SSIDS: Sparse Symmetric Indefinite Direct Solver

Major version history

2014-03-17 Version 1.0.0

Initial release

5.1 Installation

Please see the SPRAL install documentation. In particular note that:

• A CUDA compiler (nvcc) is required.
• A METIS library is required.
• A BLAS library is required (in addition to CUBLAS).

5.2 Usage overview

5.2.1 Calling sequences

Access to the package requires inclusion of either spral.h (for the entire SPRALlibrary) or spral_ssids.h (for just the relevant routines). i.e.

The following functions are available to the user:

• spral_ssids_default_options() initializes the options structure to default values.
• spral_ssids_analyse() accepts the matrix data in compressed sparse column format and optionally checks it for duplicates and out-of-range entries. The user may supply an elimination order; otherwise one is generated. Using this elimination order, spral_ssids_analyse() analyses the sparsity pattern of the matrix and prepares the data structures for the factorization.
• spral_ssids_analyse_coord() is an alternative to spral_ssids_analyse() that may be used if the user has the matrix data in coordinate format. Again, the user may supply an elimination order; otherwise one is generated. spral_ssids_analyse_coord() checks the matrix data for duplicates and out-of-range entries, stores it in compressed sparse column format and then proceeds in the same way as spral_ssids_analyse().
• spral_ssids_factor() uses the data structures set up by spral_ssids_analyse() to compute a sparse factorization. More than one call to spral_ssids_factor() may follow a call to spral_ssids_analyse() (allowing more than one matrix with the same sparsity pattern but different numerical values to be factorized without multiple calls to spral_ssids_analyse()). An option exists to scale the matrix.
• spral_ssids_solve1() and spral_ssids_solve() use the computed factors generated by spral_ssids_factor() to solve systems $AX=B$ for one or more right-hand sides $B$. Multiple calls to spral_ssids_solve1() and spral_ssids_solve() may follow a call to spral_ssids_factor(). An option is available to perform a partial solution.
• spral_ssids_free() should be called after all other calls are complete for a problem (including after an error return that does not allow the computation to continue). It frees memory and resources used by internal data structures.

In addition, the following routines may be called:

• spral_ssids_free_akeep() and spral_ssids_free_fkeep() may be called to free the memory and resources allocated by spral_ssids_analyse() or spral_ssids_factorize() to be freed separately.
• spral_ssids_enquire_posdef() may be called in the positive-definite case to obtain the pivots used.
• spral_ssids_enquire_indef() may be called in the indefinite case to obtain the pivot sequence used by the factorization and the entries of $D^{-1}$.
• spral_ssids_alter() may be called in the indefinite case to alter the entries of $D^{-1}$. Note that this means that $PLD{\left(PL\right)}^T$ is no longer a factorization of $A$.

5.2.2 Derived types

For each problem, the user must employ the derived types defined by the package to declare scalars of the types struct spral_ssids_options and struct spral_ssids_inform. The user must also declare two void * pointers akeep and fkeep for the package’s private data structures. The options data structure must be initialized using spral_ssids_default_options(), and the pointers akeep and fkeep must be initialized to NULL. The following pseudo-code illustrates this.

The components of spral_ssids_options and spral_ssids_inform are explained in Sections 5.5.1 and 5.5.2. The void * pointers are allocated using Fortran, and must be passed to spral_ssids_free_akeep(), spral_ssids_free_fkeep() and/or spral_ssids_free() to free the associated memory.

5.2.3 Achieving bit-compatibility

Care has been taken to ensure bit-compatibility is achieved using this solver. That is, consecutive runs with the same data on the same machine produces exactly the same solution.

5.2.4 Data formats

---

---

Compressed Sparse Column (CSC) Format

This standard data format consists of the following data:

Non-zero matrix entries are ordered by increasing column index and stored in the arrays row[] and val[] such that row[k] holds the row number and val[k] holds the value of the k-th entry. The ptr[] array stores column pointers such that ptr[i] is the position in row[] and val[] of the first entry in the i-th column, and ptr[n] is the total number of entries. Entries that are zero, including those on the diagonal, need not be specified.

If this format is used, SSIDS requires only the lower triangular entries of $A$, and there should be no duplicate entries. If the check argument to ssids_analyse() is true, out-of-range entries (including those in the upper triangle) will be discarded and any duplicates will be summed.

To illustrate the CSC format, the following arrays describe the matrix shown in Figure 5.1.

