycle step1
+              if (abs(val(k)) > theta*sqrt(abs(diag(i)*diag(j)))) then
+                ip = ip + 1
+                icol(ip) = icol(k)
+              end if
+            enddo
+            
+            !
+            ! If the whole strongly coupled neighborhood of I is
+            ! as yet unconnected, turn it into the next aggregate.
+            ! Same if ip==0 (in which case, neighborhood only
+            ! contains I even if it does not look like it from matrix)
+            ! The fact that DISJOINT is private and not under lock
+            ! generates a certain un-repeatability, in that between
+            ! computing DISJOINT and assigning, another thread might
+            ! alter the values of ILAGGR.
+            ! However, a certain unrepeatability is already present
+            ! because the sequence of aggregates is computed with a
+            ! different order than in serial mode.
+            ! In any case, even if the enteries of ILAGGR may be
+            ! overwritten, the important thing is that each entry is
+            ! consistent and they generate a correct aggregation map.
+            !
+            disjoint = all(ilaggr(icol(1:ip)) == -(nr+1)).or.(ip==0)
+            if (disjoint) then
+              locnaggr(kk)     = locnaggr(kk) + 1
+              itmp = (bnds(kk)-1+locnaggr(kk))*nths+kk
+              if (itmp < (bnds(kk)-1+locnaggr(kk))) then
+                !$omp atomic update
+                info = max(12345678,info)
+                !$omp end atomic
+                cycle step1
+              end if
+              !$omp atomic write
+              ilaggr(i) = itmp
+              !$omp end atomic
+              do k=1, ip
+                !$omp atomic write
+                ilaggr(icol(k)) = itmp
+                !$omp end atomic
+              end do
+            end if
+          end if
+        enddo step1
+      end do
+      !$omp end do
+
+      !$omp master
+      naggr = sum(locnaggr(0:nths-1))
+      do i=1,nths
+        locnaggr(i) = locnaggr(i) + locnaggr(i-1)
+      end do
+      do i=nths+1,1,-1
+        locnaggr(i) = locnaggr(i-1)
+      end do
+      locnaggr(0) = 0
+      !$omp end master 
+      !$omp barrier
+      !$omp  do schedule(static) 
+      do kk=0, nths-1
+        do ii=bnds(kk), bnds(kk+1)-1
+          if (ilaggr(ii) > 0) then 
+            kp = mod(ilaggr(ii),nths)
+            ilaggr(ii) = (ilaggr(ii)/nths)- (bnds(kp)-1) + locnaggr(kp)
+          end if
+        end do
+      end do
+      !$omp end do
+    end block
+    !$omp end parallel
+  end block
+  if (info /= 0) then
+    if (info == 12345678) write(0,*) 'Overflow in encoding ILAGGR'
+    info=psb_err_internal_error_
+    call psb_errpush(info,name)
+    goto 9999
+  end if
+#else
   step1: do ii=1, nr
+    if (info /= 0) cycle
     i = idxs(ii)
     if ((i<1).or.(i>nr)) then
       info=psb_err_internal_error_
       call psb_errpush(info,name)
-      goto 9999
+      cycle step1
+      !goto 9999
     end if
-    
+
     if (ilaggr(i) == -(nr+1)) then
       nz         = (acsr%irp(i+1)-acsr%irp(i))
       if ((nz<0).or.(nz>size(icol))) then
         info=psb_err_internal_error_
         call psb_errpush(info,name)
-        goto 9999
+        cycle step1
+        !goto 9999
       end if
 
       icol(1:nz) = acsr%ja(acsr%irp(i):acsr%irp(i+1)-1)
@@ -176,7 +352,7 @@ subroutine amg_d_soc1_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
       !
       ! Build the set of all strongly coupled nodes 
       !
-      ip = 0 
+      ip = 0
       do k=1, nz
         j   = icol(k)
         if ((1<=j).and.(j<=nr)) then 
@@ -194,8 +370,7 @@ subroutine amg_d_soc1_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
       ! contains I even if it does not look like it from matrix)
       !
       disjoint = all(ilaggr(icol(1:ip)) == -(nr+1)).or.(ip==0)
-      if (disjoint) then 
-        icnt      = icnt + 1 
+      if (disjoint) then       
         naggr     = naggr + 1
         do k=1, ip
           ilaggr(icol(k)) = naggr
@@ -204,16 +379,22 @@ subroutine amg_d_soc1_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
       end if
     endif
   enddo step1
-  
+#endif
   if (debug_level >= psb_debug_outer_) then 
     write(debug_unit,*) me,' ',trim(name),&
-         & ' Check 1:',count(ilaggr == -(nr+1))
+         & ' Check   1:',naggr,count(ilaggr(1:nr) == -(nr+1)), count(ilaggr(1:nr)>0),&
+         & count(ilaggr(1:nr) == -(nr+1))+count(ilaggr(1:nr)>0),nr
   end if
-
+  if (do_timings) call psb_toc(idx_soc1_p1)
+  if (do_timings) call psb_tic(idx_soc1_p2)
   !
   ! Phase two: join the neighbours
   !
+  !$omp workshare
   tmpaggr = ilaggr
+  !$omp end workshare
+  !$omp parallel do schedule(static) shared(tmpaggr,ilaggr,nr,naggr,diag,theta)& 
+  !$omp     private(ii,i,j,k,nz,icol,val,ip,cpling)
   step2: do ii=1,nr
     i = idxs(ii)
 
@@ -244,8 +425,15 @@ subroutine amg_d_soc1_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
       end if
     end if
   end do step2
+  !$omp end parallel do
+  if (do_timings) call psb_toc(idx_soc1_p2)
+  if (debug_level >= psb_debug_outer_) then 
+    write(debug_unit,*) me,' ',trim(name),&
+         & ' Check 1.5:',naggr,count(ilaggr(1:nr) == -(nr+1)), count(ilaggr(1:nr)>0),&
+         & count(ilaggr(1:nr) == -(nr+1))+count(ilaggr(1:nr)>0),nr
+  end if
 
-
+  if (do_timings) call psb_tic(idx_soc1_p3)
   !
   ! Phase three: sweep over leftovers, if any 
   !
@@ -274,7 +462,6 @@ subroutine amg_d_soc1_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
         end if
       enddo
       if (ip > 0) then
-        icnt      = icnt + 1 
         naggr     = naggr + 1
         ilaggr(i) = naggr
         do k=1, ip
@@ -292,7 +479,10 @@ subroutine amg_d_soc1_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
   end do step3
 
   ! Any leftovers?
+  !$omp parallel do schedule(static) shared(ilaggr,info)& 
+  !$omp     private(ii,i,j,k,nz,icol,val,ip)
   do i=1, nr
+    if (info /= 0) cycle
     if (ilaggr(i) < 0) then
       nz = (acsr%irp(i+1)-acsr%irp(i))
       if (nz == 1) then
@@ -303,15 +493,18 @@ subroutine amg_d_soc1_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
         ! other processes. 
         ilaggr(i) = -(nrglob+nr)
       else
+        !$omp atomic write
         info=psb_err_internal_error_
+        !$omp end atomic
         call psb_errpush(info,name,a_err='Fatal error: non-singleton leftovers')
-        goto 9999
+        cycle 
       endif
     end if
   end do
-
+  !$omp end parallel do
+  if (info /= 0) goto 9999
+  if (do_timings) call psb_toc(idx_soc1_p3)
   if (naggr > ncol) then 
-    !write(0,*) name,'Error : naggr > ncol',naggr,ncol
     info=psb_err_internal_error_
     call psb_errpush(info,name,a_err='Fatal error: naggr>ncol')
     goto 9999
@@ -336,9 +529,13 @@ subroutine amg_d_soc1_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
   nlaggr(:) = 0
   nlaggr(me+1) = naggr
   call psb_sum(ctxt,nlaggr(1:np))
+  if (debug_level >= psb_debug_outer_) then 
+    write(debug_unit,*) me,' ',trim(name),&
+         & ' Check   2:',naggr,count(ilaggr(1:nr) == -(nr+1)), count(ilaggr(1:nr)>0),&
+         & count(ilaggr(1:nr) == -(nr+1))+count(ilaggr(1:nr)>0),nr
+  end if
 
   call acsr%free()
-
   call psb_erractionrestore(err_act)
   return
 

amgprec/impl/aggregator/amg_d_soc2_map_bld.F90

@@ -68,9 +68,12 @@
 !
 subroutine amg_d_soc2_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,info)
 
-  use psb_base_mod
+  use psb_base_mod 
   use amg_base_prec_type
   use amg_d_inner_mod
+#if defined(OPENMP)
+  use omp_lib
+#endif
 
   implicit none
 
@@ -99,6 +102,9 @@ subroutine amg_d_soc2_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
   integer(psb_ipk_) :: np, me
   integer(psb_ipk_) :: nrow, ncol, n_ne
   character(len=20)  :: name, ch_err
+  integer(psb_ipk_), save :: idx_soc2_p1=-1, idx_soc2_p2=-1, idx_soc2_p3=-1
+  integer(psb_ipk_), save :: idx_soc2_p0=-1
+  logical, parameter      :: do_timings=.true.
 
   info=psb_success_
   name = 'amg_soc2_map_bld'
@@ -114,6 +120,14 @@ subroutine amg_d_soc2_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
   nrow   = desc_a%get_local_rows()
   ncol   = desc_a%get_local_cols()
   nrglob = desc_a%get_global_rows()
+  if ((do_timings).and.(idx_soc2_p0==-1))       &
+       & idx_soc2_p0 = psb_get_timer_idx("SOC2_MAP: phase0")
+  if ((do_timings).and.(idx_soc2_p1==-1))       &
+       & idx_soc2_p1 = psb_get_timer_idx("SOC2_MAP: phase1")
+  if ((do_timings).and.(idx_soc2_p2==-1))       &
+       & idx_soc2_p2 = psb_get_timer_idx("SOC2_MAP: phase2")
+  if ((do_timings).and.(idx_soc2_p3==-1))       &
+       & idx_soc2_p3 = psb_get_timer_idx("SOC2_MAP: phase3")
 
   nr = a%get_nrows()
   nc = a%get_ncols()
@@ -125,6 +139,7 @@ subroutine amg_d_soc2_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
     goto 9999
   end if
 
+  if (do_timings) call psb_tic(idx_soc2_p0)
   diag = a%get_diag(info)
   if(info /= psb_success_) then
     info=psb_err_from_subroutine_
@@ -137,55 +152,217 @@ subroutine amg_d_soc2_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
   ! 
   call a%cp_to(muij)
   if (clean_zeros) call muij%clean_zeros(info)
+  !$omp parallel do private(i,j,k) shared(nr,diag,muij) schedule(static)
   do i=1, nr
     do k=muij%irp(i),muij%irp(i+1)-1
       j = muij%ja(k)
       if (j<= nr) muij%val(k) = abs(muij%val(k))/sqrt(abs(diag(i)*diag(j)))
     end do
   end do
-
+  !$omp end parallel do 
   !
   ! Compute the 1-neigbour; mark strong links with +1, weak links with -1
   !
   call s_neigh_coo%allocate(nr,nr,muij%get_nzeros())
-  ip = 0 
+  !$omp parallel do private(i,j,k) shared(nr,diag,muij) schedule(static)
   do i=1, nr
     do k=muij%irp(i),muij%irp(i+1)-1
       j = muij%ja(k)
+      s_neigh_coo%ia(k)  = i
+      s_neigh_coo%ja(k)  = j
       if (j<=nr) then 
-        ip = ip + 1
-        s_neigh_coo%ia(ip)  = i
-        s_neigh_coo%ja(ip)  = j
         if (real(muij%val(k)) >= theta) then 
-          s_neigh_coo%val(ip) = done
+          s_neigh_coo%val(k) = done
         else
-          s_neigh_coo%val(ip) = -done
+          s_neigh_coo%val(k) = -done
         end if
+      else
+        s_neigh_coo%val(k) = -done        
       end if
     end do
   end do
+  !$omp end parallel do
   !write(*,*) 'S_NEIGH: ',nr,ip
-  call s_neigh_coo%set_nzeros(ip)
+  call s_neigh_coo%set_nzeros(muij%get_nzeros())
   call s_neigh%mv_from_coo(s_neigh_coo,info)
 
-  if (iorder == amg_aggr_ord_nat_) then 
+  if (iorder == amg_aggr_ord_nat_) then
+    
+    !$omp parallel do private(i) shared(ilaggr,idxs) schedule(static)
     do i=1, nr
       ilaggr(i) = -(nr+1)
       idxs(i)   = i 
     end do
+    !$omp end parallel do 
   else 
+    !$omp parallel do private(i) shared(ilaggr,idxs,muij)  schedule(static)
     do i=1, nr
       ilaggr(i) = -(nr+1)
       ideg(i)   = muij%irp(i+1) - muij%irp(i)
     end do
+    !$omp end parallel do
     call psb_msort(ideg,ix=idxs,dir=psb_sort_down_)
   end if
 
+  if (do_timings) call psb_toc(idx_soc2_p0)
+  if (do_timings) call psb_tic(idx_soc2_p1)
 
   !
   ! Phase one: Start with disjoint groups.
   ! 
   naggr = 0
+#if defined(OPENMP)
+  block
+    integer(psb_ipk_), allocatable  :: bnds(:), locnaggr(:)
+    integer(psb_ipk_) :: myth,nths, kk
+    ! The parallelization makes use of a locaggr(:) array; each thread
+    ! keeps its own version of naggr, and when the loop ends, a prefix is applied
+    ! to locnaggr to determine:
+    ! 1. The total number of aggregaters NAGGR;
+    ! 2. How much should each thread shift its own aggregates
+    ! Part 2 requires to keep track of which thread defined each entry
+    ! of ilaggr(), so that each entry can be adjusted correctly: even
+    ! if an entry I belongs to the range BNDS(TH)>BNDS(TH+1)-1, it may have
+    ! been set because it is strongly connected to an entry J belonging to a
+    ! different thread. 
+
+    !$omp parallel shared(s_neigh,bnds,idxs,locnaggr,ilaggr,nr,naggr,diag,theta,nths,info) &
+    !$omp private(icol,val,myth,kk) 
+    block
+      integer(psb_ipk_) :: ii,nlp,k,kp,n,ia,isz,nc,i,j,m,nz,ilg,ip,rsz,ip1,nzcnt
+      integer(psb_lpk_) :: itmp
+      !$omp master
+      nths = omp_get_num_threads()
+      allocate(bnds(0:nths),locnaggr(0:nths+1))
+      locnaggr(:) = 0
+      bnds(0) = 1
+      !$omp end master      
+      !$omp barrier
+      myth = omp_get_thread_num()
+      rsz = nr/nths
+      if (myth < mod(nr,nths)) rsz = rsz + 1
+      bnds(myth+1) = rsz
+      !$omp barrier
+      !$omp master
+      do i=1,nths
+        bnds(i) = bnds(i) + bnds(i-1)
+      end do
+      info = 0
+      !$omp end master
+      !$omp barrier
+
+      !$omp do schedule(static) private(disjoint) 
+      do kk=0, nths-1
+        step1: do ii=bnds(kk), bnds(kk+1)-1
+          i = idxs(ii)
+          if (info /= 0) then
+            write(0,*) ' Step1:',kk,ii,i,info
+            cycle step1
+          end if
+          if ((i<1).or.(i>nr)) then
+            !$omp atomic write
+            info=psb_err_internal_error_
+            !$omp end atomic 
+            call psb_errpush(info,name)
+            cycle step1
+            !goto 9999
+          end if
+
+
+          if (ilaggr(i) == -(nr+1)) then 
+            !
+            ! Get the 1-neighbourhood of I 
+            !
+            ip1 = s_neigh%irp(i)
+            nz  = s_neigh%irp(i+1)-ip1
+            !
+            ! If the neighbourhood only contains I, skip it
+            !
+            if (nz ==0) then
+              ilaggr(i) = 0
+              cycle step1
+            end if
+            if ((nz==1).and.(s_neigh%ja(ip1)==i)) then
+              ilaggr(i) = 0
+              cycle step1
+            end if
+
+            nzcnt = count(real(s_neigh%val(ip1:ip1+nz-1)) > 0)
+            icol(1:nzcnt) = pack(s_neigh%ja(ip1:ip1+nz-1),(real(s_neigh%val(ip1:ip1+nz-1)) > 0))
+            disjoint = all(ilaggr(icol(1:nzcnt)) == -(nr+1)) 
+
+            !
+            ! If the whole strongly coupled neighborhood of I is
+            ! as yet unconnected, turn it into the next aggregate.
+            ! Same if ip==0 (in which case, neighborhood only
+            ! contains I even if it does not look like it from matrix)
+            ! The fact that DISJOINT is private and not under lock
+            ! generates a certain un-repeatability, in that between
+            ! computing DISJOINT and assigning, another thread might
+            ! alter the values of ILAGGR.
+            ! However, a certain unrepeatability is already present
+            ! because the sequence of aggregates is computed with a
+            ! different order than in serial mode.
+            ! In any case, even if the enteries of ILAGGR may be
+            ! overwritten, the important thing is that each entry is
+            ! consistent and they generate a correct aggregation map.
+            !
+            if (disjoint) then
+              locnaggr(kk)     = locnaggr(kk) + 1
+              itmp = (bnds(kk)-1+locnaggr(kk))*nths+kk
+              if (itmp < (bnds(kk)-1+locnaggr(kk))) then
+                !$omp atomic update
+                info = max(12345678,info)
+                !$omp end atomic
+                cycle step1
+              end if
+              !$omp atomic write
+              ilaggr(i) = itmp
+              !$omp end atomic
+              do k=1, nzcnt
+                !$omp atomic write
+                ilaggr(icol(k)) = itmp
+                !$omp end atomic
+              end do
+            end if
+          end if
+        enddo step1
+      end do
+      !$omp end do
+
+      !$omp master
+      naggr = sum(locnaggr(0:nths-1))
+      do i=1,nths
+        locnaggr(i) = locnaggr(i) + locnaggr(i-1)
+      end do
+      do i=nths+1,1,-1
+        locnaggr(i) = locnaggr(i-1)
+      end do
+      locnaggr(0) = 0
+      !write(0,*) 'LNAG ',locnaggr(nths+1)
+      !$omp end master 
+      !$omp barrier
+      !$omp  do schedule(static) 
+      do kk=0, nths-1
+        do ii=bnds(kk), bnds(kk+1)-1
+          if (ilaggr(ii) > 0) then 
+            kp = mod(ilaggr(ii),nths)
+            ilaggr(ii) = (ilaggr(ii)/nths)- (bnds(kp)-1) + locnaggr(kp)
+          end if
+        end do
+      end do
+      !$omp end do
+    end block
+    !$omp end parallel
+  end block
+  if (info /= 0) then
+    if (info == 12345678) write(0,*) 'Overflow in encoding ILAGGR'
+    info=psb_err_internal_error_
+    call psb_errpush(info,name)
+    goto 9999
+  end if
+
+#else
   icnt = 0
   step1: do ii=1, nr
     i = idxs(ii)
@@ -224,16 +401,21 @@ subroutine amg_d_soc2_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
       end if
     endif
   enddo step1
-  
+#endif  
   if (debug_level >= psb_debug_outer_) then 
     write(debug_unit,*) me,' ',trim(name),&
          & ' Check 1:',count(ilaggr == -(nr+1))
   end if
-
+  if (do_timings) call psb_toc(idx_soc2_p1)
+  if (do_timings) call psb_tic(idx_soc2_p2)
   !
   ! Phase two: join the neighbours
   !
+  !$omp workshare
   tmpaggr = ilaggr
+  !$omp end workshare
+  !$omp parallel do schedule(static) shared(tmpaggr,ilaggr,nr,naggr,diag,muij,s_neigh)& 
+  !$omp     private(ii,i,j,k,nz,icol,val,ip,cpling)
   step2: do ii=1,nr
     i = idxs(ii)
 
@@ -259,8 +441,9 @@ subroutine amg_d_soc2_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
       end if
     end if
   end do step2
-
-
+  !$omp end parallel do
+  if (do_timings) call psb_toc(idx_soc2_p2)
+  if (do_timings) call psb_tic(idx_soc2_p3)
   !
   ! Phase three: sweep over leftovers, if any 
   !
@@ -294,6 +477,8 @@ subroutine amg_d_soc2_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
   end do step3
 
   ! Any leftovers?
+  !$omp parallel do schedule(static) shared(ilaggr,s_neigh,info)& 
+  !$omp     private(ii,i,j,k)
   do i=1, nr
     if (ilaggr(i) <= 0) then
       nz = (s_neigh%irp(i+1)-s_neigh%irp(i))
@@ -305,13 +490,17 @@ subroutine amg_d_soc2_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
         ! other processes. 
         ilaggr(i) = -(nrglob+nr)
       else
+        !$omp atomic write
         info=psb_err_internal_error_
+        !$omp end atomic
         call psb_errpush(info,name,a_err='Fatal error: non-singleton leftovers')
-        goto 9999
+        cycle
       endif
     end if
   end do
-
+  !$omp end parallel do
+  if (info /= 0) goto 9999
+  if (do_timings) call psb_toc(idx_soc2_p3)  
   if (naggr > ncol) then 
     info=psb_err_internal_error_
     call psb_errpush(info,name,a_err='Fatal error: naggr>ncol')

amgprec/impl/aggregator/amg_s_ptap_bld.f90

@@ -76,7 +76,7 @@ subroutine amg_s_ptap_bld(a_csr,desc_a,nlaggr,parms,ac,&
   integer(psb_ipk_) :: nrow, ncol, nrl, nzl, ip, nzt, i, k
   integer(psb_lpk_) ::  nrsave, ncsave, nzsave, nza
   logical, parameter :: do_timings=.false., oldstyle=.false., debug=.false.  
-  integer(psb_ipk_), save :: idx_spspmm=-1
+  integer(psb_ipk_), save :: idx_spspmm=-1, idx_cpytrans1=-1, idx_cpytrans2=-1
 
   name='amg_ptap_bld'
   if(psb_get_errstatus().ne.0) return 
@@ -93,7 +93,11 @@ subroutine amg_s_ptap_bld(a_csr,desc_a,nlaggr,parms,ac,&
   ncol  = desc_a%get_local_cols()
 
   if ((do_timings).and.(idx_spspmm==-1)) &
-       & idx_spspmm = psb_get_timer_idx("SPMM_BLD: par_spspmm")
+       & idx_spspmm = psb_get_timer_idx("PTAP_BLD: par_spspmm")
+  if ((do_timings).and.(idx_cpytrans1==-1)) &
+       & idx_cpytrans1 = psb_get_timer_idx("PTAP_BLD: cpy&trans1")
+  if ((do_timings).and.(idx_cpytrans2==-1)) &
+       & idx_cpytrans2 = psb_get_timer_idx("PTAP_BLD: cpy&trans2")
 
   naggr   = nlaggr(me+1)
   ntaggr  = sum(nlaggr)
@@ -128,6 +132,7 @@ subroutine amg_s_ptap_bld(a_csr,desc_a,nlaggr,parms,ac,&
   ! Ok first product done.
 
   if (present(desc_ax)) then
+    if (do_timings) call psb_tic(idx_cpytrans1)
     block 
       call coo_prol%cp_to_coo(coo_restr,info)
       call coo_restr%set_ncols(desc_ac%get_local_cols())
@@ -137,7 +142,7 @@ subroutine amg_s_ptap_bld(a_csr,desc_a,nlaggr,parms,ac,&
       call coo_restr%set_ncols(desc_ax%get_local_cols())
     end block
     call csr_restr%cp_from_coo(coo_restr,info)
-
+    if (do_timings) call psb_toc(idx_cpytrans1)
     if (info /= psb_success_) then 
       call psb_errpush(psb_err_from_subroutine_,name,a_err='spcnv coo_restr')
       goto 9999
@@ -167,27 +172,28 @@ subroutine amg_s_ptap_bld(a_csr,desc_a,nlaggr,parms,ac,&
 
     call coo_restr%transp()
     nzl = coo_restr%get_nzeros()
-    nrl = desc_ac%get_local_rows() 
-    i=0
+    nrl = desc_ac%get_local_rows()
+    call coo_restr%fix(info)
+    i=coo_restr%get_nzeros()
     !
     ! Only keep local rows
     !
-    do k=1, nzl
-      if ((1 <= coo_restr%ia(k)) .and.(coo_restr%ia(k) <= nrl)) then
-        i = i+1
-        coo_restr%val(i) = coo_restr%val(k)
-        coo_restr%ia(i)  = coo_restr%ia(k)
-        coo_restr%ja(i)  = coo_restr%ja(k)
+    search: do k=i,1,-1
+      if (coo_restr%ia(k) <= nrl) then
+        call coo_restr%set_nzeros(k)
+        exit search
       end if
-    end do
-    call coo_restr%set_nzeros(i)
-    call coo_restr%fix(info) 
+    end do search
+
     nzl  = coo_restr%get_nzeros()
     call coo_restr%set_nrows(desc_ac%get_local_rows())
     call coo_restr%set_ncols(desc_a%get_local_cols())
     if (debug) call check_coo(me,trim(name)//' Check 2 on coo_restr:',coo_restr)
+    if (do_timings) call psb_tic(idx_cpytrans2)
+        
     call csr_restr%cp_from_coo(coo_restr,info)
 
+    if (do_timings) call psb_toc(idx_cpytrans2)
     if (info /= psb_success_) then 
       call psb_errpush(psb_err_from_subroutine_,name,a_err='spcnv coo_restr')
       goto 9999

amgprec/impl/aggregator/amg_s_soc1_map_bld.F90

@@ -72,7 +72,9 @@ subroutine amg_s_soc1_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
   use psb_base_mod
   use amg_base_prec_type
   use amg_s_inner_mod
-
+#if defined(OPENMP)
+  use omp_lib
+#endif
   implicit none
 
   ! Arguments
@@ -85,7 +87,7 @@ subroutine amg_s_soc1_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
   integer(psb_ipk_), intent(out)               :: info
 
   ! Local variables
-  integer(psb_ipk_), allocatable :: ils(:), neigh(:), irow(:), icol(:),&
+  integer(psb_ipk_), allocatable :: neigh(:), irow(:), icol(:),&
        & ideg(:), idxs(:)
   integer(psb_lpk_), allocatable :: tmpaggr(:)
   real(psb_spk_), allocatable  :: val(:), diag(:)
@@ -99,6 +101,9 @@ subroutine amg_s_soc1_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
   integer(psb_ipk_)   :: nrow, ncol, n_ne
   integer(psb_lpk_)   :: nrglob
   character(len=20)   :: name, ch_err
+  integer(psb_ipk_), save :: idx_soc1_p1=-1, idx_soc1_p2=-1, idx_soc1_p3=-1
+  integer(psb_ipk_), save :: idx_soc1_p0=-1
+  logical, parameter      :: do_timings=.true.
 
