Updated internal docs.

mlprec/mld_c_base_aggregator_mod.f90

@@ -285,7 +285,7 @@ contains
   !!         The mapping is store in ILAGGR; for each local row index I,      
   !!         ILAGGR(I) contains the index of the aggregate to which index I
   !!         will contribute, in global numbering. 
-  !!         Many aggregation produce a binary tentative prolongator, but some
+  !!         Many aggregations produce a binary tentative prolongator, but some
   !!         do not, hence we also need the OP_PROL output.
   !!         AG_DATA is passed here just in case some of the
   !!         aggregators need it internally, most of them will ignore.

mlprec/mld_c_base_smoother_mod.f90

@@ -56,6 +56,9 @@
 !  contains a SOLVER object: the SOLVER operates locally within the
 !  current process, whereas the SMOOTHER object accounts for (possible)
 !  interactions between processes.
+!  Some solvers (MUMPS and SuperLU_DIST) can also operate on the entire
+!  distributed matrix, in which case the smoother object essentially
+!  becomes transparent. 
 !
 module mld_c_base_smoother_mod
 

mlprec/mld_c_diag_solver.f90

@@ -275,6 +275,18 @@ contains
 
 end module mld_c_diag_solver
 
+!
+! Module: mld_c_l1_diag_solver_mod
+!
+!  This module defines: 
+!  - the mld_c_l1_diag_solver_type data structure containing the 
+!    L1  diagonal solver. 
+!    The solver is defined as a diagonal containing in each element the
+!    inverse of the sum of the absolute values of the matrix entries
+!    along the corresponding row. 
+!    Combined with a Jacobi "smoother" generates 
+!    what are commonly known as the L1-Jacobi iterations
+!
 
 module mld_c_l1_diag_solver
 

mlprec/mld_c_gs_solver.f90

@@ -48,7 +48,8 @@
 !    on the block diagonal). Combined with a Jacobi smoother will generate a
 !    hybrid-Gauss-Seidel solver, i.e. Gauss-Seidel within each process, Jacobi
 !    among the processes.
-!
+!    With two objects as pre- and post-smoothers it is possible to build a
+!    Forward-Backward smoother, suitable for symmetric iterations.
 !
 module mld_c_gs_solver
 

mlprec/mld_c_ilu_fact_mod.f90

@@ -41,7 +41,7 @@
 !
 ! Module: mld_c_ilu_fact_mod
 !
-!  This module defines some interfaces used internally by the implementation if
+!  This module defines some interfaces used internally by the implementation of
 !  mld_c_ilu_solver, but not visible to the end user. 
 !
 !

mlprec/mld_c_ilu_solver.f90

@@ -51,7 +51,7 @@
 !       threshold base ILU(T,L)
 !    3. The diagonal is stored separately, so strictly speaking this is
 !       an incomplete LDU factorization; 
-!    4. The application phase is shared;
+!    4. The application phase is shared among all variants;
 !
 !
 module mld_c_ilu_solver

mlprec/mld_c_inner_mod.f90

@@ -39,8 +39,7 @@
 !
 ! Module: mld_inner_mod
 !
-!  This module defines the interfaces to the real/complex, single/double
-!  precision versions of inner MLD2P4 routines.
+!  This module defines the interfaces to inner MLD2P4 routines.
 !  The interfaces  of the user level routines are defined in mld_prec_mod.f90.
 !
 module mld_c_inner_mod

mlprec/mld_c_jac_smoother.f90

@@ -44,10 +44,10 @@
 !    the mld_c_jac_smoother_type data structure containing the
 !    smoother for a Jacobi/block Jacobi smoother.
 !  The smoother stores in ND the block off-diagonal matrix.
-!  One special case is treated separately, when the solver is DIAG
+!  One special case is treated separately, when the solver is DIAG or L1-DIAG
 !  then the ND is the entire off-diagonal part of the matrix (including the
 !  main diagonal block), so that it becomes possible to implement
-!  pure Jacobi global solver. 
+!  a pure Jacobi or L1-Jacobi global solver. 
 ! 
 module mld_c_jac_smoother
 