Coordinate Format

This standard data format consists of the following data:

The arrays should be set such that the k-th entry is in row row[k] and column col[k] with value val[k]. Entries that are zero, including those on the diagonal, need not be specified.

If this format is used, SSIDS requires that each entry of $A$ should be present only in the lower or upper triangular part. Entries present in both will be summed, as will any duplicate entries. Out-of-range entries are ignored.

To illustrate the coordinate format, the following arrays describe the matrix shown in Figure 5.1.

5.3 Basic Subroutines

5.3.1 spral_ssids_default_options()

To initialize a variable of type struct spral_ssids_options, the following routine is provided.

void spral_ssids_default_options(struct spral_ssids_options *options);

*options

is the instance to be initialized.

5.3.2 spral_ssids_analyse() and spral_ssids_analyse_coord()

To analyse the sparsity pattern and prepare for the factorization, the following routines are provided.

• for Compressed Sparse Column (CSC) format:
  void spral_ssids_analyse(bool check, int n, int *order, const int *ptr,
    const int *row, const double *val, void **akeep,
    const struct spral_ssids_options *options, struct spral_ssids_inform *inform);
• for Coordinate format:
  void spral_ssids_analyse_coord(int n, int *order, int ne, const int *row,
    const int *col, const double *val, void **akeep,
    const struct spral_ssids_options *options, struct spral_ssids_inform *inform);

Matrix data should be supplied as described in Section 5.2.4. As the package uses CSC format internally, spral_ssids_analyse() with checking disabled provides the most efficient interface.

check

determines if the matrix is checked for errors (true) or not (false). If set to true the matrix data is checked for errors and the cleaned matrix (duplicates are summed and out-of-range entries discarded) is stored in akeep. Otherwise, for data in CSC format, no checking of the matrix data is carried out and ptr[] and row[] must be passed unchanged to the factorization routines. Checking is always performed when the coordinate format is used.

n, ptr[n+1], row[ptr[n]]

specify the lower triangular part of $A$ in CSC format (see Section 5.2.4).

n, ne, row[ne], col[ne]

specify the lower (or upper) triangular part of $A$ in coordinate format (see Section 5.2.4).

order[n]

may be NULL. If options.ordering$=$0, order(:) must non-NULL and order[i] must hold the position of variable $i$ in the elimination order. On exit, order[] contains the elimination order that spral_ssids_factor() will be given (it is passed to these routines as part of *akeep); this order may give slightly more fill-in than any user-supplied order and, in the indefinite case, may be modified by spral_ssids_factor() to maintain numerical stability.

val[ne]

may be NULL, if non-NULL it must hold the numerical values of the entries of the matrix, as described in Section 5.2.4. val[] must not be NULL if a matching-based elimination ordering is required (options.ordering$=$2), and is otherwise ignored.

*akeep

is allocated to hold data about the problem being solved and must be passed unchanged to the other subroutines.

*options

specifies the algorithmic options used by the routine, as explained in Section 5.5.1.

*inform

is used to return information about the execution of the routine, as explained in Section 5.5.2.

5.3.3 spral_ssids_factor()

To factorize the matrix, the following routine is provided.
void spral_ssids_factor(bool posdef, const int *ptr, const int *row, const double *val,
  double *scale, void *akeep, void **fkeep, const struct spral_ssids_options *options,
  struct spral_ssids_inform *inform);

posdef

must be true if the matrix is positive-definite, and false if it is indefinite.

ptr[n+1]

and row[ptr[n]] may both be NULL unless spral_ssids_analyse() was called with check set to false. In this case, they must both be non-NULL and must be unchanged since that call.

val[ptr[n]]

must hold the numerical values of the entries of the matrix, as described in Section 5.2.4.

scale[n]

may be NULL. If non-NULL and options.scaling$=$0, it must contain the diagonal entries of the scaling matrix $S$. On exit, scale[i] will contain the i-th diagonal entry of the scaling matrix $S$.

akeep

must be unchanged since the call to spral_ssids_analyse() or spral_ssids_analyse_coord().

*fkeep

is allocated to hold data about the problem being solved and must be passed unchanged to the other subroutines.

*options

specifies the algorithmic options used by the subroutine, as explained in Section 5.5.1.

*inform

is used to return information about the execution of the subroutine, as explained in Section 5.5.2.

5.3.4 spral_ssids_solve()

To solve the linear system $AX=B$, after a call to spral_ssids_factor(), the following routines are provided.