   info=psb_success_
   name = 'amg_soc1_map_bld'
@@ -114,6 +119,14 @@ subroutine amg_s_soc1_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
   nrow   = desc_a%get_local_rows()
   ncol   = desc_a%get_local_cols()
   nrglob = desc_a%get_global_rows()
+  if ((do_timings).and.(idx_soc1_p0==-1))       &
+       & idx_soc1_p0 = psb_get_timer_idx("SOC1_MAP: phase0")
+  if ((do_timings).and.(idx_soc1_p1==-1))       &
+       & idx_soc1_p1 = psb_get_timer_idx("SOC1_MAP: phase1")
+  if ((do_timings).and.(idx_soc1_p2==-1))       &
+       & idx_soc1_p2 = psb_get_timer_idx("SOC1_MAP: phase2")
+  if ((do_timings).and.(idx_soc1_p3==-1))       &
+       & idx_soc1_p3 = psb_get_timer_idx("SOC1_MAP: phase3")
 
   nr = a%get_nrows()
   nc = a%get_ncols()
@@ -133,41 +146,204 @@ subroutine amg_s_soc1_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
     goto 9999
   end if
 
+  if (do_timings) call psb_tic(idx_soc1_p0)
   call a%cp_to(acsr)
+  if (do_timings) call psb_toc(idx_soc1_p0)
   if (clean_zeros) call acsr%clean_zeros(info)
   if (iorder == amg_aggr_ord_nat_) then 
+    !$omp parallel do private(i) schedule(static)
     do i=1, nr
       ilaggr(i) = -(nr+1)
-      idxs(i)   = i 
+      idxs(i)   = i
     end do
-  else 
+    !$omp end parallel do 
+  else
+    !$omp parallel do private(i) schedule(static)
     do i=1, nr
       ilaggr(i) = -(nr+1)
       ideg(i)   = acsr%irp(i+1) - acsr%irp(i)
     end do
+    !$omp end parallel do 
     call psb_msort(ideg,ix=idxs,dir=psb_sort_down_)
   end if
-
+  if (do_timings) call psb_tic(idx_soc1_p1)
 
   !
   ! Phase one: Start with disjoint groups.
   ! 
   naggr = 0
-  icnt = 0
+#if defined(OPENMP)
+  block
+    integer(psb_ipk_), allocatable  :: bnds(:), locnaggr(:)
+    integer(psb_ipk_) :: myth,nths, kk
+    ! The parallelization makes use of a locaggr(:) array; each thread
+    ! keeps its own version of naggr, and when the loop ends, a prefix is applied
+    ! to locnaggr to determine:
+    ! 1. The total number of aggregaters NAGGR;
+    ! 2. How much should each thread shift its own aggregates
+    ! Part 2 requires to keep track of which thread defined each entry
+    ! of ilaggr(), so that each entry can be adjusted correctly: even
+    ! if an entry I belongs to the range BNDS(TH)>BNDS(TH+1)-1, it may have
+    ! been set because it is strongly connected to an entry J belonging to a
+    ! different thread. 
+
+    !$omp parallel shared(bnds,idxs,locnaggr,ilaggr,nr,naggr,diag,theta,nths,info) &
+    !$omp private(icol,val,myth,kk) 
+    block
+      integer(psb_ipk_) :: ii,nlp,k,kp,n,ia,isz, nc, i,j,m, nz, ilg,  ip, rsz
+      integer(psb_lpk_) :: itmp
+      !$omp master
+      nths = omp_get_num_threads()
+      allocate(bnds(0:nths),locnaggr(0:nths+1))
+      locnaggr(:) = 0
+      bnds(0) = 1
+      !$omp end master      
+      !$omp barrier
+      myth = omp_get_thread_num()
+      rsz = nr/nths
+      if (myth < mod(nr,nths)) rsz = rsz + 1
+      bnds(myth+1) = rsz
+      !$omp barrier
+      !$omp master
+      do i=1,nths
+        bnds(i) = bnds(i) + bnds(i-1)
+      end do
+      info = 0
+      !$omp end master
+      !$omp barrier
+      
+      !$omp do schedule(static) private(disjoint) 
+      do kk=0, nths-1
+        step1: do ii=bnds(kk), bnds(kk+1)-1
+          i = idxs(ii)          
+          if (info /= 0) cycle step1
+          if ((i<1).or.(i>nr)) then
+            !$omp atomic write
+            info=psb_err_internal_error_
+            !$omp end atomic 
+            call psb_errpush(info,name)
+            cycle step1
+            !goto 9999
+          end if
+            
+          if (ilaggr(i) == -(nr+1)) then
+            nz         = (acsr%irp(i+1)-acsr%irp(i))
+            if ((nz<0).or.(nz>size(icol))) then
+              !$omp atomic write
+              info=psb_err_internal_error_
+              !$omp end atomic 
+              call psb_errpush(info,name)
+              cycle step1
+              !goto 9999
+            end if
+
+            icol(1:nz) = acsr%ja(acsr%irp(i):acsr%irp(i+1)-1)
+            val(1:nz)  = acsr%val(acsr%irp(i):acsr%irp(i+1)-1) 
+
+            !
+            ! Build the set of all strongly coupled nodes 
+            !
+            ip = 0
+            do k=1, nz
+              j   = icol(k)
+              ! If any of the neighbours is already assigned,
+              ! we will not reset. 
+              if (j>nr) cycle step1
+              if (ilaggr(j) > 0) cycle step1
+              if (abs(val(k)) > theta*sqrt(abs(diag(i)*diag(j)))) then
+                ip = ip + 1
+                icol(ip) = icol(k)
+              end if
+            enddo
+            
+            !
+            ! If the whole strongly coupled neighborhood of I is
+            ! as yet unconnected, turn it into the next aggregate.
+            ! Same if ip==0 (in which case, neighborhood only
+            ! contains I even if it does not look like it from matrix)
+            ! The fact that DISJOINT is private and not under lock
+            ! generates a certain un-repeatability, in that between
+            ! computing DISJOINT and assigning, another thread might
+            ! alter the values of ILAGGR.
+            ! However, a certain unrepeatability is already present
+            ! because the sequence of aggregates is computed with a
+            ! different order than in serial mode.
+            ! In any case, even if the enteries of ILAGGR may be
+            ! overwritten, the important thing is that each entry is
+            ! consistent and they generate a correct aggregation map.
+            !
+            disjoint = all(ilaggr(icol(1:ip)) == -(nr+1)).or.(ip==0)
+            if (disjoint) then
+              locnaggr(kk)     = locnaggr(kk) + 1
+              itmp = (bnds(kk)-1+locnaggr(kk))*nths+kk
+              if (itmp < (bnds(kk)-1+locnaggr(kk))) then
+                !$omp atomic update
+                info = max(12345678,info)
+                !$omp end atomic
+                cycle step1
+              end if
+              !$omp atomic write
+              ilaggr(i) = itmp
+              !$omp end atomic
+              do k=1, ip
+                !$omp atomic write
+                ilaggr(icol(k)) = itmp
+                !$omp end atomic
+              end do
+            end if
+          end if
+        enddo step1
+      end do
+      !$omp end do
+
+      !$omp master
+      naggr = sum(locnaggr(0:nths-1))
+      do i=1,nths
+        locnaggr(i) = locnaggr(i) + locnaggr(i-1)
+      end do
+      do i=nths+1,1,-1
+        locnaggr(i) = locnaggr(i-1)
+      end do
+      locnaggr(0) = 0
+      !$omp end master 
+      !$omp barrier
+      !$omp  do schedule(static) 
+      do kk=0, nths-1
+        do ii=bnds(kk), bnds(kk+1)-1
+          if (ilaggr(ii) > 0) then 
+            kp = mod(ilaggr(ii),nths)
+            ilaggr(ii) = (ilaggr(ii)/nths)- (bnds(kp)-1) + locnaggr(kp)
+          end if
+        end do
+      end do
+      !$omp end do
+    end block
+    !$omp end parallel
+  end block
+  if (info /= 0) then
+    if (info == 12345678) write(0,*) 'Overflow in encoding ILAGGR'
+    info=psb_err_internal_error_
+    call psb_errpush(info,name)
+    goto 9999
+  end if
+#else
   step1: do ii=1, nr
+    if (info /= 0) cycle
     i = idxs(ii)
     if ((i<1).or.(i>nr)) then
       info=psb_err_internal_error_
       call psb_errpush(info,name)
-      goto 9999
+      cycle step1
+      !goto 9999
     end if
-    
+
     if (ilaggr(i) == -(nr+1)) then
       nz         = (acsr%irp(i+1)-acsr%irp(i))
       if ((nz<0).or.(nz>size(icol))) then
         info=psb_err_internal_error_
         call psb_errpush(info,name)
-        goto 9999
+        cycle step1
+        !goto 9999
       end if
 
       icol(1:nz) = acsr%ja(acsr%irp(i):acsr%irp(i+1)-1)
@@ -176,7 +352,7 @@ subroutine amg_s_soc1_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
       !
       ! Build the set of all strongly coupled nodes 
       !
-      ip = 0 
+      ip = 0
       do k=1, nz
         j   = icol(k)
         if ((1<=j).and.(j<=nr)) then 
@@ -194,8 +370,7 @@ subroutine amg_s_soc1_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
       ! contains I even if it does not look like it from matrix)
       !
       disjoint = all(ilaggr(icol(1:ip)) == -(nr+1)).or.(ip==0)
-      if (disjoint) then 
-        icnt      = icnt + 1 
+      if (disjoint) then       
         naggr     = naggr + 1
         do k=1, ip
           ilaggr(icol(k)) = naggr
@@ -204,16 +379,22 @@ subroutine amg_s_soc1_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
       end if
     endif
   enddo step1
-  
+#endif
   if (debug_level >= psb_debug_outer_) then 
     write(debug_unit,*) me,' ',trim(name),&
-         & ' Check 1:',count(ilaggr == -(nr+1))
+         & ' Check   1:',naggr,count(ilaggr(1:nr) == -(nr+1)), count(ilaggr(1:nr)>0),&
+         & count(ilaggr(1:nr) == -(nr+1))+count(ilaggr(1:nr)>0),nr
   end if
-
+  if (do_timings) call psb_toc(idx_soc1_p1)
+  if (do_timings) call psb_tic(idx_soc1_p2)
   !
   ! Phase two: join the neighbours
   !
+  !$omp workshare
   tmpaggr = ilaggr
+  !$omp end workshare
+  !$omp parallel do schedule(static) shared(tmpaggr,ilaggr,nr,naggr,diag,theta)& 
+  !$omp     private(ii,i,j,k,nz,icol,val,ip,cpling)
   step2: do ii=1,nr
     i = idxs(ii)
 
@@ -244,8 +425,15 @@ subroutine amg_s_soc1_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
       end if
     end if
   end do step2
+  !$omp end parallel do
+  if (do_timings) call psb_toc(idx_soc1_p2)
+  if (debug_level >= psb_debug_outer_) then 
+    write(debug_unit,*) me,' ',trim(name),&
+         & ' Check 1.5:',naggr,count(ilaggr(1:nr) == -(nr+1)), count(ilaggr(1:nr)>0),&
+         & count(ilaggr(1:nr) == -(nr+1))+count(ilaggr(1:nr)>0),nr
+  end if
 
-
+  if (do_timings) call psb_tic(idx_soc1_p3)
   !
   ! Phase three: sweep over leftovers, if any 
   !
@@ -274,7 +462,6 @@ subroutine amg_s_soc1_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
         end if
       enddo
       if (ip > 0) then
-        icnt      = icnt + 1 
         naggr     = naggr + 1
         ilaggr(i) = naggr
         do k=1, ip
@@ -292,7 +479,10 @@ subroutine amg_s_soc1_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
   end do step3
 
   ! Any leftovers?
+  !$omp parallel do schedule(static) shared(ilaggr,info)& 
+  !$omp     private(ii,i,j,k,nz,icol,val,ip)
   do i=1, nr
+    if (info /= 0) cycle
     if (ilaggr(i) < 0) then
       nz = (acsr%irp(i+1)-acsr%irp(i))
       if (nz == 1) then
@@ -303,15 +493,18 @@ subroutine amg_s_soc1_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
         ! other processes. 
         ilaggr(i) = -(nrglob+nr)
       else
+        !$omp atomic write
         info=psb_err_internal_error_
+        !$omp end atomic
         call psb_errpush(info,name,a_err='Fatal error: non-singleton leftovers')
-        goto 9999
+        cycle 
       endif
     end if
   end do
-
+  !$omp end parallel do
+  if (info /= 0) goto 9999
+  if (do_timings) call psb_toc(idx_soc1_p3)
   if (naggr > ncol) then 
-    !write(0,*) name,'Error : naggr > ncol',naggr,ncol
     info=psb_err_internal_error_
     call psb_errpush(info,name,a_err='Fatal error: naggr>ncol')
     goto 9999
@@ -336,9 +529,13 @@ subroutine amg_s_soc1_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
   nlaggr(:) = 0
   nlaggr(me+1) = naggr
   call psb_sum(ctxt,nlaggr(1:np))
+  if (debug_level >= psb_debug_outer_) then 
+    write(debug_unit,*) me,' ',trim(name),&
+         & ' Check   2:',naggr,count(ilaggr(1:nr) == -(nr+1)), count(ilaggr(1:nr)>0),&
+         & count(ilaggr(1:nr) == -(nr+1))+count(ilaggr(1:nr)>0),nr
+  end if
 
   call acsr%free()
-
   call psb_erractionrestore(err_act)
   return
 

amgprec/impl/aggregator/amg_s_soc2_map_bld.F90

@@ -68,9 +68,12 @@
 !
 subroutine amg_s_soc2_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,info)
 
-  use psb_base_mod
+  use psb_base_mod 
   use amg_base_prec_type
   use amg_s_inner_mod
+#if defined(OPENMP)
+  use omp_lib
+#endif
 
   implicit none
 
@@ -99,6 +102,9 @@ subroutine amg_s_soc2_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
   integer(psb_ipk_) :: np, me
   integer(psb_ipk_) :: nrow, ncol, n_ne
   character(len=20)  :: name, ch_err
+  integer(psb_ipk_), save :: idx_soc2_p1=-1, idx_soc2_p2=-1, idx_soc2_p3=-1
+  integer(psb_ipk_), save :: idx_soc2_p0=-1
+  logical, parameter      :: do_timings=.true.
 
   info=psb_success_
   name = 'amg_soc2_map_bld'
@@ -114,6 +120,14 @@ subroutine amg_s_soc2_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
   nrow   = desc_a%get_local_rows()
   ncol   = desc_a%get_local_cols()
   nrglob = desc_a%get_global_rows()
+  if ((do_timings).and.(idx_soc2_p0==-1))       &
+       & idx_soc2_p0 = psb_get_timer_idx("SOC2_MAP: phase0")
+  if ((do_timings).and.(idx_soc2_p1==-1))       &
+       & idx_soc2_p1 = psb_get_timer_idx("SOC2_MAP: phase1")
+  if ((do_timings).and.(idx_soc2_p2==-1))       &
+       & idx_soc2_p2 = psb_get_timer_idx("SOC2_MAP: phase2")
+  if ((do_timings).and.(idx_soc2_p3==-1))       &
+       & idx_soc2_p3 = psb_get_timer_idx("SOC2_MAP: phase3")
 
   nr = a%get_nrows()
   nc = a%get_ncols()
@@ -125,6 +139,7 @@ subroutine amg_s_soc2_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
     goto 9999
   end if
 
+  if (do_timings) call psb_tic(idx_soc2_p0)
   diag = a%get_diag(info)
   if(info /= psb_success_) then
     info=psb_err_from_subroutine_
@@ -137,55 +152,217 @@ subroutine amg_s_soc2_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
   ! 
   call a%cp_to(muij)
   if (clean_zeros) call muij%clean_zeros(info)
+  !$omp parallel do private(i,j,k) shared(nr,diag,muij) schedule(static)
   do i=1, nr
     do k=muij%irp(i),muij%irp(i+1)-1
       j = muij%ja(k)
       if (j<= nr) muij%val(k) = abs(muij%val(k))/sqrt(abs(diag(i)*diag(j)))
     end do
   end do
-
+  !$omp end parallel do 
   !
   ! Compute the 1-neigbour; mark strong links with +1, weak links with -1
   !
   call s_neigh_coo%allocate(nr,nr,muij%get_nzeros())
-  ip = 0 
+  !$omp parallel do private(i,j,k) shared(nr,diag,muij) schedule(static)
   do i=1, nr
     do k=muij%irp(i),muij%irp(i+1)-1
       j = muij%ja(k)
+      s_neigh_coo%ia(k)  = i
+      s_neigh_coo%ja(k)  = j
       if (j<=nr) then 
-        ip = ip + 1
-        s_neigh_coo%ia(ip)  = i
-        s_neigh_coo%ja(ip)  = j
         if (real(muij%val(k)) >= theta) then 
-          s_neigh_coo%val(ip) = sone
+          s_neigh_coo%val(k) = sone
         else
-          s_neigh_coo%val(ip) = -sone
+          s_neigh_coo%val(k) = -sone
         end if
+      else
+        s_neigh_coo%val(k) = -sone        
       end if
     end do
   end do
+  !$omp end parallel do
   !write(*,*) 'S_NEIGH: ',nr,ip
-  call s_neigh_coo%set_nzeros(ip)
+  call s_neigh_coo%set_nzeros(muij%get_nzeros())
   call s_neigh%mv_from_coo(s_neigh_coo,info)
 
-  if (iorder == amg_aggr_ord_nat_) then 
+  if (iorder == amg_aggr_ord_nat_) then
+    
+    !$omp parallel do private(i) shared(ilaggr,idxs) schedule(static)
     do i=1, nr
       ilaggr(i) = -(nr+1)
       idxs(i)   = i 
     end do
+    !$omp end parallel do 
   else 
+    !$omp parallel do private(i) shared(ilaggr,idxs,muij)  schedule(static)
     do i=1, nr
       ilaggr(i) = -(nr+1)
       ideg(i)   = muij%irp(i+1) - muij%irp(i)
     end do
+    !$omp end parallel do
     call psb_msort(ideg,ix=idxs,dir=psb_sort_down_)
   end if
 
+  if (do_timings) call psb_toc(idx_soc2_p0)
+  if (do_timings) call psb_tic(idx_soc2_p1)
 
   !
   ! Phase one: Start with disjoint groups.
   ! 
   naggr = 0
+#if defined(OPENMP)
+  block
+    integer(psb_ipk_), allocatable  :: bnds(:), locnaggr(:)
+    integer(psb_ipk_) :: myth,nths, kk
+    ! The parallelization makes use of a locaggr(:) array; each thread
+    ! keeps its own version of naggr, and when the loop ends, a prefix is applied
+    ! to locnaggr to determine:
+    ! 1. The total number of aggregaters NAGGR;
+    ! 2. How much should each thread shift its own aggregates
+    ! Part 2 requires to keep track of which thread defined each entry
+    ! of ilaggr(), so that each entry can be adjusted correctly: even
+    ! if an entry I belongs to the range BNDS(TH)>BNDS(TH+1)-1, it may have
+    ! been set because it is strongly connected to an entry J belonging to a
+    ! different thread. 
+
+    !$omp parallel shared(s_neigh,bnds,idxs,locnaggr,ilaggr,nr,naggr,diag,theta,nths,info) &
+    !$omp private(icol,val,myth,kk) 
+    block
+      integer(psb_ipk_) :: ii,nlp,k,kp,n,ia,isz,nc,i,j,m,nz,ilg,ip,rsz,ip1,nzcnt
+      integer(psb_lpk_) :: itmp
+      !$omp master
+      nths = omp_get_num_threads()
+      allocate(bnds(0:nths),locnaggr(0:nths+1))
+      locnaggr(:) = 0
+      bnds(0) = 1
+      !$omp end master      
+      !$omp barrier
+      myth = omp_get_thread_num()
+      rsz = nr/nths
+      if (myth < mod(nr,nths)) rsz = rsz + 1
+      bnds(myth+1) = rsz
+      !$omp barrier
+      !$omp master
+      do i=1,nths
+        bnds(i) = bnds(i) + bnds(i-1)
+      end do
+      info = 0
+      !$omp end master
+      !$omp barrier
+
+      !$omp do schedule(static) private(disjoint) 
+      do kk=0, nths-1
+        step1: do ii=bnds(kk), bnds(kk+1)-1
+          i = idxs(ii)
+          if (info /= 0) then
+            write(0,*) ' Step1:',kk,ii,i,info
+            cycle step1
+          end if
+          if ((i<1).or.(i>nr)) then
+            !$omp atomic write
+            info=psb_err_internal_error_
+            !$omp end atomic 
+            call psb_errpush(info,name)
+            cycle step1
+            !goto 9999
+          end if
+
+
+          if (ilaggr(i) == -(nr+1)) then 
+            !
+            ! Get the 1-neighbourhood of I 
+            !
+            ip1 = s_neigh%irp(i)
+            nz  = s_neigh%irp(i+1)-ip1
+            !
+            ! If the neighbourhood only contains I, skip it
+            !
+            if (nz ==0) then
+              ilaggr(i) = 0
+              cycle step1
+            end if
+            if ((nz==1).and.(s_neigh%ja(ip1)==i)) then
+              ilaggr(i) = 0
+              cycle step1
+            end if
+
+            nzcnt = count(real(s_neigh%val(ip1:ip1+nz-1)) > 0)
+            icol(1:nzcnt) = pack(s_neigh%ja(ip1:ip1+nz-1),(real(s_neigh%val(ip1:ip1+nz-1)) > 0))
+            disjoint = all(ilaggr(icol(1:nzcnt)) == -(nr+1)) 
+
+            !
+            ! If the whole strongly coupled neighborhood of I is
+            ! as yet unconnected, turn it into the next aggregate.
+            ! Same if ip==0 (in which case, neighborhood only
+            ! contains I even if it does not look like it from matrix)
+            ! The fact that DISJOINT is private and not under lock
+            ! generates a certain un-repeatability, in that between
+            ! computing DISJOINT and assigning, another thread might
+            ! alter the values of ILAGGR.
+            ! However, a certain unrepeatability is already present
+            ! because the sequence of aggregates is computed with a
+            ! different order than in serial mode.
+            ! In any case, even if the enteries of ILAGGR may be
+            ! overwritten, the important thing is that each entry is
+            ! consistent and they generate a correct aggregation map.
+            !
+            if (disjoint) then
+              locnaggr(kk)     = locnaggr(kk) + 1
+              itmp = (bnds(kk)-1+locnaggr(kk))*nths+kk
+              if (itmp < (bnds(kk)-1+locnaggr(kk))) then
+                !$omp atomic update
+                info = max(12345678,info)
+                !$omp end atomic
+                cycle step1
+              end if
+              !$omp atomic write
+              ilaggr(i) = itmp
+              !$omp end atomic
+              do k=1, nzcnt
+                !$omp atomic write
+                ilaggr(icol(k)) = itmp
+                !$omp end atomic
+              end do
+            end if
+          end if
+        enddo step1
+      end do
+      !$omp end do
+
+      !$omp master
+      naggr = sum(locnaggr(0:nths-1))
+      do i=1,nths
+        locnaggr(i) = locnaggr(i) + locnaggr(i-1)
+      end do
+      do i=nths+1,1,-1
+        locnaggr(i) = locnaggr(i-1)
+      end do
+      locnaggr(0) = 0
+      !write(0,*) 'LNAG ',locnaggr(nths+1)
+      !$omp end master 
+      !$omp barrier
+      !$omp  do schedule(static) 
+      do kk=0, nths-1
+        do ii=bnds(kk), bnds(kk+1)-1
+          if (ilaggr(ii) > 0) then 
+            kp = mod(ilaggr(ii),nths)
+            ilaggr(ii) = (ilaggr(ii)/nths)- (bnds(kp)-1) + locnaggr(kp)
+          end if
+        end do
+      end do
+      !$omp end do
+    end block
+    !$omp end parallel
+  end block
+  if (info /= 0) then
+    if (info == 12345678) write(0,*) 'Overflow in encoding ILAGGR'
+    info=psb_err_internal_error_
+    call psb_errpush(info,name)
+    goto 9999
+  end if
+
+#else
   icnt = 0
   step1: do ii=1, nr
     i = idxs(ii)
@@ -224,16 +401,21 @@ subroutine amg_s_soc2_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
       end if
     endif
   enddo step1
-  
+#endif  
   if (debug_level >= psb_debug_outer_) then 
     write(debug_unit,*) me,' ',trim(name),&
          & ' Check 1:',count(ilaggr == -(nr+1))
   end if
-
+  if (do_timings) call psb_toc(idx_soc2_p1)
+  if (do_timings) call psb_tic(idx_soc2_p2)
   !
   ! Phase two: join the neighbours
   !
+  !$omp workshare
   tmpaggr = ilaggr
+  !$omp end workshare
+  !$omp parallel do schedule(static) shared(tmpaggr,ilaggr,nr,naggr,diag,muij,s_neigh)& 
+  !$omp     private(ii,i,j,k,nz,icol,val,ip,cpling)
   step2: do ii=1,nr
     i = idxs(ii)
 
@@ -259,8 +441,9 @@ subroutine amg_s_soc2_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
       end if
     end if
   end do step2
-
-
+  !$omp end parallel do
+  if (do_timings) call psb_toc(idx_soc2_p2)
+  if (do_timings) call psb_tic(idx_soc2_p3)
   !
   ! Phase three: sweep over leftovers, if any 
   !
@@ -294,6 +477,8 @@ subroutine amg_s_soc2_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
   end do step3
 