mlprec/mld_c_mumps_solver.F90

@@ -48,7 +48,7 @@
 !  - the mld_c_mumps_solver_type data structure containing the ingredients
 !    to interface with the MUMPS package. 
 !    1. The factorization can be either restricted to the diagonal block of the
-!       current image or distributed (and thus exact)
+!       current image or distributed (and thus exact). 
 !
 module mld_c_mumps_solver
   use mld_c_base_solver_mod

mlprec/mld_c_prec_mod.f90

@@ -39,7 +39,7 @@
 !
 ! Module: mld_c_prec_mod
 !
-!  This module defines the interfaces to the real/complex, single/double
+!  This module defines the user interfaces to the real/complex, single/double
 !  precision versions of the user-level MLD2P4 routines.
 !
 module mld_c_prec_mod

mlprec/mld_c_prec_type.f90

@@ -61,24 +61,24 @@ module mld_c_prec_type
   use psb_prec_mod, only : psb_cprec_type
 
   !
-  ! Type: mld_Tprec_type.
+  ! Type: mld_cprec_type.
   !
   !  This is the data type containing all the information about the multilevel
-  !  preconditioner (here and in the following 'T' denotes 'd', 's', 'c' and
-  !  'z', according to the real/complex, single/double precision version of
-  !  MLD2P4). It consists of an array of 'one-level' intermediate data structures
-  !  of type mld_Tonelev_type, each containing the information needed to apply
+  !  preconditioner ('d', 's', 'c' and 'z', according to the real/complex,
+  !  single/double precision version of  MLD2P4).
+  !  It consists of an array of 'one-level' intermediate data structures
+  !  of type mld_conelev_type, each containing the information needed to apply
   !  the smoothing and the coarse-space correction at a generic level. RT is the
   !  real data type, i.e. S for both S and C, and D for both D and Z. 
   !
-  !  type mld_Tprec_type
-  !    type(mld_Tonelev_type), allocatable :: precv(:) 
-  !  end type mld_Tprec_type
+  !  type mld_cprec_type
+  !    type(mld_conelev_type), allocatable :: precv(:) 
+  !  end type mld_cprec_type
   ! 
   !  Note that the levels are numbered in increasing order starting from
-  !  the finest one and the number of levels is given by size(precv(:)),
-  !  and that is the id of the coarsest level. 
-  !  In the multigrid literature authors often number the levels in decreasing
+  !  the level 1 as the finest one, and the number of levels is given by
+  !  size(precv(:))  which is the id of the coarsest level. 
+  !  In the multigrid literature many authors number the levels in the opposite
   !  order, with level 0 being the id of the coarsest level.
   !
   !
@@ -92,9 +92,9 @@ module mld_c_prec_type
     !
     integer(psb_ipk_)                  :: outer_sweeps = 1
     !
-    ! Coarse solver requires some tricky checks, and this needs we record the
-    ! choice in the format given by the user, to keep track against what
-    ! is put later in the multilevel array
+    ! Coarse solver requires some tricky checks, and for this we need to
+    ! record the choice in the format given by the user,
+    ! to keep track against what is put later in the multilevel array
     !
     integer(psb_ipk_)                  :: coarse_solver = -1
     

mlprec/mld_d_base_aggregator_mod.f90

@@ -285,7 +285,7 @@ contains
   !!         The mapping is store in ILAGGR; for each local row index I,      
   !!         ILAGGR(I) contains the index of the aggregate to which index I
   !!         will contribute, in global numbering. 
-  !!         Many aggregation produce a binary tentative prolongator, but some
+  !!         Many aggregations produce a binary tentative prolongator, but some
   !!         do not, hence we also need the OP_PROL output.
   !!         AG_DATA is passed here just in case some of the
   !!         aggregators need it internally, most of them will ignore.