• for a single right-hand side:
  void spral_ssids_solve1(int job, double *x1, void *akeep, void *fkeep,
    const struct spral_ssids_options *options, struct spral_ssids_inform *inform);
• for one or more right-hand sides:
  void spral_ssids_solve(int job, int nrhs, double *x, int ldx, void *akeep,
    void *fkeep, const struct spral_ssids_options *options,
    struct spral_ssids_inform *inform);

job

is used to specify the solve to be performed. In the positive-definite case, the Cholesky factorization that has been computed may be expressed in the form

$$SAS=\left(PL\right){\left(PL\right)}^T$$

where $P$ is a permutation matrix and $L$ is lower triangular. In the indefinite case, the factorization that has been computed may be expressed in the form

$$SAS=\left(PL\right)D{\left(PL\right)}^T$$

where $P$ is a permutation matrix, $L$ is unit lower triangular, and $D$ is block diagonal with blocks of order 1 and 2. $S$ is a diagonal scaling matrix ($S$ is equal to the identity, if options.scaling=0 and scale is NULL on the last call to spral_ssids_factor()). A partial solution may be computed by setting job to have one of the following values:

0

for solving $AX=B$

1

for solving $PLX=SB$

2

for solving $DX=B$ (indefinite case only)

3

for solving ${\left(PL\right)}^T{S}^{-1}X=B$

4

for solving $D{\left(PL\right)}^T{S}^{-1}X=B$ (indefinite case only)

x1[n]

must be set such that x1[i] holds the component of the right-hand side for variable $i$. On exit, x1[i] holds the solution for variable $i$.

nrhs

holds the number of right-hand sides.

x[nrhs][ldx]

must be set so that x[j][i] holds the component of the right-hand side for variable $i$ to the $j$-th system. On exit, x[j][i] holds the solution for variable $i$ to the $j$-th system.

ldx

must be set to the leading dimension of array x[].

akeep

must be unchanged since the last call to spral_ssids_factor().

fkeep

must be unchanged since the last call to spral_ssids_factor().

*options

specifies the algorithmic options used by the subroutine, as explained in Section 5.5.1.

inform

is used to return information about the execution of the subroutine, as explained in Section 5.5.2.

5.3.5 spral_ssids_free(), spral_ssids_free_akeep(), and spral_ssids_free_fkeep()

Once all other calls are complete for a problem or after an error return that does not allow the computation to continue, a call should be made to free memory and CUDA resources allocated by SSIDS and associated with the derived data types akeep and/or fkeep using calls to spral_ssids_free().

To free memory and resources, the following routines are provided.

• allocated by spral_ssids_analyse() or spral_ssids_analyse_coord():
  int spral_ssids_free_akeep(void **akeep);
• allocated by spral_ssids_factor():
  int spral_ssids_free_fkeep(void **fkeep);
• both at once:
  int spral_ssids_free(void **akeep, void **fkeep);

A non-zero return value gives a CUDA error code. This may indicate either a failure to deallocate GPU memory, or a pre-existing CUDA error condition. Note that due to the asynchronous nature of GPU execution, the reported error may have a cause errors external to SSIDS.

*akeep

must be passed unchanged by the user. On exit, associated memory and CUDA resources will have been released, and *akeep set to NULL.

*fkeep

that must be passed unchanged by the user. On exit, associated memory and CUDA resources will have been released, and *fkeep set to NULL.

5.4 Advanced subroutines

5.4.1 spral_ssids_enquire_posdef()

To obtain the matrix $D$ following a positive-definite factorization, the following routine is provided.
void spral_ssids_enquire_posdef(const void *akeep, const void *fkeep,
  const struct spral_ssids_options *options, struct spral_ssids_inform *inform,
  double *d);

akeep

must be unchanged since the call to spral_ssids_factor().

fkeep

must be unchanged since the call to spral_ssids_factor().

*options

specifies the algorithmic options used by the subroutine, as explained in Section 5.5.1.

*inform

is used to return information about the execution of the subroutine, as explained in Section 5.5.2.

d[n]

holds, on exit, the diagonal entries of the matrix $D$.

5.4.2 spral_ssids_enquire_indef()

To obtain the matrix $D^{-1}$ and/or the pivot order following an indefinite factorization, the following routine is provided.
void spral_ssids_enquire_indef(const void *akeep, const void *fkeep,
  const struct spral_ssids_options *options, struct spral_ssids_inform *inform,
  int *piv_order, double *d);

akeep

must be unchanged since the call to spral_ssids_factor().

fkeep

must be unchanged since the call to spral_ssids_factor().