   ! Any leftovers?
+  !$omp parallel do schedule(static) shared(ilaggr,s_neigh,info)& 
+  !$omp     private(ii,i,j,k)
   do i=1, nr
     if (ilaggr(i) <= 0) then
       nz = (s_neigh%irp(i+1)-s_neigh%irp(i))
@@ -305,13 +490,17 @@ subroutine amg_s_soc2_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
         ! other processes. 
         ilaggr(i) = -(nrglob+nr)
       else
+        !$omp atomic write
         info=psb_err_internal_error_
+        !$omp end atomic
         call psb_errpush(info,name,a_err='Fatal error: non-singleton leftovers')
-        goto 9999
+        cycle
       endif
     end if
   end do
-
+  !$omp end parallel do
+  if (info /= 0) goto 9999
+  if (do_timings) call psb_toc(idx_soc2_p3)  
   if (naggr > ncol) then 
     info=psb_err_internal_error_
     call psb_errpush(info,name,a_err='Fatal error: naggr>ncol')

amgprec/impl/aggregator/amg_z_ptap_bld.f90

@@ -76,7 +76,7 @@ subroutine amg_z_ptap_bld(a_csr,desc_a,nlaggr,parms,ac,&
   integer(psb_ipk_) :: nrow, ncol, nrl, nzl, ip, nzt, i, k
   integer(psb_lpk_) ::  nrsave, ncsave, nzsave, nza
   logical, parameter :: do_timings=.false., oldstyle=.false., debug=.false.  
-  integer(psb_ipk_), save :: idx_spspmm=-1
+  integer(psb_ipk_), save :: idx_spspmm=-1, idx_cpytrans1=-1, idx_cpytrans2=-1
 
   name='amg_ptap_bld'
   if(psb_get_errstatus().ne.0) return 
@@ -93,7 +93,11 @@ subroutine amg_z_ptap_bld(a_csr,desc_a,nlaggr,parms,ac,&
   ncol  = desc_a%get_local_cols()
 
   if ((do_timings).and.(idx_spspmm==-1)) &
-       & idx_spspmm = psb_get_timer_idx("SPMM_BLD: par_spspmm")
+       & idx_spspmm = psb_get_timer_idx("PTAP_BLD: par_spspmm")
+  if ((do_timings).and.(idx_cpytrans1==-1)) &
+       & idx_cpytrans1 = psb_get_timer_idx("PTAP_BLD: cpy&trans1")
+  if ((do_timings).and.(idx_cpytrans2==-1)) &
+       & idx_cpytrans2 = psb_get_timer_idx("PTAP_BLD: cpy&trans2")
 
   naggr   = nlaggr(me+1)
   ntaggr  = sum(nlaggr)
@@ -128,6 +132,7 @@ subroutine amg_z_ptap_bld(a_csr,desc_a,nlaggr,parms,ac,&
   ! Ok first product done.
 
   if (present(desc_ax)) then
+    if (do_timings) call psb_tic(idx_cpytrans1)
     block 
       call coo_prol%cp_to_coo(coo_restr,info)
       call coo_restr%set_ncols(desc_ac%get_local_cols())
@@ -137,7 +142,7 @@ subroutine amg_z_ptap_bld(a_csr,desc_a,nlaggr,parms,ac,&
       call coo_restr%set_ncols(desc_ax%get_local_cols())
     end block
     call csr_restr%cp_from_coo(coo_restr,info)
-
+    if (do_timings) call psb_toc(idx_cpytrans1)
     if (info /= psb_success_) then 
       call psb_errpush(psb_err_from_subroutine_,name,a_err='spcnv coo_restr')
       goto 9999
@@ -167,27 +172,28 @@ subroutine amg_z_ptap_bld(a_csr,desc_a,nlaggr,parms,ac,&
 
     call coo_restr%transp()
     nzl = coo_restr%get_nzeros()
-    nrl = desc_ac%get_local_rows() 
-    i=0
+    nrl = desc_ac%get_local_rows()
+    call coo_restr%fix(info)
+    i=coo_restr%get_nzeros()
     !
     ! Only keep local rows
     !
-    do k=1, nzl
-      if ((1 <= coo_restr%ia(k)) .and.(coo_restr%ia(k) <= nrl)) then
-        i = i+1
-        coo_restr%val(i) = coo_restr%val(k)
-        coo_restr%ia(i)  = coo_restr%ia(k)
-        coo_restr%ja(i)  = coo_restr%ja(k)
+    search: do k=i,1,-1
+      if (coo_restr%ia(k) <= nrl) then
+        call coo_restr%set_nzeros(k)
+        exit search
       end if
-    end do
-    call coo_restr%set_nzeros(i)
-    call coo_restr%fix(info) 
+    end do search
+
     nzl  = coo_restr%get_nzeros()
     call coo_restr%set_nrows(desc_ac%get_local_rows())
     call coo_restr%set_ncols(desc_a%get_local_cols())
     if (debug) call check_coo(me,trim(name)//' Check 2 on coo_restr:',coo_restr)
+    if (do_timings) call psb_tic(idx_cpytrans2)
+        
     call csr_restr%cp_from_coo(coo_restr,info)
 
+    if (do_timings) call psb_toc(idx_cpytrans2)
     if (info /= psb_success_) then 
       call psb_errpush(psb_err_from_subroutine_,name,a_err='spcnv coo_restr')
       goto 9999

amgprec/impl/aggregator/amg_z_soc1_map_bld.F90

@@ -72,7 +72,9 @@ subroutine amg_z_soc1_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
   use psb_base_mod
   use amg_base_prec_type
   use amg_z_inner_mod
-
+#if defined(OPENMP)
+  use omp_lib
+#endif
   implicit none
 
   ! Arguments
@@ -85,7 +87,7 @@ subroutine amg_z_soc1_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
   integer(psb_ipk_), intent(out)               :: info
 
   ! Local variables
-  integer(psb_ipk_), allocatable :: ils(:), neigh(:), irow(:), icol(:),&
+  integer(psb_ipk_), allocatable :: neigh(:), irow(:), icol(:),&
        & ideg(:), idxs(:)
   integer(psb_lpk_), allocatable :: tmpaggr(:)
   complex(psb_dpk_), allocatable  :: val(:), diag(:)
@@ -99,6 +101,9 @@ subroutine amg_z_soc1_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
   integer(psb_ipk_)   :: nrow, ncol, n_ne
   integer(psb_lpk_)   :: nrglob
   character(len=20)   :: name, ch_err
+  integer(psb_ipk_), save :: idx_soc1_p1=-1, idx_soc1_p2=-1, idx_soc1_p3=-1
+  integer(psb_ipk_), save :: idx_soc1_p0=-1
+  logical, parameter      :: do_timings=.true.
 
   info=psb_success_
   name = 'amg_soc1_map_bld'
@@ -114,6 +119,14 @@ subroutine amg_z_soc1_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
   nrow   = desc_a%get_local_rows()
   ncol   = desc_a%get_local_cols()
   nrglob = desc_a%get_global_rows()
+  if ((do_timings).and.(idx_soc1_p0==-1))       &
+       & idx_soc1_p0 = psb_get_timer_idx("SOC1_MAP: phase0")
+  if ((do_timings).and.(idx_soc1_p1==-1))       &
+       & idx_soc1_p1 = psb_get_timer_idx("SOC1_MAP: phase1")
+  if ((do_timings).and.(idx_soc1_p2==-1))       &
+       & idx_soc1_p2 = psb_get_timer_idx("SOC1_MAP: phase2")
+  if ((do_timings).and.(idx_soc1_p3==-1))       &
+       & idx_soc1_p3 = psb_get_timer_idx("SOC1_MAP: phase3")
 
   nr = a%get_nrows()
   nc = a%get_ncols()
@@ -133,41 +146,204 @@ subroutine amg_z_soc1_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
     goto 9999
   end if
 
+  if (do_timings) call psb_tic(idx_soc1_p0)
   call a%cp_to(acsr)
+  if (do_timings) call psb_toc(idx_soc1_p0)
   if (clean_zeros) call acsr%clean_zeros(info)
   if (iorder == amg_aggr_ord_nat_) then 
+    !$omp parallel do private(i) schedule(static)
     do i=1, nr
       ilaggr(i) = -(nr+1)
-      idxs(i)   = i 
+      idxs(i)   = i
     end do
-  else 
+    !$omp end parallel do 
+  else
+    !$omp parallel do private(i) schedule(static)
     do i=1, nr
       ilaggr(i) = -(nr+1)
       ideg(i)   = acsr%irp(i+1) - acsr%irp(i)
     end do
+    !$omp end parallel do 
     call psb_msort(ideg,ix=idxs,dir=psb_sort_down_)
   end if
-
+  if (do_timings) call psb_tic(idx_soc1_p1)
 
   !
   ! Phase one: Start with disjoint groups.
   ! 
   naggr = 0
-  icnt = 0
+#if defined(OPENMP)
+  block
+    integer(psb_ipk_), allocatable  :: bnds(:), locnaggr(:)
+    integer(psb_ipk_) :: myth,nths, kk
+    ! The parallelization makes use of a locaggr(:) array; each thread
+    ! keeps its own version of naggr, and when the loop ends, a prefix is applied
+    ! to locnaggr to determine:
+    ! 1. The total number of aggregaters NAGGR;
+    ! 2. How much should each thread shift its own aggregates
+    ! Part 2 requires to keep track of which thread defined each entry
+    ! of ilaggr(), so that each entry can be adjusted correctly: even
+    ! if an entry I belongs to the range BNDS(TH)>BNDS(TH+1)-1, it may have
+    ! been set because it is strongly connected to an entry J belonging to a
+    ! different thread. 
+
+    !$omp parallel shared(bnds,idxs,locnaggr,ilaggr,nr,naggr,diag,theta,nths,info) &
+    !$omp private(icol,val,myth,kk) 
+    block
+      integer(psb_ipk_) :: ii,nlp,k,kp,n,ia,isz, nc, i,j,m, nz, ilg,  ip, rsz
+      integer(psb_lpk_) :: itmp
+      !$omp master
+      nths = omp_get_num_threads()
+      allocate(bnds(0:nths),locnaggr(0:nths+1))
+      locnaggr(:) = 0
+      bnds(0) = 1
+      !$omp end master      
+      !$omp barrier
+      myth = omp_get_thread_num()
+      rsz = nr/nths
+      if (myth < mod(nr,nths)) rsz = rsz + 1
+      bnds(myth+1) = rsz
+      !$omp barrier
+      !$omp master
+      do i=1,nths
+        bnds(i) = bnds(i) + bnds(i-1)
+      end do
+      info = 0
+      !$omp end master
+      !$omp barrier
+      
+      !$omp do schedule(static) private(disjoint) 
+      do kk=0, nths-1
+        step1: do ii=bnds(kk), bnds(kk+1)-1
+          i = idxs(ii)          
+          if (info /= 0) cycle step1
+          if ((i<1).or.(i>nr)) then
+            !$omp atomic write
+            info=psb_err_internal_error_
+            !$omp end atomic 
+            call psb_errpush(info,name)
+            cycle step1
+            !goto 9999
+          end if
+            
+          if (ilaggr(i) == -(nr+1)) then
+            nz         = (acsr%irp(i+1)-acsr%irp(i))
+            if ((nz<0).or.(nz>size(icol))) then
+              !$omp atomic write
+              info=psb_err_internal_error_
+              !$omp end atomic 
+              call psb_errpush(info,name)
+              cycle step1
+              !goto 9999
+            end if
+
+            icol(1:nz) = acsr%ja(acsr%irp(i):acsr%irp(i+1)-1)
+            val(1:nz)  = acsr%val(acsr%irp(i):acsr%irp(i+1)-1) 
+
+            !
+            ! Build the set of all strongly coupled nodes 
+            !
+            ip = 0
+            do k=1, nz
+              j   = icol(k)
+              ! If any of the neighbours is already assigned,
+              ! we will not reset. 
+              if (j>nr) cycle step1
+              if (ilaggr(j) > 0) cycle step1
+              if (abs(val(k)) > theta*sqrt(abs(diag(i)*diag(j)))) then
+                ip = ip + 1
+                icol(ip) = icol(k)
+              end if
+            enddo
+            
+            !
+            ! If the whole strongly coupled neighborhood of I is
+            ! as yet unconnected, turn it into the next aggregate.
+            ! Same if ip==0 (in which case, neighborhood only
+            ! contains I even if it does not look like it from matrix)
+            ! The fact that DISJOINT is private and not under lock
+            ! generates a certain un-repeatability, in that between
+            ! computing DISJOINT and assigning, another thread might
+            ! alter the values of ILAGGR.
+            ! However, a certain unrepeatability is already present
+            ! because the sequence of aggregates is computed with a
+            ! different order than in serial mode.
+            ! In any case, even if the enteries of ILAGGR may be
+            ! overwritten, the important thing is that each entry is
+            ! consistent and they generate a correct aggregation map.
+            !
+            disjoint = all(ilaggr(icol(1:ip)) == -(nr+1)).or.(ip==0)
+            if (disjoint) then
+              locnaggr(kk)     = locnaggr(kk) + 1
+              itmp = (bnds(kk)-1+locnaggr(kk))*nths+kk
+              if (itmp < (bnds(kk)-1+locnaggr(kk))) then
+                !$omp atomic update
+                info = max(12345678,info)
+                !$omp end atomic
+                cycle step1
+              end if
+              !$omp atomic write
+              ilaggr(i) = itmp
+              !$omp end atomic
+              do k=1, ip
+                !$omp atomic write
+                ilaggr(icol(k)) = itmp
+                !$omp end atomic
+              end do
+            end if
+          end if
+        enddo step1
+      end do
+      !$omp end do
+
+      !$omp master
+      naggr = sum(locnaggr(0:nths-1))
+      do i=1,nths
+        locnaggr(i) = locnaggr(i) + locnaggr(i-1)
+      end do
+      do i=nths+1,1,-1
+        locnaggr(i) = locnaggr(i-1)
+      end do
+      locnaggr(0) = 0
+      !$omp end master 
+      !$omp barrier
+      !$omp  do schedule(static) 
+      do kk=0, nths-1
+        do ii=bnds(kk), bnds(kk+1)-1
+          if (ilaggr(ii) > 0) then 
+            kp = mod(ilaggr(ii),nths)
+            ilaggr(ii) = (ilaggr(ii)/nths)- (bnds(kp)-1) + locnaggr(kp)
+          end if
+        end do
+      end do
+      !$omp end do
+    end block
+    !$omp end parallel
+  end block
+  if (info /= 0) then
+    if (info == 12345678) write(0,*) 'Overflow in encoding ILAGGR'
+    info=psb_err_internal_error_
+    call psb_errpush(info,name)
+    goto 9999
+  end if
+#else
   step1: do ii=1, nr
+    if (info /= 0) cycle
     i = idxs(ii)
     if ((i<1).or.(i>nr)) then
       info=psb_err_internal_error_
       call psb_errpush(info,name)
-      goto 9999
+      cycle step1
+      !goto 9999
     end if
-    
+
     if (ilaggr(i) == -(nr+1)) then
       nz         = (acsr%irp(i+1)-acsr%irp(i))
       if ((nz<0).or.(nz>size(icol))) then
         info=psb_err_internal_error_
         call psb_errpush(info,name)
-        goto 9999
+        cycle step1
+        !goto 9999
       end if
 
       icol(1:nz) = acsr%ja(acsr%irp(i):acsr%irp(i+1)-1)
@@ -176,7 +352,7 @@ subroutine amg_z_soc1_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
       !
       ! Build the set of all strongly coupled nodes 
       !
-      ip = 0 
+      ip = 0
       do k=1, nz
         j   = icol(k)
         if ((1<=j).and.(j<=nr)) then 
@@ -194,8 +370,7 @@ subroutine amg_z_soc1_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
       ! contains I even if it does not look like it from matrix)
       !
       disjoint = all(ilaggr(icol(1:ip)) == -(nr+1)).or.(ip==0)
-      if (disjoint) then 
-        icnt      = icnt + 1 
+      if (disjoint) then       
         naggr     = naggr + 1
         do k=1, ip
           ilaggr(icol(k)) = naggr
@@ -204,16 +379,22 @@ subroutine amg_z_soc1_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
       end if
     endif
   enddo step1
-  
+#endif
   if (debug_level >= psb_debug_outer_) then 
     write(debug_unit,*) me,' ',trim(name),&
-         & ' Check 1:',count(ilaggr == -(nr+1))
+         & ' Check   1:',naggr,count(ilaggr(1:nr) == -(nr+1)), count(ilaggr(1:nr)>0),&
+         & count(ilaggr(1:nr) == -(nr+1))+count(ilaggr(1:nr)>0),nr
   end if
-
+  if (do_timings) call psb_toc(idx_soc1_p1)
+  if (do_timings) call psb_tic(idx_soc1_p2)
   !
   ! Phase two: join the neighbours
   !
+  !$omp workshare
   tmpaggr = ilaggr
+  !$omp end workshare
+  !$omp parallel do schedule(static) shared(tmpaggr,ilaggr,nr,naggr,diag,theta)& 
+  !$omp     private(ii,i,j,k,nz,icol,val,ip,cpling)
   step2: do ii=1,nr
     i = idxs(ii)
 
@@ -244,8 +425,15 @@ subroutine amg_z_soc1_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
       end if
     end if
   end do step2
+  !$omp end parallel do
+  if (do_timings) call psb_toc(idx_soc1_p2)
+  if (debug_level >= psb_debug_outer_) then 
+    write(debug_unit,*) me,' ',trim(name),&
+         & ' Check 1.5:',naggr,count(ilaggr(1:nr) == -(nr+1)), count(ilaggr(1:nr)>0),&
+         & count(ilaggr(1:nr) == -(nr+1))+count(ilaggr(1:nr)>0),nr
+  end if
 
-
+  if (do_timings) call psb_tic(idx_soc1_p3)
   !
   ! Phase three: sweep over leftovers, if any 
   !
@@ -274,7 +462,6 @@ subroutine amg_z_soc1_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
         end if
       enddo
       if (ip > 0) then
-        icnt      = icnt + 1 
         naggr     = naggr + 1
         ilaggr(i) = naggr
         do k=1, ip
@@ -292,7 +479,10 @@ subroutine amg_z_soc1_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
   end do step3
 
   ! Any leftovers?
+  !$omp parallel do schedule(static) shared(ilaggr,info)& 
+  !$omp     private(ii,i,j,k,nz,icol,val,ip)
   do i=1, nr
+    if (info /= 0) cycle
     if (ilaggr(i) < 0) then
       nz = (acsr%irp(i+1)-acsr%irp(i))
       if (nz == 1) then
@@ -303,15 +493,18 @@ subroutine amg_z_soc1_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
         ! other processes. 
         ilaggr(i) = -(nrglob+nr)
       else
+        !$omp atomic write
         info=psb_err_internal_error_
+        !$omp end atomic
         call psb_errpush(info,name,a_err='Fatal error: non-singleton leftovers')
-        goto 9999
+        cycle 
       endif
     end if
   end do
-
+  !$omp end parallel do
+  if (info /= 0) goto 9999
+  if (do_timings) call psb_toc(idx_soc1_p3)
   if (naggr > ncol) then 
-    !write(0,*) name,'Error : naggr > ncol',naggr,ncol
     info=psb_err_internal_error_
     call psb_errpush(info,name,a_err='Fatal error: naggr>ncol')
     goto 9999
@@ -336,9 +529,13 @@ subroutine amg_z_soc1_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
   nlaggr(:) = 0
   nlaggr(me+1) = naggr
   call psb_sum(ctxt,nlaggr(1:np))
+  if (debug_level >= psb_debug_outer_) then 
+    write(debug_unit,*) me,' ',trim(name),&
+         & ' Check   2:',naggr,count(ilaggr(1:nr) == -(nr+1)), count(ilaggr(1:nr)>0),&
+         & count(ilaggr(1:nr) == -(nr+1))+count(ilaggr(1:nr)>0),nr
+  end if
 
   call acsr%free()
-
   call psb_erractionrestore(err_act)
   return
 

amgprec/impl/aggregator/amg_z_soc2_map_bld.F90

@@ -68,9 +68,12 @@
 !
 subroutine amg_z_soc2_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,info)
 
-  use psb_base_mod
+  use psb_base_mod 
   use amg_base_prec_type
   use amg_z_inner_mod
+#if defined(OPENMP)
+  use omp_lib
+#endif
 
   implicit none
 
@@ -99,6 +102,9 @@ subroutine amg_z_soc2_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
   integer(psb_ipk_) :: np, me
   integer(psb_ipk_) :: nrow, ncol, n_ne
   character(len=20)  :: name, ch_err
+  integer(psb_ipk_), save :: idx_soc2_p1=-1, idx_soc2_p2=-1, idx_soc2_p3=-1
+  integer(psb_ipk_), save :: idx_soc2_p0=-1
+  logical, parameter      :: do_timings=.true.
 
   info=psb_success_
   name = 'amg_soc2_map_bld'
@@ -114,6 +120,14 @@ subroutine amg_z_soc2_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
   nrow   = desc_a%get_local_rows()
   ncol   = desc_a%get_local_cols()
   nrglob = desc_a%get_global_rows()
+  if ((do_timings).and.(idx_soc2_p0==-1))       &
+       & idx_soc2_p0 = psb_get_timer_idx("SOC2_MAP: phase0")
+  if ((do_timings).and.(idx_soc2_p1==-1))       &
+       & idx_soc2_p1 = psb_get_timer_idx("SOC2_MAP: phase1")
+  if ((do_timings).and.(idx_soc2_p2==-1))       &
+       & idx_soc2_p2 = psb_get_timer_idx("SOC2_MAP: phase2")
+  if ((do_timings).and.(idx_soc2_p3==-1))       &
+       & idx_soc2_p3 = psb_get_timer_idx("SOC2_MAP: phase3")
 
   nr = a%get_nrows()
   nc = a%get_ncols()
@@ -125,6 +139,7 @@ subroutine amg_z_soc2_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
     goto 9999
   end if
 
+  if (do_timings) call psb_tic(idx_soc2_p0)
   diag = a%get_diag(info)
   if(info /= psb_success_) then
     info=psb_err_from_subroutine_
@@ -137,55 +152,217 @@ subroutine amg_z_soc2_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
   ! 
   call a%cp_to(muij)
   if (clean_zeros) call muij%clean_zeros(info)
+  !$omp parallel do private(i,j,k) shared(nr,diag,muij) schedule(static)
   do i=1, nr
     do k=muij%irp(i),muij%irp(i+1)-1
       j = muij%ja(k)
       if (j<= nr) muij%val(k) = abs(muij%val(k))/sqrt(abs(diag(i)*diag(j)))
     end do
   end do
-
+  !$omp end parallel do 
   !
   ! Compute the 1-neigbour; mark strong links with +1, weak links with -1
   !
   call s_neigh_coo%allocate(nr,nr,muij%get_nzeros())
-  ip = 0 
+  !$omp parallel do private(i,j,k) shared(nr,diag,muij) schedule(static)
   do i=1, nr
     do k=muij%irp(i),muij%irp(i+1)-1
       j = muij%ja(k)
+      s_neigh_coo%ia(k)  = i
+      s_neigh_coo%ja(k)  = j
       if (j<=nr) then 
-        ip = ip + 1
-        s_neigh_coo%ia(ip)  = i
-        s_neigh_coo%ja(ip)  = j
         if (real(muij%val(k)) >= theta) then 
-          s_neigh_coo%val(ip) = done
+          s_neigh_coo%val(k) = done
         else
-          s_neigh_coo%val(ip) = -done
+          s_neigh_coo%val(k) = -done
         end if
+      else
+        s_neigh_coo%val(k) = -done        
       end if
     end do
   end do
+  !$omp end parallel do
   !write(*,*) 'S_NEIGH: ',nr,ip
-  call s_neigh_coo%set_nzeros(ip)
+  call s_neigh_coo%set_nzeros(muij%get_nzeros())
   call s_neigh%mv_from_coo(s_neigh_coo,info)
 
-  if (iorder == amg_aggr_ord_nat_) then 
+  if (iorder == amg_aggr_ord_nat_) then
+    
+    !$omp parallel do private(i) shared(ilaggr,idxs) schedule(static)
     do i=1, nr
       ilaggr(i) = -(nr+1)
       idxs(i)   = i 
     end do
+    !$omp end parallel do 
   else 
+    !$omp parallel do private(i) shared(ilaggr,idxs,muij)  schedule(static)
     do i=1, nr
       ilaggr(i) = -(nr+1)
       ideg(i)   = muij%irp(i+1) - muij%irp(i)
     end do
+    !$omp end parallel do
     call psb_msort(ideg,ix=idxs,dir=psb_sort_down_)
   end if
 
+  if (do_timings) call psb_toc(idx_soc2_p0)
+  if (do_timings) call psb_tic(idx_soc2_p1)
 