mlprec/mld_d_base_smoother_mod.f90

@@ -56,6 +56,9 @@
 !  contains a SOLVER object: the SOLVER operates locally within the
 !  current process, whereas the SMOOTHER object accounts for (possible)
 !  interactions between processes.
+!  Some solvers (MUMPS and SuperLU_DIST) can also operate on the entire
+!  distributed matrix, in which case the smoother object essentially
+!  becomes transparent. 
 !
 module mld_d_base_smoother_mod
 

mlprec/mld_d_diag_solver.f90

@@ -275,6 +275,18 @@ contains
 
 end module mld_d_diag_solver
 
+!
+! Module: mld_d_l1_diag_solver_mod
+!
+!  This module defines: 
+!  - the mld_d_l1_diag_solver_type data structure containing the 
+!    L1  diagonal solver. 
+!    The solver is defined as a diagonal containing in each element the
+!    inverse of the sum of the absolute values of the matrix entries
+!    along the corresponding row. 
+!    Combined with a Jacobi "smoother" generates 
+!    what are commonly known as the L1-Jacobi iterations
+!
 
 module mld_d_l1_diag_solver
 

mlprec/mld_d_gs_solver.f90

@@ -48,7 +48,8 @@
 !    on the block diagonal). Combined with a Jacobi smoother will generate a
 !    hybrid-Gauss-Seidel solver, i.e. Gauss-Seidel within each process, Jacobi
 !    among the processes.
-!
+!    With two objects as pre- and post-smoothers it is possible to build a
+!    Forward-Backward smoother, suitable for symmetric iterations.
 !
 module mld_d_gs_solver
 

mlprec/mld_d_ilu_fact_mod.f90

@@ -41,7 +41,7 @@
 !
 ! Module: mld_d_ilu_fact_mod
 !
-!  This module defines some interfaces used internally by the implementation if
+!  This module defines some interfaces used internally by the implementation of
 !  mld_d_ilu_solver, but not visible to the end user. 
 !
 !

mlprec/mld_d_ilu_solver.f90

@@ -51,7 +51,7 @@
 !       threshold base ILU(T,L)
 !    3. The diagonal is stored separately, so strictly speaking this is
 !       an incomplete LDU factorization; 
-!    4. The application phase is shared;
+!    4. The application phase is shared among all variants;
 !
 !
 module mld_d_ilu_solver

mlprec/mld_d_inner_mod.f90

@@ -39,8 +39,7 @@
 !
 ! Module: mld_inner_mod
 !
-!  This module defines the interfaces to the real/complex, single/double
-!  precision versions of inner MLD2P4 routines.
+!  This module defines the interfaces to inner MLD2P4 routines.
 !  The interfaces  of the user level routines are defined in mld_prec_mod.f90.
 !
 module mld_d_inner_mod

mlprec/mld_d_jac_smoother.f90

@@ -44,10 +44,10 @@
 !    the mld_d_jac_smoother_type data structure containing the
 !    smoother for a Jacobi/block Jacobi smoother.
 !  The smoother stores in ND the block off-diagonal matrix.
-!  One special case is treated separately, when the solver is DIAG
+!  One special case is treated separately, when the solver is DIAG or L1-DIAG
 !  then the ND is the entire off-diagonal part of the matrix (including the
 !  main diagonal block), so that it becomes possible to implement
-!  pure Jacobi global solver. 
+!  a pure Jacobi or L1-Jacobi global solver. 
 ! 
 module mld_d_jac_smoother
 

mlprec/mld_d_mumps_solver.F90

@@ -48,7 +48,7 @@
 !  - the mld_d_mumps_solver_type data structure containing the ingredients
 !    to interface with the MUMPS package. 
 !    1. The factorization can be either restricted to the diagonal block of the
-!       current image or distributed (and thus exact)
+!       current image or distributed (and thus exact). 
 !
 module mld_d_mumps_solver
   use mld_d_base_solver_mod

mlprec/mld_d_prec_mod.f90

@@ -39,7 +39,7 @@
 !
 ! Module: mld_d_prec_mod
 !
-!  This module defines the interfaces to the real/complex, single/double
+!  This module defines the user interfaces to the real/complex, single/double
 !  precision versions of the user-level MLD2P4 routines.
 !
 module mld_d_prec_mod

mlprec/mld_d_prec_type.f90

@@ -61,24 +61,24 @@ module mld_d_prec_type
   use psb_prec_mod, only : psb_dprec_type
 