*options

specifies the algorithmic options used by the subroutine, as explained in Section 5.5.1.

*inform

is used to return information about the execution of the subroutine, as explained in Section 5.5.2.

piv_order[n]

may be NULL. If non-NULL then, on exit, piv_order[i] holds the position of variable $i$ in the pivot order.

d[n][2]

may be NULL. If non-NULL, on exit diagonal entries of $D^{-1}$ will be placed in d[i][0], $i=1,2,\dots ,n$, the off-diagonal entries of $D^{-1}$ will be placed in d[i][1], $i=1,2,\dots ,n-1$, and d[n][1] will be set to zero.

5.4.3 spral_ssids_alter()

To alter $D^{-1}$ following an indefinite factorization, the following routine is provided.
void spral_ssids_alter(const double *d, const void *akeep, void *fkeep,
  const struct spral_ssids_options *options, struct spral_ssids_inform *inform);

Note that this routine is not compatabile with the option options.presolve$=$1.

d[n][2]

must hold the new entries of the matrix $D$. The diagonal entries of $D^{-1}$ will be altered to d[i][0], $i=1,2,\dots ,n$, and the off-diagonal entries will be altered to d[i][1], $i=1,2,\dots ,n-1$ (and $PLD{\left(PL\right)}^T$ will no longer be a factorization of $A$).

akeep

must be unchanged since the call to spral_ssids_factor().

fkeep

must be unchanged since the call to spral_ssids_factor().

*options

specifies the algorithmic options used by the subroutine, as explained in Section 5.5.1.

*inform

is used to return information about the execution of the subroutine, as explained in Section 5.5.2.

5.5 Derived types

5.5.1 struct spral_ssids_options

The structure spral_ssids_options is used to specify the options used within SSIDS. The components, that must be given default values through a call to spral_ssids_default_options(), are:

C specific options

int array_base

specifies the array indexing base. It must have the value either 0 (C indexing) or 1 (Fortran indexing). If array_base$=$1, the entries of arrays ptr[], row[], col[], order[], and piv_order[] start at 1, not 0. Further, entries of piv_order[] may also have negative sign to indicate they are part of a $2\times 2$ pivot. The default value is array_base$=$0.

Printing options

int print_level

is used to control the level of printing. The different levels are:

$<$ 0 No printing.

= 0 Error and warning messages only.

= 1 As 0, plus basic diagnostic printing.

$>$ 1 As 1, plus some additional diagnostic printing.

The default is print_level$=$0.

int unit_diagnostics

holds Fortran the unit number for diagnostic printing. Printing is suppressed if unit_diagnostics$<0$. The default is unit_diagnostics$=$6.

int unit_error

holds the Fortran unit number for error messages. Printing of error messages is suppressed if unit_error$<$0. The default is unit_error$=$6.

int unit_warning

holds the Fortran unit number for warning messages. Printing of warning messages is suppressed if unit_warning$<$0. The default is unit_warning$=$6.

Options used by spral_ssids_analyse() and spral_ssids_analyse_coord()

int ordering

controls the ordering algorithm used. If set to 0, the user must supply an elimination order in order[]; otherwise ssids_analyse() or ssids_analyse_coord() will an compute an elimination order. The options are:

0 User-supplied ordering is used.

1 METIS ordering with default settings is used.

2 A matching-based elimination ordering is computed (the Hungarian algorithm is used to identify large off-diagonal entries. A restricted METIS ordering is then used that forces these on to the subdiagonal). This option should only be chosen for indefinite systems. A scaling is also computed that may be used in spral_ssids_factor() (see options.scaling below).

The default is ordering$=$1. Restriction: ordering$=$0, 1, 2.

int nemin

controls node amalgamation. Two neighbours in the elimination tree are merged if they both involve fewer than nemin eliminations. The default is nemin$=$8. The default is used if nemin$<$1.

Options used by spral_ssids_factor()

int scaling

controls the use of scaling. The available options are:

$\le$ 0

No scaling (if scale is NULL), or user-supplied scaling (if scale is non-NULL).

$=$ 1

Compute a scaling using a weighted bipartite matching via the Hungarian Algorithm (MC64 algorithm).