   !
   ! Phase one: Start with disjoint groups.
   ! 
   naggr = 0
+#if defined(OPENMP)
+  block
+    integer(psb_ipk_), allocatable  :: bnds(:), locnaggr(:)
+    integer(psb_ipk_) :: myth,nths, kk
+    ! The parallelization makes use of a locaggr(:) array; each thread
+    ! keeps its own version of naggr, and when the loop ends, a prefix is applied
+    ! to locnaggr to determine:
+    ! 1. The total number of aggregaters NAGGR;
+    ! 2. How much should each thread shift its own aggregates
+    ! Part 2 requires to keep track of which thread defined each entry
+    ! of ilaggr(), so that each entry can be adjusted correctly: even
+    ! if an entry I belongs to the range BNDS(TH)>BNDS(TH+1)-1, it may have
+    ! been set because it is strongly connected to an entry J belonging to a
+    ! different thread. 
+
+    !$omp parallel shared(s_neigh,bnds,idxs,locnaggr,ilaggr,nr,naggr,diag,theta,nths,info) &
+    !$omp private(icol,val,myth,kk) 
+    block
+      integer(psb_ipk_) :: ii,nlp,k,kp,n,ia,isz,nc,i,j,m,nz,ilg,ip,rsz,ip1,nzcnt
+      integer(psb_lpk_) :: itmp
+      !$omp master
+      nths = omp_get_num_threads()
+      allocate(bnds(0:nths),locnaggr(0:nths+1))
+      locnaggr(:) = 0
+      bnds(0) = 1
+      !$omp end master      
+      !$omp barrier
+      myth = omp_get_thread_num()
+      rsz = nr/nths
+      if (myth < mod(nr,nths)) rsz = rsz + 1
+      bnds(myth+1) = rsz
+      !$omp barrier
+      !$omp master
+      do i=1,nths
+        bnds(i) = bnds(i) + bnds(i-1)
+      end do
+      info = 0
+      !$omp end master
+      !$omp barrier
+
+      !$omp do schedule(static) private(disjoint) 
+      do kk=0, nths-1
+        step1: do ii=bnds(kk), bnds(kk+1)-1
+          i = idxs(ii)
+          if (info /= 0) then
+            write(0,*) ' Step1:',kk,ii,i,info
+            cycle step1
+          end if
+          if ((i<1).or.(i>nr)) then
+            !$omp atomic write
+            info=psb_err_internal_error_
+            !$omp end atomic 
+            call psb_errpush(info,name)
+            cycle step1
+            !goto 9999
+          end if
+
+
+          if (ilaggr(i) == -(nr+1)) then 
+            !
+            ! Get the 1-neighbourhood of I 
+            !
+            ip1 = s_neigh%irp(i)
+            nz  = s_neigh%irp(i+1)-ip1
+            !
+            ! If the neighbourhood only contains I, skip it
+            !
+            if (nz ==0) then
+              ilaggr(i) = 0
+              cycle step1
+            end if
+            if ((nz==1).and.(s_neigh%ja(ip1)==i)) then
+              ilaggr(i) = 0
+              cycle step1
+            end if
+
+            nzcnt = count(real(s_neigh%val(ip1:ip1+nz-1)) > 0)
+            icol(1:nzcnt) = pack(s_neigh%ja(ip1:ip1+nz-1),(real(s_neigh%val(ip1:ip1+nz-1)) > 0))
+            disjoint = all(ilaggr(icol(1:nzcnt)) == -(nr+1)) 
+
+            !
+            ! If the whole strongly coupled neighborhood of I is
+            ! as yet unconnected, turn it into the next aggregate.
+            ! Same if ip==0 (in which case, neighborhood only
+            ! contains I even if it does not look like it from matrix)
+            ! The fact that DISJOINT is private and not under lock
+            ! generates a certain un-repeatability, in that between
+            ! computing DISJOINT and assigning, another thread might
+            ! alter the values of ILAGGR.
+            ! However, a certain unrepeatability is already present
+            ! because the sequence of aggregates is computed with a
+            ! different order than in serial mode.
+            ! In any case, even if the enteries of ILAGGR may be
+            ! overwritten, the important thing is that each entry is
+            ! consistent and they generate a correct aggregation map.
+            !
+            if (disjoint) then
+              locnaggr(kk)     = locnaggr(kk) + 1
+              itmp = (bnds(kk)-1+locnaggr(kk))*nths+kk
+              if (itmp < (bnds(kk)-1+locnaggr(kk))) then
+                !$omp atomic update
+                info = max(12345678,info)
+                !$omp end atomic
+                cycle step1
+              end if
+              !$omp atomic write
+              ilaggr(i) = itmp
+              !$omp end atomic
+              do k=1, nzcnt
+                !$omp atomic write
+                ilaggr(icol(k)) = itmp
+                !$omp end atomic
+              end do
+            end if
+          end if
+        enddo step1
+      end do
+      !$omp end do
+
+      !$omp master
+      naggr = sum(locnaggr(0:nths-1))
+      do i=1,nths
+        locnaggr(i) = locnaggr(i) + locnaggr(i-1)
+      end do
+      do i=nths+1,1,-1
+        locnaggr(i) = locnaggr(i-1)
+      end do
+      locnaggr(0) = 0
+      !write(0,*) 'LNAG ',locnaggr(nths+1)
+      !$omp end master 
+      !$omp barrier
+      !$omp  do schedule(static) 
+      do kk=0, nths-1
+        do ii=bnds(kk), bnds(kk+1)-1
+          if (ilaggr(ii) > 0) then 
+            kp = mod(ilaggr(ii),nths)
+            ilaggr(ii) = (ilaggr(ii)/nths)- (bnds(kp)-1) + locnaggr(kp)
+          end if
+        end do
+      end do
+      !$omp end do
+    end block
+    !$omp end parallel
+  end block
+  if (info /= 0) then
+    if (info == 12345678) write(0,*) 'Overflow in encoding ILAGGR'
+    info=psb_err_internal_error_
+    call psb_errpush(info,name)
+    goto 9999
+  end if
+
+#else
   icnt = 0
   step1: do ii=1, nr
     i = idxs(ii)
@@ -224,16 +401,21 @@ subroutine amg_z_soc2_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
       end if
     endif
   enddo step1
-  
+#endif  
   if (debug_level >= psb_debug_outer_) then 
     write(debug_unit,*) me,' ',trim(name),&
          & ' Check 1:',count(ilaggr == -(nr+1))
   end if
-
+  if (do_timings) call psb_toc(idx_soc2_p1)
+  if (do_timings) call psb_tic(idx_soc2_p2)
   !
   ! Phase two: join the neighbours
   !
+  !$omp workshare
   tmpaggr = ilaggr
+  !$omp end workshare
+  !$omp parallel do schedule(static) shared(tmpaggr,ilaggr,nr,naggr,diag,muij,s_neigh)& 
+  !$omp     private(ii,i,j,k,nz,icol,val,ip,cpling)
   step2: do ii=1,nr
     i = idxs(ii)
 
@@ -259,8 +441,9 @@ subroutine amg_z_soc2_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
       end if
     end if
   end do step2
-
-
+  !$omp end parallel do
+  if (do_timings) call psb_toc(idx_soc2_p2)
+  if (do_timings) call psb_tic(idx_soc2_p3)
   !
   ! Phase three: sweep over leftovers, if any 
   !
@@ -294,6 +477,8 @@ subroutine amg_z_soc2_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
   end do step3
 
   ! Any leftovers?
+  !$omp parallel do schedule(static) shared(ilaggr,s_neigh,info)& 
+  !$omp     private(ii,i,j,k)
   do i=1, nr
     if (ilaggr(i) <= 0) then
       nz = (s_neigh%irp(i+1)-s_neigh%irp(i))
@@ -305,13 +490,17 @@ subroutine amg_z_soc2_map_bld(iorder,theta,clean_zeros,a,desc_a,nlaggr,ilaggr,in
         ! other processes. 
         ilaggr(i) = -(nrglob+nr)
       else
+        !$omp atomic write
         info=psb_err_internal_error_
+        !$omp end atomic
         call psb_errpush(info,name,a_err='Fatal error: non-singleton leftovers')
-        goto 9999
+        cycle
       endif
     end if
   end do
-
+  !$omp end parallel do
+  if (info /= 0) goto 9999
+  if (do_timings) call psb_toc(idx_soc2_p3)  
   if (naggr > ncol) then 
     info=psb_err_internal_error_
     call psb_errpush(info,name,a_err='Fatal error: naggr>ncol')

amgprec/impl/amg_ccprecset.F90

@@ -336,7 +336,8 @@ subroutine amg_ccprecsetc(p,what,string,info,ilev,ilmax,pos,idx)
   ! Local variables
   integer(psb_ipk_)                      :: ilev_, nlev_,val,ilmax_, il
   character(len=*), parameter            :: name='amg_precsetc'
-
+  logical :: hier_asb
+  
   info = psb_success_
 
   if (.not.allocated(p%precv)) then
@@ -459,6 +460,7 @@ subroutine amg_ccprecsetc(p,what,string,info,ilev,ilmax,pos,idx)
           return
         end if
         if (nlev_ > 1) then
+          hier_asb = p%precv(nlev_)%ac%is_asb()          
           select case (psb_toupper(string))
           case('BJAC')
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
@@ -469,7 +471,11 @@ subroutine amg_ccprecsetc(p,what,string,info,ilev,ilmax,pos,idx)
 #else
             call p%precv(nlev_)%set('SUB_SOLVE',psb_ilu_n_,info,pos=pos)
 #endif
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_distr_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info)
+
           case('L1-BJAC')
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_l1_bjac_,info,pos=pos)
 #if defined(HAVE_SLU_)
@@ -479,55 +485,91 @@ subroutine amg_ccprecsetc(p,what,string,info,ilev,ilmax,pos,idx)
 #else
             call p%precv(nlev_)%set('SUB_SOLVE',psb_ilu_n_,info,pos=pos)
 #endif
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_distr_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info)
           case('SLU')
 #if defined(HAVE_SLU_)
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',amg_slu_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_repl_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_repl_mat_,info,pos=pos)
 #elif defined(HAVE_MUMPS_)
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',amg_mumps_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_distr_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 #else
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',psb_ilu_n_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_distr_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 #endif
           case('ILU')
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',psb_ilu_n_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_repl_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_repl_mat_,info,pos=pos)
           case('ILUT')
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',psb_ilu_t_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_repl_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_repl_mat_,info,pos=pos)
           case('MILU')
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',psb_milu_n_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_repl_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_repl_mat_,info,pos=pos)
           case('MUMPS')
 #if defined(HAVE_MUMPS_)
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',amg_mumps_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_distr_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 #else
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',psb_ilu_n_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_distr_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 #endif
          case('UMF')
 #if defined(HAVE_SLU_)
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',amg_slu_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_repl_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_repl_mat_,info,pos=pos)
 #elif defined(HAVE_MUMPS_)
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',amg_mumps_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_distr_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 #else
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',psb_ilu_n_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_distr_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 #endif
 
@@ -535,42 +577,69 @@ subroutine amg_ccprecsetc(p,what,string,info,ilev,ilmax,pos,idx)
 #if defined(HAVE_SLU_)
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',amg_slu_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_repl_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_repl_mat_,info,pos=pos)
 #elif defined(HAVE_MUMPS_)
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',amg_mumps_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_repl_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 #else
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',psb_ilu_n_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_distr_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 #endif
           case('JACOBI','JAC')
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',amg_diag_scale_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_distr_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 
           case('L1-JACOBI','L1-JAC')
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',amg_l1_diag_scale_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_distr_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
           case('GS','FBGS')
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',amg_gs_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_distr_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
           case('BWGS')
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',amg_bwgs_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_distr_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
           case('L1-GS','L1-FBGS')
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_l1_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',amg_gs_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_distr_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
           case('KRM')
             block
               type(amg_c_krm_solver_type)          :: krm_slv
               call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_krm_,info,pos=pos)
               call p%precv(nlev_)%set(krm_slv,info)
+              if (hier_asb) &
+                   & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                   & amg_distr_mat_)
               call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
             end block
           end select
@@ -610,11 +679,16 @@ subroutine amg_ccprecsetc(p,what,string,info,ilev,ilmax,pos,idx)
 
     case('COARSE_MAT')
       if (nlev_ > 1) then
+        hier_asb = p%precv(nlev_)%ac%is_asb()                  
+        if (hier_asb) &
+             & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+             & amg_stringval(string))
         call p%precv(nlev_)%set('COARSE_MAT',string,info,pos=pos)
       end if
 
     case('COARSE_SOLVE')
       if (nlev_ > 1) then
+        hier_asb = p%precv(nlev_)%ac%is_asb()          
         call p%precv(nlev_)%set('COARSE_SOLVE',string,info,pos=pos)
         select case (psb_toupper(trim(string)))
         case('BJAC')
@@ -626,6 +700,9 @@ subroutine amg_ccprecsetc(p,what,string,info,ilev,ilmax,pos,idx)
 #else
           call p%precv(nlev_)%set('SUB_SOLVE',psb_ilu_n_,info,pos=pos)
 #endif
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_distr_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info)
         case('L1-BJAC')
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_l1_bjac_,info,pos=pos)
@@ -636,55 +713,91 @@ subroutine amg_ccprecsetc(p,what,string,info,ilev,ilmax,pos,idx)
 #else
           call p%precv(nlev_)%set('SUB_SOLVE',psb_ilu_n_,info,pos=pos)
 #endif
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_distr_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info)
         case('SLU')
 #if defined(HAVE_SLU_)
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',amg_slu_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_repl_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_repl_mat_,info,pos=pos)
 #elif defined(HAVE_MUMPS_)
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',amg_mumps_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_distr_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 #else
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',psb_ilu_n_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_distr_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 #endif
         case('ILU')
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',psb_ilu_n_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_repl_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_repl_mat_,info,pos=pos)
         case('ILUT')
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',psb_ilu_t_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_repl_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_repl_mat_,info,pos=pos)
         case('MILU')
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',psb_milu_n_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_repl_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_repl_mat_,info,pos=pos)
         case('MUMPS')
 #if defined(HAVE_MUMPS_)
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',amg_mumps_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_distr_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 #else
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',psb_ilu_n_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_distr_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 #endif
         case('UMF')
 #if defined(HAVE_SLU_)
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',amg_slu_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_repl_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_repl_mat_,info,pos=pos)
 #elif defined(HAVE_MUMPS_)
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',amg_mumps_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_distr_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 #else
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',psb_ilu_n_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_distr_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 #endif
 
@@ -692,42 +805,69 @@ subroutine amg_ccprecsetc(p,what,string,info,ilev,ilmax,pos,idx)
 #if defined(HAVE_SLU_)
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',amg_slu_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_repl_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_repl_mat_,info,pos=pos)
 #elif defined(HAVE_MUMPS_)
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',amg_mumps_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_distr_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 #else
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',psb_ilu_n_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_distr_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 #endif
         case('JACOBI','JAC')
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',amg_diag_scale_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_distr_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 
         case('L1-JACOBI','L1-JAC')
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',amg_l1_diag_scale_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_distr_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
         case('GS','FBGS')
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',amg_gs_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_distr_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
         case('BWGS')
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',amg_bwgs_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_distr_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
         case('L1-FBGS','L1-GS')
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_l1_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',amg_gs_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_distr_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
           case('KRM')
             block
               type(amg_c_krm_solver_type)          :: krm_slv
               call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_krm_,info,pos=pos)
               call p%precv(nlev_)%set(krm_slv,info)
+              if (hier_asb) &
+                   & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                   & amg_distr_mat_)
               call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
             end block
         end select

amgprec/impl/amg_cprecinit.F90

@@ -237,7 +237,6 @@ subroutine amg_cprecinit(ctxt,prec,ptype,info)
 #else
     call prec%set('COARSE_SOLVE','ILU',info)
 #endif
-
   case default
     write(psb_err_unit,*) name,&
          &': Warning: Unknown preconditioner type request "',ptype,'"'

amgprec/impl/amg_dcprecset.F90

@@ -348,7 +348,8 @@ subroutine amg_dcprecsetc(p,what,string,info,ilev,ilmax,pos,idx)
   ! Local variables
   integer(psb_ipk_)                      :: ilev_, nlev_,val,ilmax_, il
   character(len=*), parameter            :: name='amg_precsetc'
-
+  logical :: hier_asb
+  
   info = psb_success_
 
   if (.not.allocated(p%precv)) then
@@ -471,6 +472,7 @@ subroutine amg_dcprecsetc(p,what,string,info,ilev,ilmax,pos,idx)
           return
         end if
         if (nlev_ > 1) then
+          hier_asb = p%precv(nlev_)%ac%is_asb()          
           select case (psb_toupper(string))
           case('BJAC')
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
@@ -483,7 +485,11 @@ subroutine amg_dcprecsetc(p,what,string,info,ilev,ilmax,pos,idx)
 #else
             call p%precv(nlev_)%set('SUB_SOLVE',psb_ilu_n_,info,pos=pos)
 #endif
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_distr_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info)
+
           case('L1-BJAC')
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_l1_bjac_,info,pos=pos)
 #if defined(HAVE_UMF_)
@@ -495,59 +501,98 @@ subroutine amg_dcprecsetc(p,what,string,info,ilev,ilmax,pos,idx)
 #else
             call p%precv(nlev_)%set('SUB_SOLVE',psb_ilu_n_,info,pos=pos)
 #endif
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_distr_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info)
           case('SLU')
 #if defined(HAVE_SLU_)
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',amg_slu_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_repl_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_repl_mat_,info,pos=pos)
 #elif defined(HAVE_MUMPS_)
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',amg_mumps_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_distr_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 #else
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',psb_ilu_n_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_distr_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 #endif
           case('ILU')
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',psb_ilu_n_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_repl_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_repl_mat_,info,pos=pos)
           case('ILUT')
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',psb_ilu_t_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_repl_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_repl_mat_,info,pos=pos)
           case('MILU')
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',psb_milu_n_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_repl_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_repl_mat_,info,pos=pos)
           case('MUMPS')
 #if defined(HAVE_MUMPS_)
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',amg_mumps_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_distr_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 #else
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',psb_ilu_n_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_distr_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 #endif
           case('UMF')
 #if defined(HAVE_UMF_)
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',amg_umf_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_repl_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_repl_mat_,info,pos=pos)
 #elif defined(HAVE_SLU_)
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',amg_slu_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_repl_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_repl_mat_,info,pos=pos)
 #elif defined(HAVE_MUMPS_)
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',amg_mumps_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_distr_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 #else
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',psb_ilu_n_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_distr_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 #endif
 
@@ -555,50 +600,83 @@ subroutine amg_dcprecsetc(p,what,string,info,ilev,ilmax,pos,idx)
 #if defined(HAVE_SLUDIST_)
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',amg_sludist_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_distr_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 #elif defined(HAVE_UMF_)
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',amg_umf_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_repl_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_repl_mat_,info,pos=pos)
 #elif defined(HAVE_SLU_)
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',amg_slu_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_repl_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_repl_mat_,info,pos=pos)
 #elif defined(HAVE_MUMPS_)
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',amg_mumps_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_distr_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 #else
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',psb_ilu_n_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_distr_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 #endif
           case('JACOBI','JAC')
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',amg_diag_scale_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_distr_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 
           case('L1-JACOBI','L1-JAC')
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',amg_l1_diag_scale_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_distr_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
           case('GS','FBGS')
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',amg_gs_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_distr_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
           case('BWGS')
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',amg_bwgs_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_distr_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
           case('L1-GS','L1-FBGS')
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_l1_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',amg_gs_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_distr_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
           case('KRM')
             block
               type(amg_d_krm_solver_type)          :: krm_slv
               call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_krm_,info,pos=pos)
               call p%precv(nlev_)%set(krm_slv,info)
+              if (hier_asb) &
+                   & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                   & amg_distr_mat_)
               call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
             end block
           end select
@@ -638,11 +716,16 @@ subroutine amg_dcprecsetc(p,what,string,info,ilev,ilmax,pos,idx)
 
     case('COARSE_MAT')
       if (nlev_ > 1) then
+        hier_asb = p%precv(nlev_)%ac%is_asb()                  
+        if (hier_asb) &
+             & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+             & amg_stringval(string))
         call p%precv(nlev_)%set('COARSE_MAT',string,info,pos=pos)
       end if
 
     case('COARSE_SOLVE')
       if (nlev_ > 1) then
+        hier_asb = p%precv(nlev_)%ac%is_asb()          
         call p%precv(nlev_)%set('COARSE_SOLVE',string,info,pos=pos)
         select case (psb_toupper(trim(string)))
         case('BJAC')
@@ -656,6 +739,9 @@ subroutine amg_dcprecsetc(p,what,string,info,ilev,ilmax,pos,idx)
 #else
           call p%precv(nlev_)%set('SUB_SOLVE',psb_ilu_n_,info,pos=pos)
 #endif
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_distr_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info)
         case('L1-BJAC')
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_l1_bjac_,info,pos=pos)
@@ -668,59 +754,98 @@ subroutine amg_dcprecsetc(p,what,string,info,ilev,ilmax,pos,idx)
 #else
           call p%precv(nlev_)%set('SUB_SOLVE',psb_ilu_n_,info,pos=pos)
 #endif
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_distr_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info)
         case('SLU')
 #if defined(HAVE_SLU_)
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',amg_slu_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_repl_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_repl_mat_,info,pos=pos)
 #elif defined(HAVE_MUMPS_)
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',amg_mumps_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_distr_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 #else
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',psb_ilu_n_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_distr_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 #endif
         case('ILU')
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',psb_ilu_n_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_repl_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_repl_mat_,info,pos=pos)
         case('ILUT')
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',psb_ilu_t_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_repl_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_repl_mat_,info,pos=pos)
         case('MILU')
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',psb_milu_n_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_repl_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_repl_mat_,info,pos=pos)
         case('MUMPS')
 #if defined(HAVE_MUMPS_)
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',amg_mumps_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_distr_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 #else
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',psb_ilu_n_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_distr_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 #endif
         case('UMF')
 #if defined(HAVE_UMF_)
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',amg_umf_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_repl_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_repl_mat_,info,pos=pos)
 #elif defined(HAVE_SLU_)
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',amg_slu_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_repl_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_repl_mat_,info,pos=pos)
 #elif defined(HAVE_MUMPS_)
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',amg_mumps_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_distr_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 #else
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',psb_ilu_n_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_distr_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 #endif
 
@@ -728,50 +853,83 @@ subroutine amg_dcprecsetc(p,what,string,info,ilev,ilmax,pos,idx)
 #if defined(HAVE_SLUDIST_)
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',amg_sludist_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_distr_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 #elif defined(HAVE_UMF_)
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',amg_umf_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_repl_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_repl_mat_,info,pos=pos)
 #elif defined(HAVE_SLU_)
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',amg_slu_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_repl_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_repl_mat_,info,pos=pos)
 #elif defined(HAVE_MUMPS_)
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',amg_mumps_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_distr_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 #else
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',psb_ilu_n_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_distr_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 #endif
         case('JACOBI','JAC')
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',amg_diag_scale_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_distr_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 
         case('L1-JACOBI','L1-JAC')
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',amg_l1_diag_scale_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_distr_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
         case('GS','FBGS')
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',amg_gs_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_distr_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
         case('BWGS')
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',amg_bwgs_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_distr_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
         case('L1-FBGS','L1-GS')
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_l1_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',amg_gs_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_distr_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
           case('KRM')
             block
               type(amg_d_krm_solver_type)          :: krm_slv
               call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_krm_,info,pos=pos)
               call p%precv(nlev_)%set(krm_slv,info)
+              if (hier_asb) &
+                   & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                   & amg_distr_mat_)
               call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
             end block
         end select

amgprec/impl/amg_dprecinit.F90

@@ -242,7 +242,6 @@ subroutine amg_dprecinit(ctxt,prec,ptype,info)
 #else
     call prec%set('COARSE_SOLVE','ILU',info)
 #endif
-
   case default
     write(psb_err_unit,*) name,&
          &': Warning: Unknown preconditioner type request "',ptype,'"'

amgprec/impl/amg_scprecset.F90

@@ -336,7 +336,8 @@ subroutine amg_scprecsetc(p,what,string,info,ilev,ilmax,pos,idx)
   ! Local variables
   integer(psb_ipk_)                      :: ilev_, nlev_,val,ilmax_, il
   character(len=*), parameter            :: name='amg_precsetc'
-
+  logical :: hier_asb
+  
   info = psb_success_
 
   if (.not.allocated(p%precv)) then
@@ -459,6 +460,7 @@ subroutine amg_scprecsetc(p,what,string,info,ilev,ilmax,pos,idx)
           return
         end if
         if (nlev_ > 1) then
+          hier_asb = p%precv(nlev_)%ac%is_asb()          
           select case (psb_toupper(string))
           case('BJAC')
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
@@ -469,7 +471,11 @@ subroutine amg_scprecsetc(p,what,string,info,ilev,ilmax,pos,idx)
 #else
             call p%precv(nlev_)%set('SUB_SOLVE',psb_ilu_n_,info,pos=pos)
 #endif
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_distr_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info)
+
           case('L1-BJAC')
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_l1_bjac_,info,pos=pos)
 #if defined(HAVE_SLU_)
@@ -479,55 +485,91 @@ subroutine amg_scprecsetc(p,what,string,info,ilev,ilmax,pos,idx)
 #else
             call p%precv(nlev_)%set('SUB_SOLVE',psb_ilu_n_,info,pos=pos)
 #endif
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_distr_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info)
           case('SLU')
 #if defined(HAVE_SLU_)
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',amg_slu_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_repl_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_repl_mat_,info,pos=pos)
 #elif defined(HAVE_MUMPS_)
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',amg_mumps_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_distr_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 #else
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',psb_ilu_n_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_distr_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 #endif
           case('ILU')
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',psb_ilu_n_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_repl_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_repl_mat_,info,pos=pos)
           case('ILUT')
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',psb_ilu_t_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_repl_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_repl_mat_,info,pos=pos)
           case('MILU')
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',psb_milu_n_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_repl_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_repl_mat_,info,pos=pos)
           case('MUMPS')
 #if defined(HAVE_MUMPS_)
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',amg_mumps_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_distr_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 #else
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',psb_ilu_n_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_distr_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 #endif
          case('UMF')
 #if defined(HAVE_SLU_)
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',amg_slu_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_repl_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_repl_mat_,info,pos=pos)
 #elif defined(HAVE_MUMPS_)
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',amg_mumps_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_distr_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 #else
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',psb_ilu_n_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_distr_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 #endif
 
@@ -535,42 +577,69 @@ subroutine amg_scprecsetc(p,what,string,info,ilev,ilmax,pos,idx)
 #if defined(HAVE_SLU_)
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',amg_slu_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_repl_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_repl_mat_,info,pos=pos)
 #elif defined(HAVE_MUMPS_)
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',amg_mumps_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_repl_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 #else
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',psb_ilu_n_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_distr_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 #endif
           case('JACOBI','JAC')
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',amg_diag_scale_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_distr_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 
           case('L1-JACOBI','L1-JAC')
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',amg_l1_diag_scale_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_distr_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
           case('GS','FBGS')
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',amg_gs_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_distr_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
           case('BWGS')
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',amg_bwgs_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_distr_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
           case('L1-GS','L1-FBGS')
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_l1_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',amg_gs_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_distr_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
           case('KRM')
             block
               type(amg_s_krm_solver_type)          :: krm_slv
               call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_krm_,info,pos=pos)
               call p%precv(nlev_)%set(krm_slv,info)
+              if (hier_asb) &
+                   & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                   & amg_distr_mat_)
               call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
             end block
           end select
@@ -610,11 +679,16 @@ subroutine amg_scprecsetc(p,what,string,info,ilev,ilmax,pos,idx)
 
     case('COARSE_MAT')
       if (nlev_ > 1) then
+        hier_asb = p%precv(nlev_)%ac%is_asb()                  
+        if (hier_asb) &
+             & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+             & amg_stringval(string))
         call p%precv(nlev_)%set('COARSE_MAT',string,info,pos=pos)
       end if
 
     case('COARSE_SOLVE')
       if (nlev_ > 1) then
+        hier_asb = p%precv(nlev_)%ac%is_asb()          
         call p%precv(nlev_)%set('COARSE_SOLVE',string,info,pos=pos)
         select case (psb_toupper(trim(string)))
         case('BJAC')
@@ -626,6 +700,9 @@ subroutine amg_scprecsetc(p,what,string,info,ilev,ilmax,pos,idx)
 #else
           call p%precv(nlev_)%set('SUB_SOLVE',psb_ilu_n_,info,pos=pos)
 #endif
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_distr_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info)
         case('L1-BJAC')
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_l1_bjac_,info,pos=pos)
@@ -636,55 +713,91 @@ subroutine amg_scprecsetc(p,what,string,info,ilev,ilmax,pos,idx)
 #else
           call p%precv(nlev_)%set('SUB_SOLVE',psb_ilu_n_,info,pos=pos)
 #endif
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_distr_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info)
         case('SLU')
 #if defined(HAVE_SLU_)
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',amg_slu_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_repl_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_repl_mat_,info,pos=pos)
 #elif defined(HAVE_MUMPS_)
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',amg_mumps_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_distr_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 #else
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',psb_ilu_n_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_distr_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 #endif
         case('ILU')
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',psb_ilu_n_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_repl_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_repl_mat_,info,pos=pos)
         case('ILUT')
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',psb_ilu_t_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_repl_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_repl_mat_,info,pos=pos)
         case('MILU')
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',psb_milu_n_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_repl_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_repl_mat_,info,pos=pos)
         case('MUMPS')
 #if defined(HAVE_MUMPS_)
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',amg_mumps_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_distr_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 #else
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',psb_ilu_n_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_distr_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 #endif
         case('UMF')
 #if defined(HAVE_SLU_)
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',amg_slu_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_repl_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_repl_mat_,info,pos=pos)
 #elif defined(HAVE_MUMPS_)
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',amg_mumps_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_distr_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 #else
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',psb_ilu_n_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_distr_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 #endif
 