   !
-  ! Type: mld_Tprec_type.
+  ! Type: mld_dprec_type.
   !
   !  This is the data type containing all the information about the multilevel
-  !  preconditioner (here and in the following 'T' denotes 'd', 's', 'c' and
-  !  'z', according to the real/complex, single/double precision version of
-  !  MLD2P4). It consists of an array of 'one-level' intermediate data structures
-  !  of type mld_Tonelev_type, each containing the information needed to apply
+  !  preconditioner ('d', 's', 'c' and 'z', according to the real/complex,
+  !  single/double precision version of  MLD2P4).
+  !  It consists of an array of 'one-level' intermediate data structures
+  !  of type mld_donelev_type, each containing the information needed to apply
   !  the smoothing and the coarse-space correction at a generic level. RT is the
   !  real data type, i.e. S for both S and C, and D for both D and Z. 
   !
-  !  type mld_Tprec_type
-  !    type(mld_Tonelev_type), allocatable :: precv(:) 
-  !  end type mld_Tprec_type
+  !  type mld_dprec_type
+  !    type(mld_donelev_type), allocatable :: precv(:) 
+  !  end type mld_dprec_type
   ! 
   !  Note that the levels are numbered in increasing order starting from
-  !  the finest one and the number of levels is given by size(precv(:)),
-  !  and that is the id of the coarsest level. 
-  !  In the multigrid literature authors often number the levels in decreasing
+  !  the level 1 as the finest one, and the number of levels is given by
+  !  size(precv(:))  which is the id of the coarsest level. 
+  !  In the multigrid literature many authors number the levels in the opposite
   !  order, with level 0 being the id of the coarsest level.
   !
   !
@@ -92,9 +92,9 @@ module mld_d_prec_type
     !
     integer(psb_ipk_)                  :: outer_sweeps = 1
     !
-    ! Coarse solver requires some tricky checks, and this needs we record the
-    ! choice in the format given by the user, to keep track against what
-    ! is put later in the multilevel array
+    ! Coarse solver requires some tricky checks, and for this we need to
+    ! record the choice in the format given by the user,
+    ! to keep track against what is put later in the multilevel array
     !
     integer(psb_ipk_)                  :: coarse_solver = -1
     

mlprec/mld_s_base_aggregator_mod.f90

@@ -285,7 +285,7 @@ contains
   !!         The mapping is store in ILAGGR; for each local row index I,      
   !!         ILAGGR(I) contains the index of the aggregate to which index I
   !!         will contribute, in global numbering. 
-  !!         Many aggregation produce a binary tentative prolongator, but some
+  !!         Many aggregations produce a binary tentative prolongator, but some
   !!         do not, hence we also need the OP_PROL output.
   !!         AG_DATA is passed here just in case some of the
   !!         aggregators need it internally, most of them will ignore.

mlprec/mld_s_base_smoother_mod.f90

@@ -56,6 +56,9 @@
 !  contains a SOLVER object: the SOLVER operates locally within the
 !  current process, whereas the SMOOTHER object accounts for (possible)
 !  interactions between processes.
+!  Some solvers (MUMPS and SuperLU_DIST) can also operate on the entire
+!  distributed matrix, in which case the smoother object essentially
+!  becomes transparent. 
 !
 module mld_s_base_smoother_mod
 

mlprec/mld_s_diag_solver.f90

@@ -275,6 +275,18 @@ contains
 
 end module mld_s_diag_solver
 
+!
+! Module: mld_s_l1_diag_solver_mod
+!
+!  This module defines: 
+!  - the mld_s_l1_diag_solver_type data structure containing the 
+!    L1  diagonal solver. 
+!    The solver is defined as a diagonal containing in each element the
+!    inverse of the sum of the absolute values of the matrix entries
+!    along the corresponding row. 
+!    Combined with a Jacobi "smoother" generates 
+!    what are commonly known as the L1-Jacobi iterations
+!
 