$=$ 2

Compute a scaling using a weighted bipartite matching via the Auction Algorithm (may be lower quality than that computed using the Hungarian Algorithm, but can be considerably faster).

$=$ 3

A matching-based ordering has been generated during the analyse phase using options.ordering $=$ 2. Use the scaling generated as a side-effect of this process. The scaling will be the same as that generated with options.scaling $=$ 1 if the matrix values have not changed. This option will generate an error if a matching-based ordering was not used.

$\ge$ 4

Compute a scaling using the norm-equilibration algorithm of Ruiz.

The default is scaling$=$0.

Options used by spral_ssids_factor() with posdef$=$false ($A$ indefinite)

bool action

controls the behaviour for singular matrices. If the matrix is found to be singular (has rank less than the number of non-empty rows), the computation continues after issuing a warning if action has the value true or terminates with an error if it has the value false. The default is action$=$true.

double u

holds the relative pivot tolerance $u$. The default is u$=$0.01. Values outside the range $\left[0,0.5\right]$ are treated as the closest value in that range.

Options used by spral_ssids_factor() and spral_ssids_solve()

bool use_gpu_solve

controls whether to use the CPU or GPU for the spral_ssids_solve(). If true, the GPU is used, but some more advanced features are not available (see the description of the job parameter in Section 5.3.4). Setting use_gpu_solve$=$false is only compatible with options.presolve$=$0. The default value is use_gpu_solve$=$true.

int presolve

controls the amount of extra work performed during the factorization to accelerate the solve. It can take the following values:

0

Minimal work is performed during spral_ssids_factor() to prepare for the solve.

1

The explicit inverse of the nelim$\times$nelim diagonal block in each supernode is precalculated during spral_ssids_factor() (where nelim is the number of variables eliminated at that supernode). As the matrix $L$ is overwritte, the routine spral_ssids_alter() cannot be used. This option is not compatible with options.use_gpu_solve$=$false.

The default option is presolve$=$0.

5.5.2 struct spral_ssids_inform

The structure spral_ssids_inform is used to hold parameters that give information about the progress and needs of the algorithm. The components of struct spral_ssids_inform (in alphabetical order) are:

ing flag

gives the exit status of the algorithm (details in Section 5.6).

int matrix_dup

holds, on exit from spral_ssids_analyse() with check set to true or on exit from spral_ssids_analyse_coord(), the number of duplicate entries that were found and summed.

int matrix_missing_diag

holds, on exit from spral_ssids_analyse() with check set to true, or exit from ssids_analyse_coord(), the number of diagonal entries without an explicitly provided value.

int matrix_outrange

holds, on exit from spral_ssids_analyse() with check set to true or exit from spral_ssids_analyse_coord(), the number of out-of-range entries that were found and discarded.

int matrix_rank

holds, on exit from spral_ssids_analyse() or on exit from spral_ssids_analyse_coord(), the structural rank of $A$, if available (otherwise, it is set to n). On exit from spral_ssids_factor(), it holds the computed rank of the factorized matrix.

int maxdepth

holds, on exit from spral_ssids_analyse() or spral_ssids_analyse_coord(), the maximum depth of the assembly tree.

int maxfront

holds, on exit from spral_ssids_analyse() or spral_ssids_analyse_coord(), the maximum front size in the positive-definite case (or in the indefinite case with the same pivot sequence). On exit from spral_ssids_factor(), it holds the maximum front size.

int num_delay

holds, on exit from spral_ssids_factor(), the number of eliminations that were delayed, that is, the total number of fully-summed variables that were passed to the father node because of stability considerations. If a variable is passed further up the tree, it will be counted again.

long num_factor

holds, on exit from spral_ssids_analyse() or spral_ssids_analyse_coord(), the number of entries that will be in the factor $L$ in the positive-definite case (or in the indefinite case with the same pivot sequence). On exit from ssids_factor(), it holds the actual number of entries in the factor $L$. In the indefinite case, 2n entries of $D^{-1}$ are also held.

long num_flops

holds, on exit from spral_ssids_analyse() or spral_ssids_analyse_coord(), the number of floating-point operations that will be needed to perform the factorization in the positive-definite case (or in the indefinite case with the same pivot sequence). On exit from spral_ssids_factor(), it holds the number of floating-point operations performed.

int num_neg

holds, on exit from spral_ssids_factor(), the number of negative eigenvalues of the matrix $D$.