@@ -692,42 +805,69 @@ subroutine amg_scprecsetc(p,what,string,info,ilev,ilmax,pos,idx)
 #if defined(HAVE_SLU_)
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',amg_slu_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_repl_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_repl_mat_,info,pos=pos)
 #elif defined(HAVE_MUMPS_)
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',amg_mumps_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_distr_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 #else
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',psb_ilu_n_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_distr_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 #endif
         case('JACOBI','JAC')
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',amg_diag_scale_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_distr_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 
         case('L1-JACOBI','L1-JAC')
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',amg_l1_diag_scale_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_distr_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
         case('GS','FBGS')
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',amg_gs_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_distr_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
         case('BWGS')
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',amg_bwgs_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_distr_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
         case('L1-FBGS','L1-GS')
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_l1_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',amg_gs_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_distr_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
           case('KRM')
             block
               type(amg_s_krm_solver_type)          :: krm_slv
               call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_krm_,info,pos=pos)
               call p%precv(nlev_)%set(krm_slv,info)
+              if (hier_asb) &
+                   & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                   & amg_distr_mat_)
               call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
             end block
         end select

amgprec/impl/amg_sprecinit.F90

@@ -237,7 +237,6 @@ subroutine amg_sprecinit(ctxt,prec,ptype,info)
 #else
     call prec%set('COARSE_SOLVE','ILU',info)
 #endif
-
   case default
     write(psb_err_unit,*) name,&
          &': Warning: Unknown preconditioner type request "',ptype,'"'

amgprec/impl/amg_zcprecset.F90

@@ -348,7 +348,8 @@ subroutine amg_zcprecsetc(p,what,string,info,ilev,ilmax,pos,idx)
   ! Local variables
   integer(psb_ipk_)                      :: ilev_, nlev_,val,ilmax_, il
   character(len=*), parameter            :: name='amg_precsetc'
-
+  logical :: hier_asb
+  
   info = psb_success_
 
   if (.not.allocated(p%precv)) then
@@ -471,6 +472,7 @@ subroutine amg_zcprecsetc(p,what,string,info,ilev,ilmax,pos,idx)
           return
         end if
         if (nlev_ > 1) then
+          hier_asb = p%precv(nlev_)%ac%is_asb()          
           select case (psb_toupper(string))
           case('BJAC')
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
@@ -483,7 +485,11 @@ subroutine amg_zcprecsetc(p,what,string,info,ilev,ilmax,pos,idx)
 #else
             call p%precv(nlev_)%set('SUB_SOLVE',psb_ilu_n_,info,pos=pos)
 #endif
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_distr_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info)
+
           case('L1-BJAC')
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_l1_bjac_,info,pos=pos)
 #if defined(HAVE_UMF_)
@@ -495,59 +501,98 @@ subroutine amg_zcprecsetc(p,what,string,info,ilev,ilmax,pos,idx)
 #else
             call p%precv(nlev_)%set('SUB_SOLVE',psb_ilu_n_,info,pos=pos)
 #endif
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_distr_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info)
           case('SLU')
 #if defined(HAVE_SLU_)
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',amg_slu_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_repl_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_repl_mat_,info,pos=pos)
 #elif defined(HAVE_MUMPS_)
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',amg_mumps_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_distr_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 #else
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',psb_ilu_n_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_distr_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 #endif
           case('ILU')
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',psb_ilu_n_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_repl_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_repl_mat_,info,pos=pos)
           case('ILUT')
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',psb_ilu_t_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_repl_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_repl_mat_,info,pos=pos)
           case('MILU')
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',psb_milu_n_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_repl_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_repl_mat_,info,pos=pos)
           case('MUMPS')
 #if defined(HAVE_MUMPS_)
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',amg_mumps_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_distr_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 #else
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',psb_ilu_n_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_distr_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 #endif
           case('UMF')
 #if defined(HAVE_UMF_)
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',amg_umf_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_repl_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_repl_mat_,info,pos=pos)
 #elif defined(HAVE_SLU_)
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',amg_slu_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_repl_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_repl_mat_,info,pos=pos)
 #elif defined(HAVE_MUMPS_)
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',amg_mumps_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_distr_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 #else
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',psb_ilu_n_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_distr_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 #endif
 
@@ -555,50 +600,83 @@ subroutine amg_zcprecsetc(p,what,string,info,ilev,ilmax,pos,idx)
 #if defined(HAVE_SLUDIST_)
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',amg_sludist_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_distr_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 #elif defined(HAVE_UMF_)
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',amg_umf_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_repl_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_repl_mat_,info,pos=pos)
 #elif defined(HAVE_SLU_)
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',amg_slu_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_repl_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_repl_mat_,info,pos=pos)
 #elif defined(HAVE_MUMPS_)
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',amg_mumps_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_distr_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 #else
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',psb_ilu_n_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_distr_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 #endif
           case('JACOBI','JAC')
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',amg_diag_scale_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_distr_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 
           case('L1-JACOBI','L1-JAC')
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',amg_l1_diag_scale_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_distr_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
           case('GS','FBGS')
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',amg_gs_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_distr_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
           case('BWGS')
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',amg_bwgs_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_distr_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
           case('L1-GS','L1-FBGS')
             call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_l1_bjac_,info,pos=pos)
             call p%precv(nlev_)%set('SUB_SOLVE',amg_gs_,info,pos=pos)
+            if (hier_asb) &
+                 & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                 & amg_distr_mat_)
             call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
           case('KRM')
             block
               type(amg_z_krm_solver_type)          :: krm_slv
               call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_krm_,info,pos=pos)
               call p%precv(nlev_)%set(krm_slv,info)
+              if (hier_asb) &
+                   & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                   & amg_distr_mat_)
               call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
             end block
           end select
@@ -638,11 +716,16 @@ subroutine amg_zcprecsetc(p,what,string,info,ilev,ilmax,pos,idx)
 
     case('COARSE_MAT')
       if (nlev_ > 1) then
+        hier_asb = p%precv(nlev_)%ac%is_asb()                  
+        if (hier_asb) &
+             & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+             & amg_stringval(string))
         call p%precv(nlev_)%set('COARSE_MAT',string,info,pos=pos)
       end if
 
     case('COARSE_SOLVE')
       if (nlev_ > 1) then
+        hier_asb = p%precv(nlev_)%ac%is_asb()          
         call p%precv(nlev_)%set('COARSE_SOLVE',string,info,pos=pos)
         select case (psb_toupper(trim(string)))
         case('BJAC')
@@ -656,6 +739,9 @@ subroutine amg_zcprecsetc(p,what,string,info,ilev,ilmax,pos,idx)
 #else
           call p%precv(nlev_)%set('SUB_SOLVE',psb_ilu_n_,info,pos=pos)
 #endif
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_distr_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info)
         case('L1-BJAC')
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_l1_bjac_,info,pos=pos)
@@ -668,59 +754,98 @@ subroutine amg_zcprecsetc(p,what,string,info,ilev,ilmax,pos,idx)
 #else
           call p%precv(nlev_)%set('SUB_SOLVE',psb_ilu_n_,info,pos=pos)
 #endif
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_distr_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info)
         case('SLU')
 #if defined(HAVE_SLU_)
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',amg_slu_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_repl_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_repl_mat_,info,pos=pos)
 #elif defined(HAVE_MUMPS_)
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',amg_mumps_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_distr_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 #else
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',psb_ilu_n_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_distr_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 #endif
         case('ILU')
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',psb_ilu_n_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_repl_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_repl_mat_,info,pos=pos)
         case('ILUT')
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',psb_ilu_t_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_repl_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_repl_mat_,info,pos=pos)
         case('MILU')
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',psb_milu_n_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_repl_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_repl_mat_,info,pos=pos)
         case('MUMPS')
 #if defined(HAVE_MUMPS_)
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',amg_mumps_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_distr_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 #else
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',psb_ilu_n_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_distr_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 #endif
         case('UMF')
 #if defined(HAVE_UMF_)
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',amg_umf_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_repl_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_repl_mat_,info,pos=pos)
 #elif defined(HAVE_SLU_)
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',amg_slu_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_repl_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_repl_mat_,info,pos=pos)
 #elif defined(HAVE_MUMPS_)
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',amg_mumps_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_distr_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 #else
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',psb_ilu_n_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_distr_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 #endif
 
@@ -728,50 +853,83 @@ subroutine amg_zcprecsetc(p,what,string,info,ilev,ilmax,pos,idx)
 #if defined(HAVE_SLUDIST_)
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',amg_sludist_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_distr_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 #elif defined(HAVE_UMF_)
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',amg_umf_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_repl_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_repl_mat_,info,pos=pos)
 #elif defined(HAVE_SLU_)
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',amg_slu_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_repl_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_repl_mat_,info,pos=pos)
 #elif defined(HAVE_MUMPS_)
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',amg_mumps_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_distr_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 #else
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',psb_ilu_n_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_distr_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 #endif
         case('JACOBI','JAC')
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',amg_diag_scale_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_distr_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
 
         case('L1-JACOBI','L1-JAC')
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',amg_l1_diag_scale_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_distr_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
         case('GS','FBGS')
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',amg_gs_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_distr_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
         case('BWGS')
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',amg_bwgs_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_distr_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
         case('L1-FBGS','L1-GS')
           call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_l1_bjac_,info,pos=pos)
           call p%precv(nlev_)%set('SUB_SOLVE',amg_gs_,info,pos=pos)
+          if (hier_asb) &
+               & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+               & amg_distr_mat_)
           call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
           case('KRM')
             block
               type(amg_z_krm_solver_type)          :: krm_slv
               call p%precv(nlev_)%set('SMOOTHER_TYPE',amg_krm_,info,pos=pos)
               call p%precv(nlev_)%set(krm_slv,info)
+              if (hier_asb) &
+                   & call amg_warn_coarse_mat(p%precv(nlev_)%parms%get_coarse_mat(),&
+                   & amg_distr_mat_)
               call p%precv(nlev_)%set('COARSE_MAT',amg_distr_mat_,info,pos=pos)
             end block
         end select

amgprec/impl/amg_zprecinit.F90

@@ -242,7 +242,6 @@ subroutine amg_zprecinit(ctxt,prec,ptype,info)
 #else
     call prec%set('COARSE_SOLVE','ILU',info)
 #endif
-
   case default
     write(psb_err_unit,*) name,&
          &': Warning: Unknown preconditioner type request "',ptype,'"'

amgprec/impl/level/Makefile

@@ -17,6 +17,7 @@ amg_c_base_onelev_csetr.o \
 amg_c_base_onelev_descr.o \
 amg_c_base_onelev_dump.o \
 amg_c_base_onelev_free.o \
+amg_c_base_onelev_free_smoothers.o \
 amg_c_base_onelev_mat_asb.o \
 amg_c_base_onelev_setag.o \
 amg_c_base_onelev_setsm.o \
@@ -32,6 +33,7 @@ amg_d_base_onelev_csetr.o \
 amg_d_base_onelev_descr.o \
 amg_d_base_onelev_dump.o \
 amg_d_base_onelev_free.o \
+amg_d_base_onelev_free_smoothers.o \
 amg_d_base_onelev_mat_asb.o \
 amg_d_base_onelev_setag.o \
 amg_d_base_onelev_setsm.o \
@@ -47,6 +49,7 @@ amg_s_base_onelev_csetr.o \
 amg_s_base_onelev_descr.o \
 amg_s_base_onelev_dump.o \
 amg_s_base_onelev_free.o \
+amg_s_base_onelev_free_smoothers.o \
 amg_s_base_onelev_mat_asb.o \
 amg_s_base_onelev_setag.o \
 amg_s_base_onelev_setsm.o \
@@ -62,6 +65,7 @@ amg_z_base_onelev_csetr.o \
 amg_z_base_onelev_descr.o \
 amg_z_base_onelev_dump.o \
 amg_z_base_onelev_free.o \
+amg_z_base_onelev_free_smoothers.o \
 amg_z_base_onelev_mat_asb.o \
 amg_z_base_onelev_setag.o \
 amg_z_base_onelev_setsm.o \

amgprec/impl/level/amg_c_base_onelev_free_smoothers.f90

@@ -0,0 +1,60 @@
+!  
+!   
+!                             AMG4PSBLAS version 1.0
+!    Algebraic Multigrid Package
+!               based on PSBLAS (Parallel Sparse BLAS version 3.7)
+!    
+!    (C) Copyright 2021 
+!  
+!        Salvatore Filippone  
+!        Pasqua D'Ambra   
+!        Fabio Durastante        
+!   
+!    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 AMG4PSBLAS 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 AMG4PSBLAS 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.
+!   
+!  
+subroutine amg_c_base_onelev_free_smoothers(lv,info)
+  
+  use psb_base_mod
+  use amg_c_onelev_mod, amg_protect_name =>  amg_c_base_onelev_free_smoothers
+  implicit none 
+
+  class(amg_c_onelev_type), intent(inout) :: lv
+  integer(psb_ipk_), intent(out)          :: info
+  integer(psb_ipk_) :: i
+
+  info = psb_success_
+
+  ! We might just deallocate the top level array, except 
+  ! that there may be inner objects containing C pointers,
+  ! e.g.  UMFPACK, SLU or CUDA stuff.
+  ! We really need FINALs. 
+  if (allocated(lv%sm)) &
+       & call lv%sm%free(info)
+
+  if (allocated(lv%sm2a)) &
+       & call lv%sm2a%free(info)
+
+end subroutine amg_c_base_onelev_free_smoothers

amgprec/impl/level/amg_d_base_onelev_free_smoothers.f90

@@ -0,0 +1,60 @@
+!  
+!   
+!                             AMG4PSBLAS version 1.0
+!    Algebraic Multigrid Package
+!               based on PSBLAS (Parallel Sparse BLAS version 3.7)
+!    
+!    (C) Copyright 2021 
+!  
+!        Salvatore Filippone  
+!        Pasqua D'Ambra   
+!        Fabio Durastante        
+!   
+!    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 AMG4PSBLAS 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 AMG4PSBLAS 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.
+!   
+!  
+subroutine amg_d_base_onelev_free_smoothers(lv,info)
+  
+  use psb_base_mod
+  use amg_d_onelev_mod, amg_protect_name =>  amg_d_base_onelev_free_smoothers
+  implicit none 
+
+  class(amg_d_onelev_type), intent(inout) :: lv
+  integer(psb_ipk_), intent(out)          :: info
+  integer(psb_ipk_) :: i
+
+  info = psb_success_
+
+  ! We might just deallocate the top level array, except 
+  ! that there may be inner objects containing C pointers,
+  ! e.g.  UMFPACK, SLU or CUDA stuff.
+  ! We really need FINALs. 
+  if (allocated(lv%sm)) &
+       & call lv%sm%free(info)
+
+  if (allocated(lv%sm2a)) &
+       & call lv%sm2a%free(info)
+
+end subroutine amg_d_base_onelev_free_smoothers

amgprec/impl/level/amg_s_base_onelev_free_smoothers.f90

@@ -0,0 +1,60 @@
+!  
+!   
+!                             AMG4PSBLAS version 1.0
+!    Algebraic Multigrid Package
+!               based on PSBLAS (Parallel Sparse BLAS version 3.7)
+!    
+!    (C) Copyright 2021 
+!  
+!        Salvatore Filippone  
+!        Pasqua D'Ambra   
+!        Fabio Durastante        
+!   
+!    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 AMG4PSBLAS 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 AMG4PSBLAS 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.
+!   
+!  
+subroutine amg_s_base_onelev_free_smoothers(lv,info)
+  
+  use psb_base_mod
+  use amg_s_onelev_mod, amg_protect_name =>  amg_s_base_onelev_free_smoothers
+  implicit none 
+
+  class(amg_s_onelev_type), intent(inout) :: lv
+  integer(psb_ipk_), intent(out)          :: info
+  integer(psb_ipk_) :: i
+
+  info = psb_success_
+
+  ! We might just deallocate the top level array, except 
+  ! that there may be inner objects containing C pointers,
+  ! e.g.  UMFPACK, SLU or CUDA stuff.
+  ! We really need FINALs. 
+  if (allocated(lv%sm)) &
+       & call lv%sm%free(info)
+
+  if (allocated(lv%sm2a)) &
+       & call lv%sm2a%free(info)
+
+end subroutine amg_s_base_onelev_free_smoothers

amgprec/impl/level/amg_z_base_onelev_free_smoothers.f90

@@ -0,0 +1,60 @@
+!  
+!   
+!                             AMG4PSBLAS version 1.0
+!    Algebraic Multigrid Package
+!               based on PSBLAS (Parallel Sparse BLAS version 3.7)
+!    
+!    (C) Copyright 2021 
+!  
+!        Salvatore Filippone  
+!        Pasqua D'Ambra   
+!        Fabio Durastante        
+!   
+!    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 AMG4PSBLAS 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 AMG4PSBLAS 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.
+!   
+!  
+subroutine amg_z_base_onelev_free_smoothers(lv,info)
+  
+  use psb_base_mod
+  use amg_z_onelev_mod, amg_protect_name =>  amg_z_base_onelev_free_smoothers
+  implicit none 
+
+  class(amg_z_onelev_type), intent(inout) :: lv
+  integer(psb_ipk_), intent(out)          :: info
+  integer(psb_ipk_) :: i
+
+  info = psb_success_
+
+  ! We might just deallocate the top level array, except 
+  ! that there may be inner objects containing C pointers,
+  ! e.g.  UMFPACK, SLU or CUDA stuff.
+  ! We really need FINALs. 
+  if (allocated(lv%sm)) &
+       & call lv%sm%free(info)
+
+  if (allocated(lv%sm2a)) &
+       & call lv%sm2a%free(info)
+
+end subroutine amg_z_base_onelev_free_smoothers

amgprec/impl/solver/Makefile

@@ -197,6 +197,7 @@ amg_z_mumps_solver_bld.o \
 amg_s_ainv_solver_bld.o \
 amg_s_ainv_solver_check.o \
 amg_s_ainv_solver_clone.o \
+amg_s_ainv_solver_clone_settings.o \
 amg_s_ainv_solver_csetc.o \
 amg_s_ainv_solver_cseti.o \
 amg_s_ainv_solver_csetr.o \
@@ -204,6 +205,7 @@ amg_s_ainv_solver_descr.o \
 amg_d_ainv_solver_bld.o \
 amg_d_ainv_solver_check.o \
 amg_d_ainv_solver_clone.o \
+amg_d_ainv_solver_clone_settings.o \
 amg_d_ainv_solver_csetc.o \
 amg_d_ainv_solver_cseti.o \
 amg_d_ainv_solver_csetr.o \
@@ -211,6 +213,7 @@ amg_d_ainv_solver_descr.o \
 amg_c_ainv_solver_bld.o \
 amg_c_ainv_solver_check.o \
 amg_c_ainv_solver_clone.o \
+amg_c_ainv_solver_clone_settings.o \
 amg_c_ainv_solver_csetc.o \
 amg_c_ainv_solver_cseti.o \
 amg_c_ainv_solver_csetr.o \
@@ -218,6 +221,7 @@ amg_c_ainv_solver_descr.o \
 amg_z_ainv_solver_bld.o \
 amg_z_ainv_solver_check.o \
 amg_z_ainv_solver_clone.o \
+amg_z_ainv_solver_clone_settings.o \
 amg_z_ainv_solver_csetc.o \
 amg_z_ainv_solver_cseti.o \
 amg_z_ainv_solver_csetr.o \
@@ -249,45 +253,53 @@ amg_z_base_ainv_update_a.o \
 amg_d_invt_solver_bld.o \
 amg_d_invt_solver_check.o \
 amg_d_invt_solver_clone.o \
+amg_d_invt_solver_clone_settings.o \
 amg_d_invt_solver_cseti.o \
 amg_d_invt_solver_csetr.o \
 amg_d_invt_solver_descr.o \
 amg_d_invk_solver_bld.o \
 amg_d_invk_solver_check.o \
 amg_d_invk_solver_clone.o \
+amg_d_invk_solver_clone_settings.o \
 amg_d_invk_solver_cseti.o \
 amg_d_invk_solver_descr.o \
 amg_s_invt_solver_bld.o \
 amg_s_invt_solver_check.o \
 amg_s_invt_solver_clone.o \
+amg_s_invt_solver_clone_settings.o \
 amg_s_invt_solver_cseti.o \
 amg_s_invt_solver_csetr.o \
 amg_s_invt_solver_descr.o \
 amg_s_invk_solver_bld.o \
 amg_s_invk_solver_check.o \
 amg_s_invk_solver_clone.o \
+amg_s_invk_solver_clone_settings.o \
 amg_s_invk_solver_cseti.o \
 amg_s_invk_solver_descr.o \
 amg_c_invt_solver_bld.o \
 amg_c_invt_solver_check.o \
 amg_c_invt_solver_clone.o \
+amg_c_invt_solver_clone_settings.o \
 amg_c_invt_solver_cseti.o \
 amg_c_invt_solver_csetr.o \
 amg_c_invt_solver_descr.o \
 amg_c_invk_solver_bld.o \
 amg_c_invk_solver_check.o \
 amg_c_invk_solver_clone.o \
+amg_c_invk_solver_clone_settings.o \
 amg_c_invk_solver_cseti.o \
 amg_c_invk_solver_descr.o \
 amg_z_invt_solver_bld.o \
 amg_z_invt_solver_check.o \
 amg_z_invt_solver_clone.o \
+amg_z_invt_solver_clone_settings.o \
 amg_z_invt_solver_cseti.o \
 amg_z_invt_solver_csetr.o \
 amg_z_invt_solver_descr.o \
 amg_z_invk_solver_bld.o \
 amg_z_invk_solver_check.o \
 amg_z_invk_solver_clone.o \
+amg_z_invk_solver_clone_settings.o \
 amg_z_invk_solver_cseti.o \
 amg_z_invk_solver_descr.o \
 amg_z_krm_solver_impl.o 

amgprec/impl/solver/amg_c_ainv_solver_clone_settings.f90

@@ -0,0 +1,84 @@
+!
+!
+!                             AMG4PSBLAS version 1.0
+!    MultiLevel Domain Decomposition Parallel Preconditioners Package
+!               based on PSBLAS (Parallel Sparse BLAS version 3.0)
+!
+!    (C) Copyright 2008,2009,2010,2010,2012
+!
+!                        Salvatore Filippone  University of Rome Tor Vergata
+!                        Alfredo Buttari      CNRS-IRIT, Toulouse
+!                        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 AMG4PSBLAS 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 AMG4PSBLAS 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.
+!
+!
+subroutine amg_c_ainv_solver_clone_settings(sv,svout,info)
+
+  use psb_base_mod
+  use amg_c_ainv_solver, amg_protect_name => amg_c_ainv_solver_clone_settings
+
+  Implicit None
+
+  ! Arguments
+  class(amg_c_ainv_solver_type), intent(inout)              :: sv
+  class(amg_c_base_solver_type), allocatable, intent(inout) :: svout
+  integer(psb_ipk_), intent(out)                            :: info
+  ! Local variables
+  integer(psb_ipk_) :: err_act
+
+
+  info=psb_success_
+  call psb_erractionsave(err_act)
+
+  if (allocated(svout)) then
+    call svout%free(info)
+    if (info == psb_success_) deallocate(svout, stat=info)
+  end if
+  if (info == psb_success_) &
+       & allocate(amg_c_ainv_solver_type :: svout, stat=info)
+  if (info /= 0) then
+    info = psb_err_alloc_dealloc_
+    goto 9999
+  end if
+  select type(svo => svout)
+  type is (amg_c_ainv_solver_type)
+    svo%alg      = sv%alg
+    svo%fill_in  = sv%fill_in
+    svo%thresh   = sv%thresh
+
+  class default
+    info = psb_err_internal_error_
+  end select
+
+  if (info /= 0) goto 9999
+
+  call psb_erractionrestore(err_act)
+  return
+
+9999 call psb_error_handler(err_act)
+  return
+end subroutine amg_c_ainv_solver_clone_settings

amgprec/impl/solver/amg_c_bwgs_solver_bld.f90

@@ -56,6 +56,8 @@ subroutine amg_c_bwgs_solver_bld(a,desc_a,sv,info,b,amold,vmold,imold)
   type(psb_ctxt_type) :: ctxt
   integer(psb_ipk_)   :: np, me, i, err_act, debug_unit, debug_level
   character(len=20)   :: name='d_bwgs_solver_bld', ch_err
+  integer(psb_ipk_), save :: idx_tril=-1
+  logical, parameter      :: do_timings=.true.
 
   info=psb_success_
   call psb_erractionsave(err_act)
@@ -65,6 +67,8 @@ subroutine amg_c_bwgs_solver_bld(a,desc_a,sv,info,b,amold,vmold,imold)
   call psb_info(ctxt, me, np)
   if (debug_level >= psb_debug_outer_) &
        & write(debug_unit,*) me,' ',trim(name),' start'
+  if ((do_timings).and.(idx_tril==-1))       &
+       & idx_tril = psb_get_timer_idx("BWGS_BLD: tril")
 
 
   n_row  = desc_a%get_local_rows()
@@ -77,7 +81,10 @@ subroutine amg_c_bwgs_solver_bld(a,desc_a,sv,info,b,amold,vmold,imold)
     ! This cuts out the off-diagonal part, because it's supposed to
     ! be handled by the outer Jacobi smoother.
     !
+    !write(0,*) 'Calling A%TRIL in bwgs_solver_bld'
+    if (do_timings) call psb_tic(idx_tril)
     call a%tril(sv%l,info,diag=-ione,jmax=nrow_a,u=sv%u)     
+    if (do_timings) call psb_toc(idx_tril)
 
   else
 

amgprec/impl/solver/amg_c_gs_solver_bld.f90

@@ -56,6 +56,8 @@ subroutine amg_c_gs_solver_bld(a,desc_a,sv,info,b,amold,vmold,imold)
   type(psb_ctxt_type) :: ctxt
   integer(psb_ipk_)   :: np, me, i, err_act, debug_unit, debug_level
   character(len=20)   :: name='c_gs_solver_bld', ch_err
+  integer(psb_ipk_), save :: idx_tril=-1
+  logical, parameter      :: do_timings=.true.
 