 module mld_s_l1_diag_solver
 

mlprec/mld_s_gs_solver.f90

@@ -48,7 +48,8 @@
 !    on the block diagonal). Combined with a Jacobi smoother will generate a
 !    hybrid-Gauss-Seidel solver, i.e. Gauss-Seidel within each process, Jacobi
 !    among the processes.
-!
+!    With two objects as pre- and post-smoothers it is possible to build a
+!    Forward-Backward smoother, suitable for symmetric iterations.
 !
 module mld_s_gs_solver
 

mlprec/mld_s_ilu_fact_mod.f90

@@ -41,7 +41,7 @@
 !
 ! Module: mld_s_ilu_fact_mod
 !
-!  This module defines some interfaces used internally by the implementation if
+!  This module defines some interfaces used internally by the implementation of
 !  mld_s_ilu_solver, but not visible to the end user. 
 !
 !

mlprec/mld_s_ilu_solver.f90

@@ -51,7 +51,7 @@
 !       threshold base ILU(T,L)
 !    3. The diagonal is stored separately, so strictly speaking this is
 !       an incomplete LDU factorization; 
-!    4. The application phase is shared;
+!    4. The application phase is shared among all variants;
 !
 !
 module mld_s_ilu_solver

mlprec/mld_s_inner_mod.f90

@@ -39,8 +39,7 @@
 !
 ! Module: mld_inner_mod
 !
-!  This module defines the interfaces to the real/complex, single/double
-!  precision versions of inner MLD2P4 routines.
+!  This module defines the interfaces to inner MLD2P4 routines.
 !  The interfaces  of the user level routines are defined in mld_prec_mod.f90.
 !
 module mld_s_inner_mod

mlprec/mld_s_jac_smoother.f90

@@ -44,10 +44,10 @@
 !    the mld_s_jac_smoother_type data structure containing the
 !    smoother for a Jacobi/block Jacobi smoother.
 !  The smoother stores in ND the block off-diagonal matrix.
-!  One special case is treated separately, when the solver is DIAG
+!  One special case is treated separately, when the solver is DIAG or L1-DIAG
 !  then the ND is the entire off-diagonal part of the matrix (including the
 !  main diagonal block), so that it becomes possible to implement
-!  pure Jacobi global solver. 
+!  a pure Jacobi or L1-Jacobi global solver. 
 ! 
 module mld_s_jac_smoother
 

mlprec/mld_s_mumps_solver.F90

@@ -48,7 +48,7 @@
 !  - the mld_s_mumps_solver_type data structure containing the ingredients
 !    to interface with the MUMPS package. 
 !    1. The factorization can be either restricted to the diagonal block of the
-!       current image or distributed (and thus exact)
+!       current image or distributed (and thus exact). 
 !
 module mld_s_mumps_solver
   use mld_s_base_solver_mod

mlprec/mld_s_prec_mod.f90

@@ -39,7 +39,7 @@
 !
 ! Module: mld_s_prec_mod
 !
-!  This module defines the interfaces to the real/complex, single/double
+!  This module defines the user interfaces to the real/complex, single/double
 !  precision versions of the user-level MLD2P4 routines.
 !
 module mld_s_prec_mod

mlprec/mld_s_prec_type.f90

@@ -61,24 +61,24 @@ module mld_s_prec_type
   use psb_prec_mod, only : psb_sprec_type
 
   !
-  ! Type: mld_Tprec_type.
+  ! Type: mld_sprec_type.
   !
   !  This is the data type containing all the information about the multilevel
-  !  preconditioner (here and in the following 'T' denotes 'd', 's', 'c' and
-  !  'z', according to the real/complex, single/double precision version of
-  !  MLD2P4). It consists of an array of 'one-level' intermediate data structures
-  !  of type mld_Tonelev_type, each containing the information needed to apply
+  !  preconditioner ('d', 's', 'c' and 'z', according to the real/complex,
+  !  single/double precision version of  MLD2P4).
+  !  It consists of an array of 'one-level' intermediate data structures
+  !  of type mld_sonelev_type, each containing the information needed to apply
   !  the smoothing and the coarse-space correction at a generic level. RT is the
   !  real data type, i.e. S for both S and C, and D for both D and Z. 
   !
-  !  type mld_Tprec_type
-  !    type(mld_Tonelev_type), allocatable :: precv(:) 
-  !  end type mld_Tprec_type
+  !  type mld_sprec_type
+  !    type(mld_sonelev_type), allocatable :: precv(:) 
+  !  end type mld_sprec_type
   ! 
   !  Note that the levels are numbered in increasing order starting from
-  !  the finest one and the number of levels is given by size(precv(:)),
-  !  and that is the id of the coarsest level. 
-  !  In the multigrid literature authors often number the levels in decreasing
+  !  the level 1 as the finest one, and the number of levels is given by
+  !  size(precv(:))  which is the id of the coarsest level. 
+  !  In the multigrid literature many authors number the levels in the opposite
   !  order, with level 0 being the id of the coarsest level.
   !
   !
@@ -92,9 +92,9 @@ module mld_s_prec_type
     !
     integer(psb_ipk_)                  :: outer_sweeps = 1
     !
-    ! Coarse solver requires some tricky checks, and this needs we record the
-    ! choice in the format given by the user, to keep track against what
-    ! is put later in the multilevel array
+    ! Coarse solver requires some tricky checks, and for this we need to
+    ! record the choice in the format given by the user,
+    ! to keep track against what is put later in the multilevel array
     !
     integer(psb_ipk_)                  :: coarse_solver = -1
     