int num_sup

holds, on exit from spral_ssids_analyse() or spral_ssids_analyse_coord(), the number of supernodes in the problem.

int num_two

holds, on exit from spral_ssids_factor(), the number of $2\times 2$ pivots used by the factorization, that is, the number of $2\times 2$ blocks in $D$.

int stat

holds, in the event of an allocation or deallocation error, the Fortran stat parameter if it is available (and is set to 0 otherwise).

int cublas_error

holds, in the event of an error return from the CUBLAS library, the error code returned (and is 0 otherwise).

int cuda_error

holds, in the event of a CUDA error, the error code returned (and is 0 otherwise). Note that due to the asynchronous nature of GPU execution, the reported error may have a cause errors external to SSIDS.

5.6 Error flags

A successful return from a routine in the package is indicated by inform.flag having the value zero. A negative value is associated with an error message that by default will be output on unit options.unit_error.

Possible negative values are:

$-$1 An error has been made in the sequence of calls (this includes calling a subroutine after an error that cannot be recovered from).

$-$2 Returned by spral_ssids_analyse() and spral_ssids_analyse_coord() if n$<$0. Also returned by spral_ssids_analyse_coord() if ne$<$1.

$-$3 Returned by spral_ssids_analyse() if there is an error in ptr[].

$-$4 Returned by spral_ssids_analyse() if all the variable indices in one or more columns are out-of-range. Also returned by spral_ssids_analyse_coord() if all entries are out-of-range.

$-$5 Returned by spral_ssids_factor() if posdef$=$false and options.action = false when the matrix is found to be singular. The user may reset the matrix values in val[] and recall spral_ssids_factor().

$-$6 Returned by spral_ssids_factor() if posdef$=$true and the matrix is found to be not positive definite. This may be because the Hungarian scaling determined that the matrix was structurally singular. The user may reset the matrix values in val[] and recall spral_ssids_factor().

$-$7 Returned by spral_ssids_factor() if spral_ssids_analyse() was called with check set to false but ptr[] and/or row[] was NULL.

$-$8 Returned by spral_ssids_analyse() and spral_ssids_analyse_coord() if options.ordering is out-of-range, or options.ordering$=$0 and the user has either failed to provide an elimination order or an error has been found in the user-supplied elimination order (as supplied in order[]).

$-$9 Returned by spral_ssids_analyse() and spral_ssids_analyse_coord() if options.ordering$=$2, but val[] was NULL.

$-$10 Returned by spral_ssids_solve() if there is an error in the size of array x[][] (that is, ldx$<$n or nrhs$<$1). The user may reset ldx and/or nrhs and recall spral_ssids_solve().

$-$11 Returned by spral_ssids_solve() if job is out-of-range. The user may reset job and recall
spral_ssids_solve().

$-$12 Returned by spral_ssids_solve() and spral_ssids_alter() if the selected combination of options.use_gpu_solve and options.presolve are not compatible with the requested operation.

$-$13 Returned by spral_ssids_enquire_posdef() if posdef$=$false on the last call to spral_ssids_factor().

$-$14 Returned by spral_ssids_enquire_indef() if posdef$=$true on the last call to spral_ssids_factor().

$-$15 Returned by spral_ssids_factor() if options.scaling$=$3 but a matching based ordering was not used during the call to ssids_analyse() or ssids_analyse_coord() (i.e. was called with
options.ordering$\ne$2).

$-$50 Allocation error. If available, the Fortran stat parameter is returned in inform.stat.

$-$51 CUDA error. The CUDA error return value is returned in inform.cuda_error.

$-$52 CUBLAS error. The CUBLAS error return value is returned in inform.cublas_error.

A positive value of inform.flag is used to warn the user that the input matrix data may be faulty or that the subroutine cannot guarantee the solution obtained. Possible values are:

$+$1 Returned by spral_ssids_analyse() and spral_ssids_analyse_coord() if out-of-range variable indices found. Any such entries are ignored and the computation continues. inform.matrix_outrange is set to the number of such entries.

$+$2 Returned by spral_ssids_analyse() and spral_ssids_analyse_coord() if duplicated indices found. Duplicates are recorded and the corresponding entries are summed. inform.matrix_dup is set to the number of such entries.

$+$3 Returned by spral_ssids_analyse() and spral_ssids_analyse_coord() if both out-of-range and duplicated variable indices found.

$+$4 Returned by spral_ssids_analyse() and spral_ssids_analyse_coord() if one and more diagonal entries of $A$ is missing.