   info=psb_success_
   call psb_erractionsave(err_act)
@@ -65,6 +67,8 @@ subroutine amg_c_gs_solver_bld(a,desc_a,sv,info,b,amold,vmold,imold)
   call psb_info(ctxt, me, np)
   if (debug_level >= psb_debug_outer_) &
        & write(debug_unit,*) me,' ',trim(name),' start'
+  if ((do_timings).and.(idx_tril==-1))       &
+       & idx_tril = psb_get_timer_idx("GS_BLD: tril")
 
 
   n_row  = desc_a%get_local_rows()
@@ -76,9 +80,12 @@ subroutine amg_c_gs_solver_bld(a,desc_a,sv,info,b,amold,vmold,imold)
     !
     ! This cuts out the off-diagonal part, because it's supposed to
     ! be handled by the outer Jacobi smoother.
-    ! 
+    !
+    !write(0,*) 'Calling A%TRIL in gs_solver_bld'
+    if (do_timings) call psb_tic(idx_tril)
     call a%tril(sv%l,info,diag=izero,jmax=nrow_a,u=sv%u)
-
+    if (do_timings) call psb_toc(idx_tril)
+    !write(0,*) 'From A%TRIL in gs_solver_bld',a%get_nzeros(),sv%l%get_nzeros(),sv%u%get_nzeros()
   else
 
     info = psb_err_missing_override_method_

amgprec/impl/solver/amg_c_invk_solver_clone_settings.f90

@@ -0,0 +1,83 @@
+!  
+!
+!                             AMG4PSBLAS version 1.0
+!    MultiLevel Domain Decomposition Parallel Preconditioners Package
+!               based on PSBLAS (Parallel Sparse BLAS version 3.0)
+!
+!    (C) Copyright 2008,2009,2010,2010,2012
+!
+!                        Salvatore Filippone  University of Rome Tor Vergata
+!                        Alfredo Buttari      CNRS-IRIT, Toulouse
+!                        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 AMG4PSBLAS 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 AMG4PSBLAS 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.
+!
+!
+subroutine amg_c_invk_solver_clone_settings(sv,svout,info)
+
+  use psb_base_mod
+  use amg_c_invk_solver, amg_protect_name => amg_c_invk_solver_clone_settings
+
+  Implicit None
+
+  ! Arguments
+  class(amg_c_invk_solver_type), intent(inout)              :: sv
+  class(amg_c_base_solver_type), allocatable, intent(inout) :: svout
+  integer(psb_ipk_), intent(out)                            :: info
+  ! Local variables
+  integer(psb_ipk_) :: err_act
+
+
+  info=psb_success_
+  call psb_erractionsave(err_act)
+
+  if (allocated(svout)) then
+    call svout%free(info)
+    if (info == psb_success_) deallocate(svout, stat=info)
+  end if
+  if (info == psb_success_) &
+       & allocate(amg_c_invk_solver_type :: svout, stat=info)
+  if (info /= 0) then
+    info = psb_err_alloc_dealloc_
+    goto 9999
+  end if
+  select type(svo => svout)
+  type is (amg_c_invk_solver_type)
+    svo%fill_in  = sv%fill_in
+    svo%inv_fill = sv%inv_fill
+
+  class default
+    info = psb_err_internal_error_
+  end select
+
+  if (info /= 0) goto 9999
+
+  call psb_erractionrestore(err_act)
+  return
+
+9999 call psb_error_handler(err_act)
+  return
+end subroutine amg_c_invk_solver_clone_settings

amgprec/impl/solver/amg_c_invt_solver_clone_settings.f90

@@ -0,0 +1,85 @@
+!  
+!
+!                             AMG4PSBLAS version 1.0
+!    MultiLevel Domain Decomposition Parallel Preconditioners Package
+!               based on PSBLAS (Parallel Sparse BLAS version 3.0)
+!
+!    (C) Copyright 2008,2009,2010,2010,2012
+!
+!                        Salvatore Filippone  University of Rome Tor Vergata
+!                        Alfredo Buttari      CNRS-IRIT, Toulouse
+!                        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 AMG4PSBLAS 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 AMG4PSBLAS 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.
+!
+!
+subroutine amg_c_invt_solver_clone_settings(sv,svout,info)
+
+  use psb_base_mod
+  use amg_c_invt_solver, amg_protect_name => amg_c_invt_solver_clone_settings
+
+  Implicit None
+
+  ! Arguments
+  class(amg_c_invt_solver_type), intent(inout)              :: sv
+  class(amg_c_base_solver_type), allocatable, intent(inout) :: svout
+  integer(psb_ipk_), intent(out)                            :: info
+  ! Local variables
+  integer(psb_ipk_) :: err_act
+
+
+  info=psb_success_
+  call psb_erractionsave(err_act)
+
+  if (allocated(svout)) then
+    call svout%free(info)
+    if (info == psb_success_) deallocate(svout, stat=info)
+  end if
+  if (info == psb_success_) &
+       & allocate(amg_c_invt_solver_type :: svout, stat=info)
+  if (info /= 0) then
+    info = psb_err_alloc_dealloc_
+    goto 9999
+  end if
+  select type(svo => svout)
+  type is (amg_c_invt_solver_type)
+    svo%fill_in    = sv%fill_in
+    svo%inv_fill   = sv%inv_fill
+    svo%thresh     = sv%thresh
+    svo%inv_thresh = sv%inv_thresh
+
+  class default
+    info = psb_err_internal_error_
+  end select
+
+  if (info /= 0) goto 9999
+
+  call psb_erractionrestore(err_act)
+  return
+
+9999 call psb_error_handler(err_act)
+  return
+end subroutine amg_c_invt_solver_clone_settings

amgprec/impl/solver/amg_d_ainv_solver_clone_settings.f90

@@ -0,0 +1,84 @@
+!
+!
+!                             AMG4PSBLAS version 1.0
+!    MultiLevel Domain Decomposition Parallel Preconditioners Package
+!               based on PSBLAS (Parallel Sparse BLAS version 3.0)
+!
+!    (C) Copyright 2008,2009,2010,2010,2012
+!
+!                        Salvatore Filippone  University of Rome Tor Vergata
+!                        Alfredo Buttari      CNRS-IRIT, Toulouse
+!                        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 AMG4PSBLAS 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 AMG4PSBLAS 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.
+!
+!
+subroutine amg_d_ainv_solver_clone_settings(sv,svout,info)
+
+  use psb_base_mod
+  use amg_d_ainv_solver, amg_protect_name => amg_d_ainv_solver_clone_settings
+
+  Implicit None
+
+  ! Arguments
+  class(amg_d_ainv_solver_type), intent(inout)              :: sv
+  class(amg_d_base_solver_type), allocatable, intent(inout) :: svout
+  integer(psb_ipk_), intent(out)                            :: info
+  ! Local variables
+  integer(psb_ipk_) :: err_act
+
+
+  info=psb_success_
+  call psb_erractionsave(err_act)
+
+  if (allocated(svout)) then
+    call svout%free(info)
+    if (info == psb_success_) deallocate(svout, stat=info)
+  end if
+  if (info == psb_success_) &
+       & allocate(amg_d_ainv_solver_type :: svout, stat=info)
+  if (info /= 0) then
+    info = psb_err_alloc_dealloc_
+    goto 9999
+  end if
+  select type(svo => svout)
+  type is (amg_d_ainv_solver_type)
+    svo%alg      = sv%alg
+    svo%fill_in  = sv%fill_in
+    svo%thresh   = sv%thresh
+
+  class default
+    info = psb_err_internal_error_
+  end select
+
+  if (info /= 0) goto 9999
+
+  call psb_erractionrestore(err_act)
+  return
+
+9999 call psb_error_handler(err_act)
+  return
+end subroutine amg_d_ainv_solver_clone_settings

amgprec/impl/solver/amg_d_bwgs_solver_bld.f90

@@ -56,6 +56,8 @@ subroutine amg_d_bwgs_solver_bld(a,desc_a,sv,info,b,amold,vmold,imold)
   type(psb_ctxt_type) :: ctxt
   integer(psb_ipk_)   :: np, me, i, err_act, debug_unit, debug_level
   character(len=20)   :: name='d_bwgs_solver_bld', ch_err
+  integer(psb_ipk_), save :: idx_tril=-1
+  logical, parameter      :: do_timings=.true.
 
   info=psb_success_
   call psb_erractionsave(err_act)
@@ -65,6 +67,8 @@ subroutine amg_d_bwgs_solver_bld(a,desc_a,sv,info,b,amold,vmold,imold)
   call psb_info(ctxt, me, np)
   if (debug_level >= psb_debug_outer_) &
        & write(debug_unit,*) me,' ',trim(name),' start'
+  if ((do_timings).and.(idx_tril==-1))       &
+       & idx_tril = psb_get_timer_idx("BWGS_BLD: tril")
 
 
   n_row  = desc_a%get_local_rows()
@@ -77,7 +81,10 @@ subroutine amg_d_bwgs_solver_bld(a,desc_a,sv,info,b,amold,vmold,imold)
     ! This cuts out the off-diagonal part, because it's supposed to
     ! be handled by the outer Jacobi smoother.
     !
+    !write(0,*) 'Calling A%TRIL in bwgs_solver_bld'
+    if (do_timings) call psb_tic(idx_tril)
     call a%tril(sv%l,info,diag=-ione,jmax=nrow_a,u=sv%u)     
+    if (do_timings) call psb_toc(idx_tril)
 
   else
 

amgprec/impl/solver/amg_d_gs_solver_bld.f90

@@ -56,6 +56,8 @@ subroutine amg_d_gs_solver_bld(a,desc_a,sv,info,b,amold,vmold,imold)
   type(psb_ctxt_type) :: ctxt
   integer(psb_ipk_)   :: np, me, i, err_act, debug_unit, debug_level
   character(len=20)   :: name='d_gs_solver_bld', ch_err
+  integer(psb_ipk_), save :: idx_tril=-1
+  logical, parameter      :: do_timings=.true.
 
   info=psb_success_
   call psb_erractionsave(err_act)
@@ -65,6 +67,8 @@ subroutine amg_d_gs_solver_bld(a,desc_a,sv,info,b,amold,vmold,imold)
   call psb_info(ctxt, me, np)
   if (debug_level >= psb_debug_outer_) &
        & write(debug_unit,*) me,' ',trim(name),' start'
+  if ((do_timings).and.(idx_tril==-1))       &
+       & idx_tril = psb_get_timer_idx("GS_BLD: tril")
 
 
   n_row  = desc_a%get_local_rows()
@@ -76,9 +80,12 @@ subroutine amg_d_gs_solver_bld(a,desc_a,sv,info,b,amold,vmold,imold)
     !
     ! This cuts out the off-diagonal part, because it's supposed to
     ! be handled by the outer Jacobi smoother.
-    ! 
+    !
+    !write(0,*) 'Calling A%TRIL in gs_solver_bld'
+    if (do_timings) call psb_tic(idx_tril)
     call a%tril(sv%l,info,diag=izero,jmax=nrow_a,u=sv%u)
-
+    if (do_timings) call psb_toc(idx_tril)
+    !write(0,*) 'From A%TRIL in gs_solver_bld',a%get_nzeros(),sv%l%get_nzeros(),sv%u%get_nzeros()
   else
 
     info = psb_err_missing_override_method_

amgprec/impl/solver/amg_d_invk_solver_clone_settings.f90

@@ -0,0 +1,83 @@
+!  
+!
+!                             AMG4PSBLAS version 1.0
+!    MultiLevel Domain Decomposition Parallel Preconditioners Package
+!               based on PSBLAS (Parallel Sparse BLAS version 3.0)
+!
+!    (C) Copyright 2008,2009,2010,2010,2012
+!
+!                        Salvatore Filippone  University of Rome Tor Vergata
+!                        Alfredo Buttari      CNRS-IRIT, Toulouse
+!                        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 AMG4PSBLAS 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 AMG4PSBLAS 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.
+!
+!
+subroutine amg_d_invk_solver_clone_settings(sv,svout,info)
+
+  use psb_base_mod
+  use amg_d_invk_solver, amg_protect_name => amg_d_invk_solver_clone_settings
+
+  Implicit None
+
+  ! Arguments
+  class(amg_d_invk_solver_type), intent(inout)              :: sv
+  class(amg_d_base_solver_type), allocatable, intent(inout) :: svout
+  integer(psb_ipk_), intent(out)                            :: info
+  ! Local variables
+  integer(psb_ipk_) :: err_act
+
+
+  info=psb_success_
+  call psb_erractionsave(err_act)
+
+  if (allocated(svout)) then
+    call svout%free(info)
+    if (info == psb_success_) deallocate(svout, stat=info)
+  end if
+  if (info == psb_success_) &
+       & allocate(amg_d_invk_solver_type :: svout, stat=info)
+  if (info /= 0) then
+    info = psb_err_alloc_dealloc_
+    goto 9999
+  end if
+  select type(svo => svout)
+  type is (amg_d_invk_solver_type)
+    svo%fill_in  = sv%fill_in
+    svo%inv_fill = sv%inv_fill
+
+  class default
+    info = psb_err_internal_error_
+  end select
+
+  if (info /= 0) goto 9999
+
+  call psb_erractionrestore(err_act)
+  return
+
+9999 call psb_error_handler(err_act)
+  return
+end subroutine amg_d_invk_solver_clone_settings

amgprec/impl/solver/amg_d_invt_solver_clone_settings.f90

@@ -0,0 +1,85 @@
+!  
+!
+!                             AMG4PSBLAS version 1.0
+!    MultiLevel Domain Decomposition Parallel Preconditioners Package
+!               based on PSBLAS (Parallel Sparse BLAS version 3.0)
+!
+!    (C) Copyright 2008,2009,2010,2010,2012
+!
+!                        Salvatore Filippone  University of Rome Tor Vergata
+!                        Alfredo Buttari      CNRS-IRIT, Toulouse
+!                        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 AMG4PSBLAS 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 AMG4PSBLAS 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.
+!
+!
+subroutine amg_d_invt_solver_clone_settings(sv,svout,info)
+
+  use psb_base_mod
+  use amg_d_invt_solver, amg_protect_name => amg_d_invt_solver_clone_settings
+
+  Implicit None
+
+  ! Arguments
+  class(amg_d_invt_solver_type), intent(inout)              :: sv
+  class(amg_d_base_solver_type), allocatable, intent(inout) :: svout
+  integer(psb_ipk_), intent(out)                            :: info
+  ! Local variables
+  integer(psb_ipk_) :: err_act
+
+
+  info=psb_success_
+  call psb_erractionsave(err_act)
+
+  if (allocated(svout)) then
+    call svout%free(info)
+    if (info == psb_success_) deallocate(svout, stat=info)
+  end if
+  if (info == psb_success_) &
+       & allocate(amg_d_invt_solver_type :: svout, stat=info)
+  if (info /= 0) then
+    info = psb_err_alloc_dealloc_
+    goto 9999
+  end if
+  select type(svo => svout)
+  type is (amg_d_invt_solver_type)
+    svo%fill_in    = sv%fill_in
+    svo%inv_fill   = sv%inv_fill
+    svo%thresh     = sv%thresh
+    svo%inv_thresh = sv%inv_thresh
+
+  class default
+    info = psb_err_internal_error_
+  end select
+
+  if (info /= 0) goto 9999
+
+  call psb_erractionrestore(err_act)
+  return
+
+9999 call psb_error_handler(err_act)
+  return
+end subroutine amg_d_invt_solver_clone_settings

amgprec/impl/solver/amg_s_ainv_solver_clone_settings.f90

@@ -0,0 +1,84 @@
+!
+!
+!                             AMG4PSBLAS version 1.0
+!    MultiLevel Domain Decomposition Parallel Preconditioners Package
+!               based on PSBLAS (Parallel Sparse BLAS version 3.0)
+!
+!    (C) Copyright 2008,2009,2010,2010,2012
+!
+!                        Salvatore Filippone  University of Rome Tor Vergata
+!                        Alfredo Buttari      CNRS-IRIT, Toulouse
+!                        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 AMG4PSBLAS 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 AMG4PSBLAS 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.
+!
+!
+subroutine amg_s_ainv_solver_clone_settings(sv,svout,info)
+
+  use psb_base_mod
+  use amg_s_ainv_solver, amg_protect_name => amg_s_ainv_solver_clone_settings
+
+  Implicit None
+
+  ! Arguments
+  class(amg_s_ainv_solver_type), intent(inout)              :: sv
+  class(amg_s_base_solver_type), allocatable, intent(inout) :: svout
+  integer(psb_ipk_), intent(out)                            :: info
+  ! Local variables
+  integer(psb_ipk_) :: err_act
+
+
+  info=psb_success_
+  call psb_erractionsave(err_act)
+
+  if (allocated(svout)) then
+    call svout%free(info)
+    if (info == psb_success_) deallocate(svout, stat=info)
+  end if
+  if (info == psb_success_) &
+       & allocate(amg_s_ainv_solver_type :: svout, stat=info)
+  if (info /= 0) then
+    info = psb_err_alloc_dealloc_
+    goto 9999
+  end if
+  select type(svo => svout)
+  type is (amg_s_ainv_solver_type)
+    svo%alg      = sv%alg
+    svo%fill_in  = sv%fill_in
+    svo%thresh   = sv%thresh
+
+  class default
+    info = psb_err_internal_error_
+  end select
+
+  if (info /= 0) goto 9999
+
+  call psb_erractionrestore(err_act)
+  return
+
+9999 call psb_error_handler(err_act)
+  return
+end subroutine amg_s_ainv_solver_clone_settings

amgprec/impl/solver/amg_s_bwgs_solver_bld.f90

@@ -56,6 +56,8 @@ subroutine amg_s_bwgs_solver_bld(a,desc_a,sv,info,b,amold,vmold,imold)
   type(psb_ctxt_type) :: ctxt
   integer(psb_ipk_)   :: np, me, i, err_act, debug_unit, debug_level
   character(len=20)   :: name='d_bwgs_solver_bld', ch_err
+  integer(psb_ipk_), save :: idx_tril=-1
+  logical, parameter      :: do_timings=.true.
 
   info=psb_success_
   call psb_erractionsave(err_act)
@@ -65,6 +67,8 @@ subroutine amg_s_bwgs_solver_bld(a,desc_a,sv,info,b,amold,vmold,imold)
   call psb_info(ctxt, me, np)
   if (debug_level >= psb_debug_outer_) &
        & write(debug_unit,*) me,' ',trim(name),' start'
+  if ((do_timings).and.(idx_tril==-1))       &
+       & idx_tril = psb_get_timer_idx("BWGS_BLD: tril")
 
 
   n_row  = desc_a%get_local_rows()
@@ -77,7 +81,10 @@ subroutine amg_s_bwgs_solver_bld(a,desc_a,sv,info,b,amold,vmold,imold)
     ! This cuts out the off-diagonal part, because it's supposed to
     ! be handled by the outer Jacobi smoother.
     !
+    !write(0,*) 'Calling A%TRIL in bwgs_solver_bld'
+    if (do_timings) call psb_tic(idx_tril)
     call a%tril(sv%l,info,diag=-ione,jmax=nrow_a,u=sv%u)     
+    if (do_timings) call psb_toc(idx_tril)
 
   else
 

amgprec/impl/solver/amg_s_gs_solver_bld.f90

@@ -56,6 +56,8 @@ subroutine amg_s_gs_solver_bld(a,desc_a,sv,info,b,amold,vmold,imold)
   type(psb_ctxt_type) :: ctxt
   integer(psb_ipk_)   :: np, me, i, err_act, debug_unit, debug_level
   character(len=20)   :: name='s_gs_solver_bld', ch_err
+  integer(psb_ipk_), save :: idx_tril=-1
+  logical, parameter      :: do_timings=.true.
 
   info=psb_success_
   call psb_erractionsave(err_act)
@@ -65,6 +67,8 @@ subroutine amg_s_gs_solver_bld(a,desc_a,sv,info,b,amold,vmold,imold)
   call psb_info(ctxt, me, np)
   if (debug_level >= psb_debug_outer_) &
        & write(debug_unit,*) me,' ',trim(name),' start'
+  if ((do_timings).and.(idx_tril==-1))       &
+       & idx_tril = psb_get_timer_idx("GS_BLD: tril")
 
 
   n_row  = desc_a%get_local_rows()
@@ -76,9 +80,12 @@ subroutine amg_s_gs_solver_bld(a,desc_a,sv,info,b,amold,vmold,imold)
     !
     ! This cuts out the off-diagonal part, because it's supposed to
     ! be handled by the outer Jacobi smoother.
-    ! 
+    !
+    !write(0,*) 'Calling A%TRIL in gs_solver_bld'
+    if (do_timings) call psb_tic(idx_tril)
     call a%tril(sv%l,info,diag=izero,jmax=nrow_a,u=sv%u)
-
+    if (do_timings) call psb_toc(idx_tril)
+    !write(0,*) 'From A%TRIL in gs_solver_bld',a%get_nzeros(),sv%l%get_nzeros(),sv%u%get_nzeros()
   else
 
     info = psb_err_missing_override_method_

amgprec/impl/solver/amg_s_invk_solver_clone_settings.f90

@@ -0,0 +1,83 @@
+!  
+!
+!                             AMG4PSBLAS version 1.0
+!    MultiLevel Domain Decomposition Parallel Preconditioners Package
+!               based on PSBLAS (Parallel Sparse BLAS version 3.0)
+!
+!    (C) Copyright 2008,2009,2010,2010,2012
+!
+!                        Salvatore Filippone  University of Rome Tor Vergata
+!                        Alfredo Buttari      CNRS-IRIT, Toulouse
+!                        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 AMG4PSBLAS 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 AMG4PSBLAS 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.
+!
+!
+subroutine amg_s_invk_solver_clone_settings(sv,svout,info)
+
+  use psb_base_mod
+  use amg_s_invk_solver, amg_protect_name => amg_s_invk_solver_clone_settings
+
+  Implicit None
+
+  ! Arguments
+  class(amg_s_invk_solver_type), intent(inout)              :: sv
+  class(amg_s_base_solver_type), allocatable, intent(inout) :: svout
+  integer(psb_ipk_), intent(out)                            :: info
+  ! Local variables
+  integer(psb_ipk_) :: err_act
+
+
+  info=psb_success_
+  call psb_erractionsave(err_act)
+
+  if (allocated(svout)) then
+    call svout%free(info)
+    if (info == psb_success_) deallocate(svout, stat=info)
+  end if
+  if (info == psb_success_) &
+       & allocate(amg_s_invk_solver_type :: svout, stat=info)
+  if (info /= 0) then
+    info = psb_err_alloc_dealloc_
+    goto 9999
+  end if
+  select type(svo => svout)
+  type is (amg_s_invk_solver_type)
+    svo%fill_in  = sv%fill_in
+    svo%inv_fill = sv%inv_fill
+
+  class default
+    info = psb_err_internal_error_
+  end select
+
+  if (info /= 0) goto 9999
+
+  call psb_erractionrestore(err_act)
+  return
+
+9999 call psb_error_handler(err_act)
+  return
+end subroutine amg_s_invk_solver_clone_settings

amgprec/impl/solver/amg_s_invt_solver_clone_settings.f90

@@ -0,0 +1,85 @@
+!  
+!
+!                             AMG4PSBLAS version 1.0
+!    MultiLevel Domain Decomposition Parallel Preconditioners Package
+!               based on PSBLAS (Parallel Sparse BLAS version 3.0)
+!
+!    (C) Copyright 2008,2009,2010,2010,2012
+!
+!                        Salvatore Filippone  University of Rome Tor Vergata
+!                        Alfredo Buttari      CNRS-IRIT, Toulouse
+!                        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 AMG4PSBLAS 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 AMG4PSBLAS 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.
+!
+!
+subroutine amg_s_invt_solver_clone_settings(sv,svout,info)
+
+  use psb_base_mod
+  use amg_s_invt_solver, amg_protect_name => amg_s_invt_solver_clone_settings
+
+  Implicit None
+
+  ! Arguments
+  class(amg_s_invt_solver_type), intent(inout)              :: sv
+  class(amg_s_base_solver_type), allocatable, intent(inout) :: svout
+  integer(psb_ipk_), intent(out)                            :: info
+  ! Local variables
+  integer(psb_ipk_) :: err_act
+
+
+  info=psb_success_
+  call psb_erractionsave(err_act)
+
+  if (allocated(svout)) then
+    call svout%free(info)
+    if (info == psb_success_) deallocate(svout, stat=info)
+  end if
+  if (info == psb_success_) &
+       & allocate(amg_s_invt_solver_type :: svout, stat=info)
+  if (info /= 0) then
+    info = psb_err_alloc_dealloc_
+    goto 9999
+  end if
+  select type(svo => svout)
+  type is (amg_s_invt_solver_type)
+    svo%fill_in    = sv%fill_in
+    svo%inv_fill   = sv%inv_fill
+    svo%thresh     = sv%thresh
+    svo%inv_thresh = sv%inv_thresh
+
+  class default
+    info = psb_err_internal_error_
+  end select
+
+  if (info /= 0) goto 9999
+
+  call psb_erractionrestore(err_act)
+  return
+
+9999 call psb_error_handler(err_act)
+  return
+end subroutine amg_s_invt_solver_clone_settings

amgprec/impl/solver/amg_z_ainv_solver_clone_settings.f90

@@ -0,0 +1,84 @@
+!
+!
+!                             AMG4PSBLAS version 1.0
+!    MultiLevel Domain Decomposition Parallel Preconditioners Package
+!               based on PSBLAS (Parallel Sparse BLAS version 3.0)
+!
+!    (C) Copyright 2008,2009,2010,2010,2012
+!
+!                        Salvatore Filippone  University of Rome Tor Vergata
+!                        Alfredo Buttari      CNRS-IRIT, Toulouse
+!                        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 AMG4PSBLAS 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 AMG4PSBLAS 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.
+!
+!
+subroutine amg_z_ainv_solver_clone_settings(sv,svout,info)
+
+  use psb_base_mod
+  use amg_z_ainv_solver, amg_protect_name => amg_z_ainv_solver_clone_settings
+
+  Implicit None
+
+  ! Arguments
+  class(amg_z_ainv_solver_type), intent(inout)              :: sv
+  class(amg_z_base_solver_type), allocatable, intent(inout) :: svout
+  integer(psb_ipk_), intent(out)                            :: info
+  ! Local variables
+  integer(psb_ipk_) :: err_act
+
+
+  info=psb_success_
+  call psb_erractionsave(err_act)
+
+  if (allocated(svout)) then
+    call svout%free(info)
+    if (info == psb_success_) deallocate(svout, stat=info)
+  end if
+  if (info == psb_success_) &
+       & allocate(amg_z_ainv_solver_type :: svout, stat=info)
+  if (info /= 0) then
+    info = psb_err_alloc_dealloc_
+    goto 9999
+  end if
+  select type(svo => svout)
+  type is (amg_z_ainv_solver_type)
+    svo%alg      = sv%alg
+    svo%fill_in  = sv%fill_in
+    svo%thresh   = sv%thresh
+
+  class default
+    info = psb_err_internal_error_
+  end select
+
+  if (info /= 0) goto 9999
+
+  call psb_erractionrestore(err_act)
+  return
+
+9999 call psb_error_handler(err_act)
+  return
+end subroutine amg_z_ainv_solver_clone_settings

amgprec/impl/solver/amg_z_bwgs_solver_bld.f90

@@ -56,6 +56,8 @@ subroutine amg_z_bwgs_solver_bld(a,desc_a,sv,info,b,amold,vmold,imold)
   type(psb_ctxt_type) :: ctxt
   integer(psb_ipk_)   :: np, me, i, err_act, debug_unit, debug_level
   character(len=20)   :: name='d_bwgs_solver_bld', ch_err
+  integer(psb_ipk_), save :: idx_tril=-1
+  logical, parameter      :: do_timings=.true.
 