mlprec/mld_z_base_aggregator_mod.f90

@@ -285,7 +285,7 @@ contains
   !!         The mapping is store in ILAGGR; for each local row index I,      
   !!         ILAGGR(I) contains the index of the aggregate to which index I
   !!         will contribute, in global numbering. 
-  !!         Many aggregation produce a binary tentative prolongator, but some
+  !!         Many aggregations produce a binary tentative prolongator, but some
   !!         do not, hence we also need the OP_PROL output.
   !!         AG_DATA is passed here just in case some of the
   !!         aggregators need it internally, most of them will ignore.

mlprec/mld_z_base_smoother_mod.f90

@@ -56,6 +56,9 @@
 !  contains a SOLVER object: the SOLVER operates locally within the
 !  current process, whereas the SMOOTHER object accounts for (possible)
 !  interactions between processes.
+!  Some solvers (MUMPS and SuperLU_DIST) can also operate on the entire
+!  distributed matrix, in which case the smoother object essentially
+!  becomes transparent. 
 !
 module mld_z_base_smoother_mod
 

mlprec/mld_z_diag_solver.f90

@@ -275,6 +275,18 @@ contains
 
 end module mld_z_diag_solver
 
+!
+! Module: mld_z_l1_diag_solver_mod
+!
+!  This module defines: 
+!  - the mld_z_l1_diag_solver_type data structure containing the 
+!    L1  diagonal solver. 
+!    The solver is defined as a diagonal containing in each element the
+!    inverse of the sum of the absolute values of the matrix entries
+!    along the corresponding row. 
+!    Combined with a Jacobi "smoother" generates 
+!    what are commonly known as the L1-Jacobi iterations
+!
 
 module mld_z_l1_diag_solver
 

mlprec/mld_z_gs_solver.f90

@@ -48,7 +48,8 @@
 !    on the block diagonal). Combined with a Jacobi smoother will generate a
 !    hybrid-Gauss-Seidel solver, i.e. Gauss-Seidel within each process, Jacobi
 !    among the processes.
-!
+!    With two objects as pre- and post-smoothers it is possible to build a
+!    Forward-Backward smoother, suitable for symmetric iterations.
 !
 module mld_z_gs_solver
 

mlprec/mld_z_ilu_fact_mod.f90

@@ -41,7 +41,7 @@
 !
 ! Module: mld_z_ilu_fact_mod
 !
-!  This module defines some interfaces used internally by the implementation if
+!  This module defines some interfaces used internally by the implementation of
 !  mld_z_ilu_solver, but not visible to the end user. 
 !
 !

mlprec/mld_z_ilu_solver.f90

@@ -51,7 +51,7 @@
 !       threshold base ILU(T,L)
 !    3. The diagonal is stored separately, so strictly speaking this is
 !       an incomplete LDU factorization; 
-!    4. The application phase is shared;
+!    4. The application phase is shared among all variants;
 !
 !
 module mld_z_ilu_solver

mlprec/mld_z_inner_mod.f90

@@ -39,8 +39,7 @@
 !
 ! Module: mld_inner_mod
 !
-!  This module defines the interfaces to the real/complex, single/double
-!  precision versions of inner MLD2P4 routines.
+!  This module defines the interfaces to inner MLD2P4 routines.
 !  The interfaces  of the user level routines are defined in mld_prec_mod.f90.
 !
 module mld_z_inner_mod

mlprec/mld_z_jac_smoother.f90

@@ -44,10 +44,10 @@
 !    the mld_z_jac_smoother_type data structure containing the
 !    smoother for a Jacobi/block Jacobi smoother.
 !  The smoother stores in ND the block off-diagonal matrix.
-!  One special case is treated separately, when the solver is DIAG
+!  One special case is treated separately, when the solver is DIAG or L1-DIAG
 !  then the ND is the entire off-diagonal part of the matrix (including the
 !  main diagonal block), so that it becomes possible to implement
-!  pure Jacobi global solver. 
+!  a pure Jacobi or L1-Jacobi global solver. 
 ! 
 module mld_z_jac_smoother
 

mlprec/mld_z_mumps_solver.F90

@@ -48,7 +48,7 @@
 !  - the mld_z_mumps_solver_type data structure containing the ingredients
 !    to interface with the MUMPS package. 
 !    1. The factorization can be either restricted to the diagonal block of the
-!       current image or distributed (and thus exact)
+!       current image or distributed (and thus exact). 
 !
 module mld_z_mumps_solver
   use mld_z_base_solver_mod

mlprec/mld_z_prec_mod.f90

@@ -39,7 +39,7 @@
 !
 ! Module: mld_z_prec_mod
 !
-!  This module defines the interfaces to the real/complex, single/double
+!  This module defines the user interfaces to the real/complex, single/double
 !  precision versions of the user-level MLD2P4 routines.
 !
 module mld_z_prec_mod

mlprec/mld_z_prec_type.f90

@@ -61,24 +61,24 @@ module mld_z_prec_type
   use psb_prec_mod, only : psb_zprec_type
 
   !
-  ! Type: mld_Tprec_type.
+  ! Type: mld_zprec_type.
   !
   !  This is the data type containing all the information about the multilevel
-  !  preconditioner (here and in the following 'T' denotes 'd', 's', 'c' and
-  !  'z', according to the real/complex, single/double precision version of
-  !  MLD2P4). It consists of an array of 'one-level' intermediate data structures
-  !  of type mld_Tonelev_type, each containing the information needed to apply
+  !  preconditioner ('d', 's', 'c' and 'z', according to the real/complex,
+  !  single/double precision version of  MLD2P4).
+  !  It consists of an array of 'one-level' intermediate data structures
+  !  of type mld_zonelev_type, each containing the information needed to apply
   !  the smoothing and the coarse-space correction at a generic level. RT is the
   !  real data type, i.e. S for both S and C, and D for both D and Z. 
   !
-  !  type mld_Tprec_type
-  !    type(mld_Tonelev_type), allocatable :: precv(:) 
-  !  end type mld_Tprec_type
+  !  type mld_zprec_type
+  !    type(mld_zonelev_type), allocatable :: precv(:) 
+  !  end type mld_zprec_type
   ! 
   !  Note that the levels are numbered in increasing order starting from
-  !  the finest one and the number of levels is given by size(precv(:)),
-  !  and that is the id of the coarsest level. 
-  !  In the multigrid literature authors often number the levels in decreasing
+  !  the level 1 as the finest one, and the number of levels is given by
+  !  size(precv(:))  which is the id of the coarsest level. 
+  !  In the multigrid literature many authors number the levels in the opposite
   !  order, with level 0 being the id of the coarsest level.
   !
   !
@@ -92,9 +92,9 @@ module mld_z_prec_type
     !
     integer(psb_ipk_)                  :: outer_sweeps = 1
     !
-    ! Coarse solver requires some tricky checks, and this needs we record the
-    ! choice in the format given by the user, to keep track against what
-    ! is put later in the multilevel array
+    ! Coarse solver requires some tricky checks, and for this we need to
+    ! record the choice in the format given by the user,
+    ! to keep track against what is put later in the multilevel array
     !
     integer(psb_ipk_)                  :: coarse_solver = -1