   info=psb_success_
   call psb_erractionsave(err_act)
@@ -65,6 +67,8 @@ subroutine amg_z_bwgs_solver_bld(a,desc_a,sv,info,b,amold,vmold,imold)
   call psb_info(ctxt, me, np)
   if (debug_level >= psb_debug_outer_) &
        & write(debug_unit,*) me,' ',trim(name),' start'
+  if ((do_timings).and.(idx_tril==-1))       &
+       & idx_tril = psb_get_timer_idx("BWGS_BLD: tril")
 
 
   n_row  = desc_a%get_local_rows()
@@ -77,7 +81,10 @@ subroutine amg_z_bwgs_solver_bld(a,desc_a,sv,info,b,amold,vmold,imold)
     ! This cuts out the off-diagonal part, because it's supposed to
     ! be handled by the outer Jacobi smoother.
     !
+    !write(0,*) 'Calling A%TRIL in bwgs_solver_bld'
+    if (do_timings) call psb_tic(idx_tril)
     call a%tril(sv%l,info,diag=-ione,jmax=nrow_a,u=sv%u)     
+    if (do_timings) call psb_toc(idx_tril)
 
   else
 

amgprec/impl/solver/amg_z_gs_solver_bld.f90

@@ -56,6 +56,8 @@ subroutine amg_z_gs_solver_bld(a,desc_a,sv,info,b,amold,vmold,imold)
   type(psb_ctxt_type) :: ctxt
   integer(psb_ipk_)   :: np, me, i, err_act, debug_unit, debug_level
   character(len=20)   :: name='z_gs_solver_bld', ch_err
+  integer(psb_ipk_), save :: idx_tril=-1
+  logical, parameter      :: do_timings=.true.
 
   info=psb_success_
   call psb_erractionsave(err_act)
@@ -65,6 +67,8 @@ subroutine amg_z_gs_solver_bld(a,desc_a,sv,info,b,amold,vmold,imold)
   call psb_info(ctxt, me, np)
   if (debug_level >= psb_debug_outer_) &
        & write(debug_unit,*) me,' ',trim(name),' start'
+  if ((do_timings).and.(idx_tril==-1))       &
+       & idx_tril = psb_get_timer_idx("GS_BLD: tril")
 
 
   n_row  = desc_a%get_local_rows()
@@ -76,9 +80,12 @@ subroutine amg_z_gs_solver_bld(a,desc_a,sv,info,b,amold,vmold,imold)
     !
     ! This cuts out the off-diagonal part, because it's supposed to
     ! be handled by the outer Jacobi smoother.
-    ! 
+    !
+    !write(0,*) 'Calling A%TRIL in gs_solver_bld'
+    if (do_timings) call psb_tic(idx_tril)
     call a%tril(sv%l,info,diag=izero,jmax=nrow_a,u=sv%u)
-
+    if (do_timings) call psb_toc(idx_tril)
+    !write(0,*) 'From A%TRIL in gs_solver_bld',a%get_nzeros(),sv%l%get_nzeros(),sv%u%get_nzeros()
   else
 
     info = psb_err_missing_override_method_

amgprec/impl/solver/amg_z_invk_solver_clone_settings.f90

@@ -0,0 +1,83 @@
+!  
+!
+!                             AMG4PSBLAS version 1.0
+!    MultiLevel Domain Decomposition Parallel Preconditioners Package
+!               based on PSBLAS (Parallel Sparse BLAS version 3.0)
+!
+!    (C) Copyright 2008,2009,2010,2010,2012
+!
+!                        Salvatore Filippone  University of Rome Tor Vergata
+!                        Alfredo Buttari      CNRS-IRIT, Toulouse
+!                        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 AMG4PSBLAS 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 AMG4PSBLAS 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.
+!
+!
+subroutine amg_z_invk_solver_clone_settings(sv,svout,info)
+
+  use psb_base_mod
+  use amg_z_invk_solver, amg_protect_name => amg_z_invk_solver_clone_settings
+
+  Implicit None
+
+  ! Arguments
+  class(amg_z_invk_solver_type), intent(inout)              :: sv
+  class(amg_z_base_solver_type), allocatable, intent(inout) :: svout
+  integer(psb_ipk_), intent(out)                            :: info
+  ! Local variables
+  integer(psb_ipk_) :: err_act
+
+
+  info=psb_success_
+  call psb_erractionsave(err_act)
+
+  if (allocated(svout)) then
+    call svout%free(info)
+    if (info == psb_success_) deallocate(svout, stat=info)
+  end if
+  if (info == psb_success_) &
+       & allocate(amg_z_invk_solver_type :: svout, stat=info)
+  if (info /= 0) then
+    info = psb_err_alloc_dealloc_
+    goto 9999
+  end if
+  select type(svo => svout)
+  type is (amg_z_invk_solver_type)
+    svo%fill_in  = sv%fill_in
+    svo%inv_fill = sv%inv_fill
+
+  class default
+    info = psb_err_internal_error_
+  end select
+
+  if (info /= 0) goto 9999
+
+  call psb_erractionrestore(err_act)
+  return
+
+9999 call psb_error_handler(err_act)
+  return
+end subroutine amg_z_invk_solver_clone_settings

amgprec/impl/solver/amg_z_invt_solver_clone_settings.f90

@@ -0,0 +1,85 @@
+!  
+!
+!                             AMG4PSBLAS version 1.0
+!    MultiLevel Domain Decomposition Parallel Preconditioners Package
+!               based on PSBLAS (Parallel Sparse BLAS version 3.0)
+!
+!    (C) Copyright 2008,2009,2010,2010,2012
+!
+!                        Salvatore Filippone  University of Rome Tor Vergata
+!                        Alfredo Buttari      CNRS-IRIT, Toulouse
+!                        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 AMG4PSBLAS 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 AMG4PSBLAS 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.
+!
+!
+subroutine amg_z_invt_solver_clone_settings(sv,svout,info)
+
+  use psb_base_mod
+  use amg_z_invt_solver, amg_protect_name => amg_z_invt_solver_clone_settings
+
+  Implicit None
+
+  ! Arguments
+  class(amg_z_invt_solver_type), intent(inout)              :: sv
+  class(amg_z_base_solver_type), allocatable, intent(inout) :: svout
+  integer(psb_ipk_), intent(out)                            :: info
+  ! Local variables
+  integer(psb_ipk_) :: err_act
+
+
+  info=psb_success_
+  call psb_erractionsave(err_act)
+
+  if (allocated(svout)) then
+    call svout%free(info)
+    if (info == psb_success_) deallocate(svout, stat=info)
+  end if
+  if (info == psb_success_) &
+       & allocate(amg_z_invt_solver_type :: svout, stat=info)
+  if (info /= 0) then
+    info = psb_err_alloc_dealloc_
+    goto 9999
+  end if
+  select type(svo => svout)
+  type is (amg_z_invt_solver_type)
+    svo%fill_in    = sv%fill_in
+    svo%inv_fill   = sv%inv_fill
+    svo%thresh     = sv%thresh
+    svo%inv_thresh = sv%inv_thresh
+
+  class default
+    info = psb_err_internal_error_
+  end select
+
+  if (info /= 0) goto 9999
+
+  call psb_erractionrestore(err_act)
+  return
+
+9999 call psb_error_handler(err_act)
+  return
+end subroutine amg_z_invt_solver_clone_settings

config/pac.m4

@@ -608,8 +608,12 @@ dnl Maybe new structure with UMFPACK UFconfig AMD?
 fi
 
 
-if test "x$pac_umf_header_ok" == "xyes" ; then 
-      UMF_LIBS="$amg4psblas_cv_umfpack $UMF_LIBDIR"
+if test "x$pac_umf_header_ok" == "xyes" ; then
+      if  test "x$UMF_LIBDIR" == "x" ; then
+	  UMF_LIBS="$amg4psblas_cv_umfpack"
+      else
+          UMF_LIBS="$UMF_LIBDIR $amg4psblas_cv_umfpack"
+      fi
       LIBS="$UMF_LIBS -lm $LIBS $EXTRA_LIBS";
       AC_MSG_CHECKING([for umfpack_di_symbolic in $UMF_LIBS])
       AC_TRY_LINK_FUNC(umfpack_di_symbolic, 
@@ -619,8 +623,8 @@ if test "x$pac_umf_header_ok" == "xyes" ; then
      if test "x$pac_umf_lib_ok" == "xno" ; then 
         dnl Maybe Lib or lib? 
         UMF_LIBDIR="-L$amg4psblas_cv_umfpackdir/Lib -L$amg4psblas_cv_umfpackdir/lib"
-        UMF_LIBS="$amg4psblas_cv_umfpack $UMF_LIBDIR -lm $SAVE_LIBS  $EXTRA_LIBS"
-        LIBS="$UMF_LIBS"
+        UMF_LIBS="$amg4psblas_cv_umfpack $UMF_LIBDIR -lm"
+        LIBS="$UMF_LIBS $SAVE_LIBS  $EXTRA_LIBS "
         
       AC_MSG_CHECKING([for umfpack_di_symbolic in $UMF_LIBS])
       AC_TRY_LINK_FUNC(umfpack_di_symbolic, 
@@ -631,8 +635,8 @@ if test "x$pac_umf_header_ok" == "xyes" ; then
      if test "x$pac_umf_lib_ok" == "xno" ; then 
         dnl Maybe UMFPACK/Lib? 
         UMF_LIBDIR="-L$amg4psblas_cv_umfpackdir/AMD/Lib -L$amg4psblas_cv_umfpackdir/UMFPACK/Lib"
-        UMF_LIBS="$amg4psblas_cv_umfpack $UMF_LIBDIR -lm $SAVE_LIBS $EXTRA_LIBS"
-             LIBS="$UMF_LIBS"
+        UMF_LIBS="$amg4psblas_cv_umfpack $UMF_LIBDIR -lm"
+        LIBS="$UMF_LIBS $SAVE_LIBS $EXTRA_LIBS"
       AC_MSG_CHECKING([for umfpack_di_symbolic in $UMF_LIBS])
       AC_TRY_LINK_FUNC(umfpack_di_symbolic, 
        [amg4psblas_cv_have_umfpack=yes;pac_umf_lib_ok=yes; ],
@@ -716,7 +720,11 @@ dnl Maybe Include subdirs?
 fi
 
 if test "x$pac_slu_header_ok" == "xyes" ; then 
- SLU_LIBS="$amg4psblas_cv_superlu $SLU_LIBS"
+   if  test "x$SLU_LIBS" == "x" ; then
+       SLU_LIBS="$amg4psblas_cv_superlu"
+   else
+     SLU_LIBS="$SLU_LIBS $amg4psblas_cv_superlu"
+   fi  
  LIBS="$SLU_LIBS -lm $save_LIBS";
  AC_MSG_CHECKING([for superlu_malloc in $SLU_LIBS])
  AC_TRY_LINK_FUNC(superlu_malloc, 
@@ -813,7 +821,7 @@ elif test "x$amg4psblas_cv_superludistdir" != "x"; then
    SLUDIST_LIBS="-L$amg4psblas_cv_superludistdir"
 fi
 
-LIBS="$SLUDIST_LIBS $save_LIBS"
+LIBS="$save_LIBS $SLUDIST_LIBS"
 CPPFLAGS="$SLUDIST_INCLUDES $save_CPPFLAGS"
 
 AC_CHECK_HEADERS([superlu_ddefs.h],
@@ -841,7 +849,11 @@ dnl Maybe Include subdirs?
 fi
 
 if test "x$pac_sludist_header_ok" == "xyes" ; then 
-      SLUDIST_LIBS="$amg4psblas_cv_superludist $SLUDIST_LIBS"
+      if  test "x$SLUDIST_LIBS" == "x" ; then
+         SLUDIST_LIBS="$amg4psblas_cv_superludist"
+      else
+         SLUDIST_LIBS="$SLUDIST_LIBS $amg4psblas_cv_superludist"
+      fi  
       LIBS="$SLUDIST_LIBS -lm $save_LIBS";
       AC_MSG_CHECKING([for superlu_malloc_dist in $SLUDIST_LIBS])
       AC_TRY_LINK_FUNC(superlu_malloc_dist, 
@@ -1150,7 +1162,11 @@ if test "x$pac_mumps_fmods_ok" == "xno" ; then
 fi   
 
 if test "x$pac_mumps_fmods_ok" == "xyes" || test "x$pac_mumps_fincs_ok" == "xyes" ; then  
-    MUMPS_LIBS="$amg4psblas_cv_mumps $MUMPS_LIBS"
+    if  test "x$MUMPS_LIBS" == "x" ; then
+        MUMPS_LIBS="$amg4psblas_cv_mumps"
+    else
+      MUMPS_LIBS="$MUMPS_LIBS $amg4psblas_cv_mumps"
+    fi  
     LIBS="$MUMPS_LIBS  $save_LIBS  $EXTRA_LIBS";
     AC_MSG_CHECKING([for dmumps in $MUMPS_LIBS])
 			AC_TRY_LINK_FUNC(dmumps, 

configure.ac

@@ -98,6 +98,7 @@ AC_PREREQ(2.59)
 #
 #
 AC_PROG_INSTALL
+INSTALL="${INSTALL} -p"
 
 AC_MSG_CHECKING([where to install])
 case $prefix in  

install-sh

@@ -1,251 +1,541 @@
 #!/bin/sh
-#
 # install - install a program, script, or datafile
-# This comes from X11R5 (mit/util/scripts/install.sh).
+
+scriptversion=2020-11-14.01; # UTC
+
+# This originates from X11R5 (mit/util/scripts/install.sh), which was
+# later released in X11R6 (xc/config/util/install.sh) with the
+# following copyright and license.
 #
-# Copyright 1991 by the Massachusetts Institute of Technology
+# Copyright (C) 1994 X Consortium
 #
-# Permission to use, copy, modify, distribute, and sell this software and its
-# documentation for any purpose is hereby granted without fee, provided that
-# the above copyright notice appear in all copies and that both that
-# copyright notice and this permission notice appear in supporting
-# documentation, and that the name of M.I.T. not be used in advertising or
-# publicity pertaining to distribution of the software without specific,
-# written prior permission.  M.I.T. makes no representations about the
-# suitability of this software for any purpose.  It is provided "as is"
-# without express or implied warranty.
+# Permission is hereby granted, free of charge, to any person obtaining a copy
+# of this software and associated documentation files (the "Software"), to
+# deal in the Software without restriction, including without limitation the
+# rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
+# sell copies of the Software, and to permit persons to whom the Software is
+# furnished to do so, subject to the following conditions:
+#
+# The above copyright notice and this permission notice shall be included in
+# all copies or substantial portions of the Software.
+#
+# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+# IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL THE
+# X CONSORTIUM BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+# AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNEC-
+# TION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+#
+# Except as contained in this notice, the name of the X Consortium shall not
+# be used in advertising or otherwise to promote the sale, use or other deal-
+# ings in this Software without prior written authorization from the X Consor-
+# tium.
+#
+#
+# FSF changes to this file are in the public domain.
 #
 # Calling this script install-sh is preferred over install.sh, to prevent
-# `make' implicit rules from creating a file called install from it
+# 'make' implicit rules from creating a file called install from it
 # when there is no Makefile.
 #
 # This script is compatible with the BSD install script, but was written
-# from scratch.  It can only install one file at a time, a restriction
-# shared with many OS's install programs.
+# from scratch.
 
-
-# set DOITPROG to echo to test this script
-
-# Don't use :- since 4.3BSD and earlier shells don't like it.
-doit="${DOITPROG-}"
-
-
-# put in absolute paths if you don't have them in your path; or use env. vars.
-
-mvprog="${MVPROG-mv}"
-cpprog="${CPPROG-cp}"
-chmodprog="${CHMODPROG-chmod}"
-chownprog="${CHOWNPROG-chown}"
-chgrpprog="${CHGRPPROG-chgrp}"
-stripprog="${STRIPPROG-strip}"
-rmprog="${RMPROG-rm}"
-mkdirprog="${MKDIRPROG-mkdir}"
-
-transformbasename=""
-transform_arg=""
-instcmd="$mvprog"
-chmodcmd="$chmodprog 0755"
-chowncmd=""
-chgrpcmd=""
-stripcmd=""
-rmcmd="$rmprog -f"
-mvcmd="$mvprog"
-src=""
-dst=""
-dir_arg=""
-
-while [ x"$1" != x ]; do
-    case $1 in
-	-c) instcmd="$cpprog"
-	    shift
-	    continue;;
-
-	-d) dir_arg=true
-	    shift
-	    continue;;
-
-	-m) chmodcmd="$chmodprog $2"
-	    shift
-	    shift
-	    continue;;
-
-	-o) chowncmd="$chownprog $2"
-	    shift
-	    shift
-	    continue;;
-
-	-g) chgrpcmd="$chgrpprog $2"
-	    shift
-	    shift
-	    continue;;
-
-	-s) stripcmd="$stripprog"
-	    shift
-	    continue;;
-
-	-t=*) transformarg=`echo $1 | sed 's/-t=//'`
-	    shift
-	    continue;;
-
-	-b=*) transformbasename=`echo $1 | sed 's/-b=//'`
-	    shift
-	    continue;;
-
-	*)  if [ x"$src" = x ]
-	    then
-		src=$1
-	    else
-		# this colon is to work around a 386BSD /bin/sh bug
-		:
-		dst=$1
-	    fi
-	    shift
-	    continue;;
-    esac
-done
-
-if [ x"$src" = x ]
-then
-	echo "install:	no input file specified"
-	exit 1
-else
-	true
-fi
-
-if [ x"$dir_arg" != x ]; then
-	dst=$src
-	src=""
-	
-	if [ -d $dst ]; then
-		instcmd=:
-		chmodcmd=""
-	else
-		instcmd=mkdir
-	fi
-else
-
-# Waiting for this to be detected by the "$instcmd $src $dsttmp" command
-# might cause directories to be created, which would be especially bad 
-# if $src (and thus $dsttmp) contains '*'.
-
-	if [ -f $src -o -d $src ]
-	then
-		true
-	else
-		echo "install:  $src does not exist"
-		exit 1
-	fi
-	
-	if [ x"$dst" = x ]
-	then
-		echo "install:	no destination specified"
-		exit 1
-	else
-		true
-	fi
-
-# If destination is a directory, append the input filename; if your system
-# does not like double slashes in filenames, you may need to add some logic
-
-	if [ -d $dst ]
-	then
-		dst="$dst"/`basename $src`
-	else
-		true
-	fi
-fi
-
-## this sed command emulates the dirname command
-dstdir=`echo $dst | sed -e 's,[^/]*$,,;s,/$,,;s,^$,.,'`
-
-# Make sure that the destination directory exists.
-#  this part is taken from Noah Friedman's mkinstalldirs script
-
-# Skip lots of stat calls in the usual case.
-if [ ! -d "$dstdir" ]; then
-defaultIFS='	
+tab='	'
+nl='
 '
-IFS="${IFS-${defaultIFS}}"
+IFS=" $tab$nl"
 
-oIFS="${IFS}"
-# Some sh's can't handle IFS=/ for some reason.
-IFS='%'
-set - `echo ${dstdir} | sed -e 's@/@%@g' -e 's@^%@/@'`
-IFS="${oIFS}"
+# Set DOITPROG to "echo" to test this script.
 
-pathcomp=''
+doit=${DOITPROG-}
+doit_exec=${doit:-exec}
 
-while [ $# -ne 0 ] ; do
-	pathcomp="${pathcomp}${1}"
-	shift
+# Put in absolute file names if you don't have them in your path;
+# or use environment vars.
 
-	if [ ! -d "${pathcomp}" ] ;
-        then
-		$mkdirprog "${pathcomp}"
-	else
-		true
-	fi
+chgrpprog=${CHGRPPROG-chgrp}
+chmodprog=${CHMODPROG-chmod}
+chownprog=${CHOWNPROG-chown}
+cmpprog=${CMPPROG-cmp}
+cpprog=${CPPROG-cp}
+mkdirprog=${MKDIRPROG-mkdir}
+mvprog=${MVPROG-mv}
+rmprog=${RMPROG-rm}
+stripprog=${STRIPPROG-strip}
 
-	pathcomp="${pathcomp}/"
+posix_mkdir=
+
+# Desired mode of installed file.
+mode=0755
+
+# Create dirs (including intermediate dirs) using mode 755.
+# This is like GNU 'install' as of coreutils 8.32 (2020).
+mkdir_umask=22
+
+backupsuffix=
+chgrpcmd=
+chmodcmd=$chmodprog
+chowncmd=
+mvcmd=$mvprog
+rmcmd="$rmprog -f"
+stripcmd=
+
+src=
+dst=
+dir_arg=
+dst_arg=
+
+copy_on_change=false
+is_target_a_directory=possibly
+
+usage="\
+Usage: $0 [OPTION]... [-T] SRCFILE DSTFILE
+   or: $0 [OPTION]... SRCFILES... DIRECTORY
+   or: $0 [OPTION]... -t DIRECTORY SRCFILES...
+   or: $0 [OPTION]... -d DIRECTORIES...
+
+In the 1st form, copy SRCFILE to DSTFILE.
+In the 2nd and 3rd, copy all SRCFILES to DIRECTORY.
+In the 4th, create DIRECTORIES.
+
+Options:
+     --help     display this help and exit.
+     --version  display version info and exit.
+
+  -c            (ignored)
+  -C            install only if different (preserve data modification time)
+  -d            create directories instead of installing files.
+  -g GROUP      $chgrpprog installed files to GROUP.
+  -m MODE       $chmodprog installed files to MODE.
+  -o USER       $chownprog installed files to USER.
+  -p            pass -p to $cpprog.
+  -s            $stripprog installed files.
+  -S SUFFIX     attempt to back up existing files, with suffix SUFFIX.
+  -t DIRECTORY  install into DIRECTORY.
+  -T            report an error if DSTFILE is a directory.
+
+Environment variables override the default commands:
+  CHGRPPROG CHMODPROG CHOWNPROG CMPPROG CPPROG MKDIRPROG MVPROG
+  RMPROG STRIPPROG
+
+By default, rm is invoked with -f; when overridden with RMPROG,
+it's up to you to specify -f if you want it.
+
+If -S is not specified, no backups are attempted.
+
+Email bug reports to bug-automake@gnu.org.
+Automake home page: https://www.gnu.org/software/automake/
+"
+
+while test $# -ne 0; do
+  case $1 in
+    -c) ;;
+
+    -C) copy_on_change=true;;
+
+    -d) dir_arg=true;;
+
+    -g) chgrpcmd="$chgrpprog $2"
+        shift;;
+
+    --help) echo "$usage"; exit $?;;
+
+    -m) mode=$2
+        case $mode in
+          *' '* | *"$tab"* | *"$nl"* | *'*'* | *'?'* | *'['*)
+            echo "$0: invalid mode: $mode" >&2
+            exit 1;;
+        esac
+        shift;;
+
+    -o) chowncmd="$chownprog $2"
+        shift;;
+
+    -p) cpprog="$cpprog -p";;
+
+    -s) stripcmd=$stripprog;;
+
+    -S) backupsuffix="$2"
+        shift;;
+
+    -t)
+        is_target_a_directory=always
+        dst_arg=$2
+        # Protect names problematic for 'test' and other utilities.
+        case $dst_arg in
+          -* | [=\(\)!]) dst_arg=./$dst_arg;;
+        esac
+        shift;;
+
+    -T) is_target_a_directory=never;;
+
+    --version) echo "$0 $scriptversion"; exit $?;;
+
+    --) shift
+        break;;
+
+    -*) echo "$0: invalid option: $1" >&2
+        exit 1;;
+
+    *)  break;;
+  esac
+  shift
 done
+
+# We allow the use of options -d and -T together, by making -d
+# take the precedence; this is for compatibility with GNU install.
+
+if test -n "$dir_arg"; then
+  if test -n "$dst_arg"; then
+    echo "$0: target directory not allowed when installing a directory." >&2
+    exit 1
+  fi
 fi
 
-if [ x"$dir_arg" != x ]
-then
-	$doit $instcmd $dst &&
+if test $# -ne 0 && test -z "$dir_arg$dst_arg"; then
+  # When -d is used, all remaining arguments are directories to create.
+  # When -t is used, the destination is already specified.
+  # Otherwise, the last argument is the destination.  Remove it from $@.
+  for arg
+  do
+    if test -n "$dst_arg"; then
+      # $@ is not empty: it contains at least $arg.
+      set fnord "$@" "$dst_arg"
+      shift # fnord
+    fi
+    shift # arg
+    dst_arg=$arg
+    # Protect names problematic for 'test' and other utilities.
+    case $dst_arg in
+      -* | [=\(\)!]) dst_arg=./$dst_arg;;
+    esac
+  done
+fi
 
-	if [ x"$chowncmd" != x ]; then $doit $chowncmd $dst; else true ; fi &&
-	if [ x"$chgrpcmd" != x ]; then $doit $chgrpcmd $dst; else true ; fi &&
-	if [ x"$stripcmd" != x ]; then $doit $stripcmd $dst; else true ; fi &&
-	if [ x"$chmodcmd" != x ]; then $doit $chmodcmd $dst; else true ; fi
-else
+if test $# -eq 0; then
+  if test -z "$dir_arg"; then
+    echo "$0: no input file specified." >&2
+    exit 1
+  fi
+  # It's OK to call 'install-sh -d' without argument.
+  # This can happen when creating conditional directories.
+  exit 0
+fi
 
-# If we're going to rename the final executable, determine the name now.
+if test -z "$dir_arg"; then
+  if test $# -gt 1 || test "$is_target_a_directory" = always; then
+    if test ! -d "$dst_arg"; then
+      echo "$0: $dst_arg: Is not a directory." >&2
+      exit 1
+    fi
+  fi
+fi
 
-	if [ x"$transformarg" = x ] 
+if test -z "$dir_arg"; then
+  do_exit='(exit $ret); exit $ret'
+  trap "ret=129; $do_exit" 1
+  trap "ret=130; $do_exit" 2
+  trap "ret=141; $do_exit" 13
+  trap "ret=143; $do_exit" 15
+
+  # Set umask so as not to create temps with too-generous modes.
+  # However, 'strip' requires both read and write access to temps.
+  case $mode in
+    # Optimize common cases.
+    *644) cp_umask=133;;
+    *755) cp_umask=22;;
+
+    *[0-7])
+      if test -z "$stripcmd"; then
+        u_plus_rw=
+      else
+        u_plus_rw='% 200'
+      fi
+      cp_umask=`expr '(' 777 - $mode % 1000 ')' $u_plus_rw`;;
+    *)
+      if test -z "$stripcmd"; then
+        u_plus_rw=
+      else
+        u_plus_rw=,u+rw
+      fi
+      cp_umask=$mode$u_plus_rw;;
+  esac
+fi
+
+for src
+do
+  # Protect names problematic for 'test' and other utilities.
+  case $src in
+    -* | [=\(\)!]) src=./$src;;
+  esac
+
+  if test -n "$dir_arg"; then
+    dst=$src
+    dstdir=$dst
+    test -d "$dstdir"
+    dstdir_status=$?
+    # Don't chown directories that already exist.
+    if test $dstdir_status = 0; then
+      chowncmd=""
+    fi
+  else
+
+    # Waiting for this to be detected by the "$cpprog $src $dsttmp" command
+    # might cause directories to be created, which would be especially bad
+    # if $src (and thus $dsttmp) contains '*'.
+    if test ! -f "$src" && test ! -d "$src"; then
+      echo "$0: $src does not exist." >&2
+      exit 1
+    fi
+
+    if test -z "$dst_arg"; then
+      echo "$0: no destination specified." >&2
+      exit 1
+    fi
+    dst=$dst_arg
+
+    # If destination is a directory, append the input filename.
+    if test -d "$dst"; then
+      if test "$is_target_a_directory" = never; then
+        echo "$0: $dst_arg: Is a directory" >&2
+        exit 1
+      fi
+      dstdir=$dst
+      dstbase=`basename "$src"`
+      case $dst in
+	*/) dst=$dst$dstbase;;
+	*)  dst=$dst/$dstbase;;
+      esac
+      dstdir_status=0
+    else
+      dstdir=`dirname "$dst"`
+      test -d "$dstdir"
+      dstdir_status=$?
+    fi
+  fi
+
+  case $dstdir in
+    */) dstdirslash=$dstdir;;
+    *)  dstdirslash=$dstdir/;;
+  esac
+
+  obsolete_mkdir_used=false
+
+  if test $dstdir_status != 0; then
+    case $posix_mkdir in
+      '')
+        # With -d, create the new directory with the user-specified mode.
+        # Otherwise, rely on $mkdir_umask.
+        if test -n "$dir_arg"; then
+          mkdir_mode=-m$mode
+        else
+          mkdir_mode=
+        fi
+
+        posix_mkdir=false
+	# The $RANDOM variable is not portable (e.g., dash).  Use it
+	# here however when possible just to lower collision chance.
+	tmpdir=${TMPDIR-/tmp}/ins$RANDOM-$$
+
+	trap '
+	  ret=$?
+	  rmdir "$tmpdir/a/b" "$tmpdir/a" "$tmpdir" 2>/dev/null
+	  exit $ret
+	' 0
+
+	# Because "mkdir -p" follows existing symlinks and we likely work
+	# directly in world-writeable /tmp, make sure that the '$tmpdir'
+	# directory is successfully created first before we actually test
+	# 'mkdir -p'.
+	if (umask $mkdir_umask &&
+	    $mkdirprog $mkdir_mode "$tmpdir" &&
+	    exec $mkdirprog $mkdir_mode -p -- "$tmpdir/a/b") >/dev/null 2>&1
 	then
-		dstfile=`basename $dst`
+	  if test -z "$dir_arg" || {
+	       # Check for POSIX incompatibilities with -m.
+	       # HP-UX 11.23 and IRIX 6.5 mkdir -m -p sets group- or
+	       # other-writable bit of parent directory when it shouldn't.
+	       # FreeBSD 6.1 mkdir -m -p sets mode of existing directory.
+	       test_tmpdir="$tmpdir/a"
+	       ls_ld_tmpdir=`ls -ld "$test_tmpdir"`
+	       case $ls_ld_tmpdir in
+		 d????-?r-*) different_mode=700;;
+		 d????-?--*) different_mode=755;;
+		 *) false;;
+	       esac &&
+	       $mkdirprog -m$different_mode -p -- "$test_tmpdir" && {
+		 ls_ld_tmpdir_1=`ls -ld "$test_tmpdir"`
+		 test "$ls_ld_tmpdir" = "$ls_ld_tmpdir_1"
+	       }
+	     }
+	  then posix_mkdir=:
+	  fi
+	  rmdir "$tmpdir/a/b" "$tmpdir/a" "$tmpdir"
 	else
-		dstfile=`basename $dst $transformbasename | 
-			sed $transformarg`$transformbasename
+	  # Remove any dirs left behind by ancient mkdir implementations.
+	  rmdir ./$mkdir_mode ./-p ./-- "$tmpdir" 2>/dev/null
 	fi
+	trap '' 0;;
+    esac
 
-# don't allow the sed command to completely eliminate the filename
+    if
+      $posix_mkdir && (
+        umask $mkdir_umask &&
+        $doit_exec $mkdirprog $mkdir_mode -p -- "$dstdir"
+      )
+    then :
+    else
 
-	if [ x"$dstfile" = x ] 
-	then
-		dstfile=`basename $dst`
-	else
-		true
-	fi
+      # mkdir does not conform to POSIX,
+      # or it failed possibly due to a race condition.  Create the
+      # directory the slow way, step by step, checking for races as we go.
 
-# Make a temp file name in the proper directory.
+      case $dstdir in
+        /*) prefix='/';;
+        [-=\(\)!]*) prefix='./';;
+        *)  prefix='';;
+      esac
 
-	dsttmp=$dstdir/#inst.$$#
+      oIFS=$IFS
+      IFS=/
+      set -f
+      set fnord $dstdir
+      shift
+      set +f
+      IFS=$oIFS
 
-# Move or copy the file name to the temp name
+      prefixes=
 
-	$doit $instcmd $src $dsttmp &&
+      for d
+      do
+        test X"$d" = X && continue
 
-	trap "rm -f ${dsttmp}" 0 &&
+        prefix=$prefix$d
+        if test -d "$prefix"; then
+          prefixes=
+        else
+          if $posix_mkdir; then
+            (umask $mkdir_umask &&
+             $doit_exec $mkdirprog $mkdir_mode -p -- "$dstdir") && break
+            # Don't fail if two instances are running concurrently.
+            test -d "$prefix" || exit 1
+          else
+            case $prefix in
+              *\'*) qprefix=`echo "$prefix" | sed "s/'/'\\\\\\\\''/g"`;;
+              *) qprefix=$prefix;;
+            esac
+            prefixes="$prefixes '$qprefix'"
+          fi
+        fi
+        prefix=$prefix/
+      done
 
-# and set any options; do chmod last to preserve setuid bits
+      if test -n "$prefixes"; then
+        # Don't fail if two instances are running concurrently.
+        (umask $mkdir_umask &&
+         eval "\$doit_exec \$mkdirprog $prefixes") ||
+          test -d "$dstdir" || exit 1
+        obsolete_mkdir_used=true
+      fi
+    fi
+  fi
 
-# If any of these fail, we abort the whole thing.  If we want to
-# ignore errors from any of these, just make sure not to ignore
-# errors from the above "$doit $instcmd $src $dsttmp" command.
+  if test -n "$dir_arg"; then
+    { test -z "$chowncmd" || $doit $chowncmd "$dst"; } &&
+    { test -z "$chgrpcmd" || $doit $chgrpcmd "$dst"; } &&
+    { test "$obsolete_mkdir_used$chowncmd$chgrpcmd" = false ||
+      test -z "$chmodcmd" || $doit $chmodcmd $mode "$dst"; } || exit 1
+  else
 
-	if [ x"$chowncmd" != x ]; then $doit $chowncmd $dsttmp; else true;fi &&
-	if [ x"$chgrpcmd" != x ]; then $doit $chgrpcmd $dsttmp; else true;fi &&
-	if [ x"$stripcmd" != x ]; then $doit $stripcmd $dsttmp; else true;fi &&
-	if [ x"$chmodcmd" != x ]; then $doit $chmodcmd $dsttmp; else true;fi &&
+    # Make a couple of temp file names in the proper directory.
+    dsttmp=${dstdirslash}_inst.$$_
+    rmtmp=${dstdirslash}_rm.$$_
 
-# Now rename the file to the real destination.
+    # Trap to clean up those temp files at exit.
+    trap 'ret=$?; rm -f "$dsttmp" "$rmtmp" && exit $ret' 0
 
-	$doit $rmcmd -f $dstdir/$dstfile &&
-	$doit $mvcmd $dsttmp $dstdir/$dstfile 
+    # Copy the file name to the temp name.
+    (umask $cp_umask &&
+     { test -z "$stripcmd" || {
+	 # Create $dsttmp read-write so that cp doesn't create it read-only,
+	 # which would cause strip to fail.
+	 if test -z "$doit"; then
+	   : >"$dsttmp" # No need to fork-exec 'touch'.
+	 else
+	   $doit touch "$dsttmp"
+	 fi
+       }
+     } &&
+     $doit_exec $cpprog "$src" "$dsttmp") &&
 
-fi &&
+    # and set any options; do chmod last to preserve setuid bits.
+    #
+    # If any of these fail, we abort the whole thing.  If we want to
+    # ignore errors from any of these, just make sure not to ignore
+    # errors from the above "$doit $cpprog $src $dsttmp" command.
+    #
+    { test -z "$chowncmd" || $doit $chowncmd "$dsttmp"; } &&
+    { test -z "$chgrpcmd" || $doit $chgrpcmd "$dsttmp"; } &&
+    { test -z "$stripcmd" || $doit $stripcmd "$dsttmp"; } &&
+    { test -z "$chmodcmd" || $doit $chmodcmd $mode "$dsttmp"; } &&
 
+    # If -C, don't bother to copy if it wouldn't change the file.
+    if $copy_on_change &&
+       old=`LC_ALL=C ls -dlL "$dst"     2>/dev/null` &&
+       new=`LC_ALL=C ls -dlL "$dsttmp"  2>/dev/null` &&
+       set -f &&
+       set X $old && old=:$2:$4:$5:$6 &&
+       set X $new && new=:$2:$4:$5:$6 &&
+       set +f &&
+       test "$old" = "$new" &&
+       $cmpprog "$dst" "$dsttmp" >/dev/null 2>&1
+    then
+      rm -f "$dsttmp"
+    else
+      # If $backupsuffix is set, and the file being installed
+      # already exists, attempt a backup.  Don't worry if it fails,
+      # e.g., if mv doesn't support -f.
+      if test -n "$backupsuffix" && test -f "$dst"; then
+        $doit $mvcmd -f "$dst" "$dst$backupsuffix" 2>/dev/null
+      fi
 
-exit 0
+      # Rename the file to the real destination.
+      $doit $mvcmd -f "$dsttmp" "$dst" 2>/dev/null ||
+
+      # The rename failed, perhaps because mv can't rename something else
+      # to itself, or perhaps because mv is so ancient that it does not
+      # support -f.
+      {
+        # Now remove or move aside any old file at destination location.
+        # We try this two ways since rm can't unlink itself on some
+        # systems and the destination file might be busy for other
+        # reasons.  In this case, the final cleanup might fail but the new
+        # file should still install successfully.
+        {
+          test ! -f "$dst" ||
+          $doit $rmcmd "$dst" 2>/dev/null ||
+          { $doit $mvcmd -f "$dst" "$rmtmp" 2>/dev/null &&
+            { $doit $rmcmd "$rmtmp" 2>/dev/null; :; }
+          } ||
+          { echo "$0: cannot unlink or rename $dst" >&2
+            (exit 1); exit 1
+          }
+        } &&
+
+        # Now rename the file to the real destination.
+        $doit $mvcmd "$dsttmp" "$dst"
+      }
+    fi || exit 1
+
+    trap '' 0
+  fi
+done
+
+# Local variables:
+# eval: (add-hook 'before-save-hook 'time-stamp)
+# time-stamp-start: "scriptversion="
+# time-stamp-format: "%:y-%02m-%02d.%02H"
+# time-stamp-time-zone: "UTC0"
+# time-stamp-end: "; # UTC"
+# End:

samples/advanced/pdegen/amg_d_pde2d.F90

@@ -73,6 +73,9 @@ program amg_d_pde2d
   use amg_d_pde2d_exp_mod
   use amg_d_pde2d_box_mod
   use amg_d_genpde_mod
+#if defined(OPENMP)
+  use omp_lib
+#endif
   implicit none
 
   ! input parameters
@@ -93,7 +96,7 @@ program amg_d_pde2d
   type(psb_d_vect_type) :: x,b,r
   ! parallel environment
   type(psb_ctxt_type) :: ctxt
-  integer(psb_ipk_)   :: iam, np
+  integer(psb_ipk_)   :: iam, np, nth
 
   ! solver parameters
   integer(psb_ipk_)        :: iter, itmax,itrace, istopc, irst, nlv
@@ -197,6 +200,15 @@ program amg_d_pde2d
 
   call psb_init(ctxt)
   call psb_info(ctxt,iam,np)
+#if defined(OPENMP)
+  !$OMP parallel shared(nth)
+  !$OMP master
+  nth = omp_get_num_threads()
+  !$OMP end master
+  !$OMP end parallel
+#else
+  nth = 1
+#endif
 
   if (iam < 0) then
     ! This should not happen, but just in case
@@ -451,12 +463,14 @@ program amg_d_pde2d
   call psb_sum(ctxt,precsize)
   call prec%descr(info,iout=psb_out_unit)
   if (iam == psb_root_) then
-    write(psb_out_unit,'("Computed solution on ",i8," processors")')   np
+    write(psb_out_unit,'("Computed solution on ",i8," process(es)")')  np
+    write(psb_out_unit,'("Number of threads                  : ",i12)') nth
+    write(psb_out_unit,'("Total number of tasks              : ",i12)') nth*np
     write(psb_out_unit,'("Linear system size                 : ",i12)') system_size
-    write(psb_out_unit,'("PDE Coefficients                   : ",a)')  trim(pdecoeff)
-    write(psb_out_unit,'("Krylov method                      : ",a)')  trim(s_choice%kmethd)
-    write(psb_out_unit,'("Preconditioner                     : ",a)')  trim(p_choice%descr)
-    write(psb_out_unit,'("Iterations to convergence          : ",i12)')   iter
+    write(psb_out_unit,'("PDE Coefficients                   : ",a)') trim(pdecoeff)
+    write(psb_out_unit,'("Krylov method                      : ",a)') trim(s_choice%kmethd)
+    write(psb_out_unit,'("Preconditioner                     : ",a)') trim(p_choice%descr)
+    write(psb_out_unit,'("Iterations to convergence          : ",i12)')    iter
     write(psb_out_unit,'("Relative error estimate on exit    : ",es12.5)') err
     write(psb_out_unit,'("Number of levels in hierarchy      : ",i12)')    prec%get_nlevs()
     write(psb_out_unit,'("Time to build hierarchy            : ",es12.5)') thier

samples/advanced/pdegen/amg_d_pde3d.F90

@@ -74,6 +74,9 @@ program amg_d_pde3d
   use amg_d_pde3d_exp_mod
   use amg_d_pde3d_gauss_mod
   use amg_d_genpde_mod
+#if defined(OPENMP)
+  use omp_lib
+#endif
   implicit none
 
   ! input parameters
@@ -94,7 +97,7 @@ program amg_d_pde3d
   type(psb_d_vect_type) :: x,b,r
   ! parallel environment
   type(psb_ctxt_type) :: ctxt
-  integer(psb_ipk_)   :: iam, np
+  integer(psb_ipk_)   :: iam, np, nth
 
   ! solver parameters
   integer(psb_ipk_)        :: iter, itmax,itrace, istopc, irst, nlv
@@ -192,12 +195,21 @@ program amg_d_pde3d
   ! other variables
   integer(psb_ipk_)  :: info, i, k
   character(len=20)  :: name,ch_err
-
+  type(psb_d_csr_sparse_mat) :: amold
   info=psb_success_
 
 
   call psb_init(ctxt)
   call psb_info(ctxt,iam,np)
+#if defined(OPENMP)
+  !$OMP parallel shared(nth)
+  !$OMP master
+  nth = omp_get_num_threads()
+  !$OMP end master
+  !$OMP end parallel
+#else
+  nth = 1
+#endif
 
   if (iam < 0) then
     ! This should not happen, but just in case
@@ -390,7 +402,7 @@ program amg_d_pde3d
   end if
   call psb_barrier(ctxt)
   t1 = psb_wtime()
-  call prec%smoothers_build(a,desc_a,info)
+  call prec%smoothers_build(a,desc_a,info,amold=amold)
   tprec = psb_wtime()-t1
   if (info /= psb_success_) then
     call psb_errpush(psb_err_from_subroutine_,name,a_err='amg_smoothers_bld')
@@ -455,7 +467,9 @@ program amg_d_pde3d
   call psb_sum(ctxt,precsize)
   call prec%descr(info,iout=psb_out_unit)
   if (iam == psb_root_) then
-    write(psb_out_unit,'("Computed solution on ",i8," processors")')  np
+    write(psb_out_unit,'("Computed solution on ",i8," process(es)")')  np
+    write(psb_out_unit,'("Number of threads                  : ",i12)') nth
+    write(psb_out_unit,'("Total number of tasks              : ",i12)') nth*np
     write(psb_out_unit,'("Linear system size                 : ",i12)') system_size
     write(psb_out_unit,'("PDE Coefficients                   : ",a)') trim(pdecoeff)
     write(psb_out_unit,'("Krylov method                      : ",a)') trim(s_choice%kmethd)
@@ -478,7 +492,7 @@ program amg_d_pde3d
     write(psb_out_unit,'("Storage format for DESC_A          : ",a  )') desc_a%get_fmt()
 
   end if
-
+  call psb_print_timers(ctxt)
   !
   !  cleanup storage and exit
   !

samples/advanced/pdegen/amg_s_pde2d.F90

@@ -73,6 +73,9 @@ program amg_s_pde2d
   use amg_s_pde2d_exp_mod
   use amg_s_pde2d_box_mod
   use amg_s_genpde_mod
+#if defined(OPENMP)
+  use omp_lib
+#endif
   implicit none
 
   ! input parameters
@@ -93,7 +96,7 @@ program amg_s_pde2d
   type(psb_s_vect_type) :: x,b,r
   ! parallel environment
   type(psb_ctxt_type) :: ctxt
-  integer(psb_ipk_)   :: iam, np
+  integer(psb_ipk_)   :: iam, np, nth
 
   ! solver parameters
   integer(psb_ipk_)        :: iter, itmax,itrace, istopc, irst, nlv
@@ -197,6 +200,15 @@ program amg_s_pde2d
 
   call psb_init(ctxt)
   call psb_info(ctxt,iam,np)
+#if defined(OPENMP)
+  !$OMP parallel shared(nth)
+  !$OMP master
+  nth = omp_get_num_threads()
+  !$OMP end master
+  !$OMP end parallel
+#else
+  nth = 1
+#endif
 
   if (iam < 0) then
     ! This should not happen, but just in case
@@ -451,12 +463,14 @@ program amg_s_pde2d
   call psb_sum(ctxt,precsize)
   call prec%descr(info,iout=psb_out_unit)
   if (iam == psb_root_) then
-    write(psb_out_unit,'("Computed solution on ",i8," processors")')   np
+    write(psb_out_unit,'("Computed solution on ",i8," process(es)")')  np
+    write(psb_out_unit,'("Number of threads                  : ",i12)') nth
+    write(psb_out_unit,'("Total number of tasks              : ",i12)') nth*np
     write(psb_out_unit,'("Linear system size                 : ",i12)') system_size
-    write(psb_out_unit,'("PDE Coefficients                   : ",a)')  trim(pdecoeff)
-    write(psb_out_unit,'("Krylov method                      : ",a)')  trim(s_choice%kmethd)
-    write(psb_out_unit,'("Preconditioner                     : ",a)')  trim(p_choice%descr)
-    write(psb_out_unit,'("Iterations to convergence          : ",i12)')   iter
+    write(psb_out_unit,'("PDE Coefficients                   : ",a)') trim(pdecoeff)
+    write(psb_out_unit,'("Krylov method                      : ",a)') trim(s_choice%kmethd)
+    write(psb_out_unit,'("Preconditioner                     : ",a)') trim(p_choice%descr)
+    write(psb_out_unit,'("Iterations to convergence          : ",i12)')    iter
     write(psb_out_unit,'("Relative error estimate on exit    : ",es12.5)') err
     write(psb_out_unit,'("Number of levels in hierarchy      : ",i12)')    prec%get_nlevs()
     write(psb_out_unit,'("Time to build hierarchy            : ",es12.5)') thier

samples/advanced/pdegen/amg_s_pde3d.F90

@@ -74,6 +74,9 @@ program amg_s_pde3d
   use amg_s_pde3d_exp_mod
   use amg_s_pde3d_gauss_mod
   use amg_s_genpde_mod
+#if defined(OPENMP)
+  use omp_lib
+#endif
   implicit none
 
   ! input parameters
@@ -94,7 +97,7 @@ program amg_s_pde3d
   type(psb_s_vect_type) :: x,b,r
   ! parallel environment
   type(psb_ctxt_type) :: ctxt
-  integer(psb_ipk_)   :: iam, np
+  integer(psb_ipk_)   :: iam, np, nth
 
   ! solver parameters
   integer(psb_ipk_)        :: iter, itmax,itrace, istopc, irst, nlv
@@ -192,12 +195,21 @@ program amg_s_pde3d
   ! other variables
   integer(psb_ipk_)  :: info, i, k
   character(len=20)  :: name,ch_err
-
+  type(psb_s_csr_sparse_mat) :: amold
   info=psb_success_
 
 
   call psb_init(ctxt)
   call psb_info(ctxt,iam,np)
+#if defined(OPENMP)
+  !$OMP parallel shared(nth)
+  !$OMP master
+  nth = omp_get_num_threads()
+  !$OMP end master
+  !$OMP end parallel
+#else
+  nth = 1
+#endif
 
   if (iam < 0) then
     ! This should not happen, but just in case
@@ -390,7 +402,7 @@ program amg_s_pde3d
   end if
   call psb_barrier(ctxt)
   t1 = psb_wtime()
-  call prec%smoothers_build(a,desc_a,info)
+  call prec%smoothers_build(a,desc_a,info,amold=amold)
   tprec = psb_wtime()-t1
   if (info /= psb_success_) then
     call psb_errpush(psb_err_from_subroutine_,name,a_err='amg_smoothers_bld')
@@ -455,7 +467,9 @@ program amg_s_pde3d
   call psb_sum(ctxt,precsize)
   call prec%descr(info,iout=psb_out_unit)
   if (iam == psb_root_) then
-    write(psb_out_unit,'("Computed solution on ",i8," processors")')  np
+    write(psb_out_unit,'("Computed solution on ",i8," process(es)")')  np
+    write(psb_out_unit,'("Number of threads                  : ",i12)') nth
+    write(psb_out_unit,'("Total number of tasks              : ",i12)') nth*np
     write(psb_out_unit,'("Linear system size                 : ",i12)') system_size
     write(psb_out_unit,'("PDE Coefficients                   : ",a)') trim(pdecoeff)
     write(psb_out_unit,'("Krylov method                      : ",a)') trim(s_choice%kmethd)
@@ -478,7 +492,7 @@ program amg_s_pde3d
     write(psb_out_unit,'("Storage format for DESC_A          : ",a  )') desc_a%get_fmt()
 
   end if
-
+  call psb_print_timers(ctxt)
   !
   !  cleanup storage and exit
   !

samples/advanced/pdegen/runs/amg_pde3d.inp

@@ -1,6 +1,6 @@
 %%%%%%%%%%%  General  arguments % Lines starting with % are ignored.
 CSR                         ! Storage format CSR COO JAD
-0200                       ! IDIM; domain size. Linear system size is IDIM**3
+0200                        ! IDIM; domain size. Linear system size is IDIM**3
 CONST                       ! PDECOEFF: CONST, EXP, GAUSS Coefficients of the PDE
 BICGSTAB                    ! Iterative method: BiCGSTAB BiCGSTABL BiCG CG CGS FCG GCR RGMRES
 2                           ! ISTOPC
@@ -9,7 +9,7 @@ BICGSTAB                    ! Iterative method: BiCGSTAB BiCGSTABL BiCG CG CGS F
 30                          ! IRST (restart for RGMRES and BiCGSTABL)
 1.d-6                       ! EPS
 %%%%%%%%%%%  Main preconditioner choices %%%%%%%%%%%%%%%%
-ML-VCYCLE-BJAC-D-BJAC       ! Longer descriptive name for preconditioner (up to 20 chars)
+ML-VBM-VCYCLE-FBGS-D-BJAC       ! Longer descriptive name for preconditioner (up to 20 chars)
 ML                          ! Preconditioner type: NONE JACOBI GS FBGS BJAC AS ML
 %%%%%%%%%%%  First smoother (for all levels but coarsest) %%%%%%%%%%%%%%%%
 FBGS                        ! Smoother type JACOBI FBGS GS BWGS BJAC AS. For 1-level, repeats previous.
@@ -39,8 +39,8 @@ VCYCLE                      ! Type of multilevel CYCLE: VCYCLE WCYCLE KCYCLE MUL
 -3                          ! Max Number of levels in a multilevel preconditioner; if <0, lib default
 -3                          ! Target coarse matrix size per process; if <0, lib default
 SMOOTHED                    ! Type of aggregation: SMOOTHED UNSMOOTHED
-COUPLED                     ! Parallel aggregation: DEC, SYMDEC, COUPLED
-MATCHBOXP                   ! aggregation measure SOC1, MATCHBOXP
+DEC                         ! Parallel aggregation: DEC, SYMDEC, COUPLED
+SOC1                        ! aggregation measure SOC1, MATCHBOXP
 8                           ! Requested size of the aggregates for MATCHBOXP
 NATURAL                     ! Ordering of aggregation NATURAL DEGREE
 -1.5                        ! Coarsening ratio, if < 0 use library default