$+$5 Returned by spral_ssids_analyse() and spral_ssids_analyse_coord() if one and more diagonal entries of $A$ is missing and out-of-range and/or duplicated variable indices have been found.

$+$6 Returned by spral_ssids_analyse() and spral_ssids_analyse_coord() if $A$ is found be (structurally) singular. This will overwrite any of the above warnings.

$+$7 Returned by spral_ssids_factor() if options.action is set to true and the matrix is found to be (structurally or numerically) singular.

$+$8 Returned by spral_ssids_factor() if options.ordering$=$2 (i.e. a matching-based ordering was used) but the associated scaling was not (i.e. options.scaling$\ne$ 3).

5.7 Method

spral_ssids_analyse() and spral_ssids_analyse_coord()

If check is set to true on the call to spral_ssids_analyse() or if spral_ssids_analyse_coord() is called, the user-supplied matrix data is checked for errors. The cleaned integer matrix data (duplicates are summed and out-of-range indices discarded) is stored in akeep. The use of checking is optional on a call to spral_ssids_analyse() as it incurs both time and memory overheads. However, it is recommended since the behaviour of the other routines in the package is unpredictable if duplicates and/or out-of-range variable indices are entered.

If the user has supplied an elimination order it is checked for errors. Otherwise, an elimination order is generated by the package. The elimination order is used to construct an assembly tree. On exit from spral_ssids_analyse() (and spral_ssids_analyse_coord()), order[] is set so that order[i] holds the position of variable $i$ in the elimination order. If an ordering was supplied by the user, this order may differ, but will be equivalent in terms of fill-in.

spral_ssids_factor()

spral_ssids_factor() optionally computes a scaling and then performs the numerical factorization. The user must specify whether or not the matrix is positive definite. If posdef is set to true, no pivoting is performed and the computation will terminate with an error if a non-positive pivot is encountered.

The factorization uses the assembly tree that was set up by the analyse phase. At each node, entries from $A$ and, if it is not a leaf node, the generated elements and any delayed pivots from its child nodes must be assembled. Separate kernels handle each of these.

The kernel that performs the assembly from the child nodes considers one parent-child assembly at a time. Each generated element from a child is divided into a number of tiles, and a thread block launched to assemble each tile into a dense submatrix using a simple mapping array to determine the destination row and column of each entry. Bit-compatibility is achieved by ensuring the child entries are always assembled in the same order.

A dense partial factorization of the fully summed columns is then performed. The fully summed columns are split into a number of tiles that are each handled by an associated block. Factorization proceeds one column of tiles at a time. The pivoting condition is chosen to ensure that all entries of $L$ have absolute value less than ${\mathtt{\text{u}}}^{-1}$. This limits the growth of the entries of the $D$ factor and ensures that any solves will be backwards stable. The details are described in [1].

If a pivot candidate does not pass the pivot tests at a node, it is delayed to its parent node, where further elimination operations may make it acceptable. Delaying pivots leads to additional fill-in and floating-point operations beyond that predicted by spral_ssids_analyse() (and spral_ssids_analyse_coord()), and may result in additional memory allocations being required. The number of delayed pivots can often be reduced by using appropriate scaling.

At each non-root node, the majority of the floating-point operations involve the formation of the generated element. This is handled by a single dedicated kernel; again, see [1] for details.

At the end of the factorization, data structures for use in future calls to spral_ssids_solve() are prepared. If options.presolve=1, the block of $L$ corresponding to the eliminated variables is explicitly inverted to accelerate future calls to spral_ssids_solve() at the cost of making spral_ssids_factor() slower.

spral_ssids_solve()

If options.use_gpu_solve$=$false, data is moved to the CPU if required and the BLAS calls are used to perform a solve using the assembly tree and factors generated on previous calls.

Otherwise, the solve is conducted on the GPU in a similar fashion. If options.presolve=0, custom GPU implementations of _trsv() and _gemv() are used to handle multiple independent operations. If multiple right-hand sides are to be solved for, the single right-hand side solve is looped over. If options.presolve=1, _trsv() can be replaced by the much more parallel (and hence faster) _gemv(). In this case multiple right-hand sides are handled at the same time.

References

[1] J.D. Hogg, E. Ovtchinnikov and J.A. Scott. (2014). A sparse symmetric indefinite direct solver for GPU architectures. RAL Technical Report. RAL-P-2014-0xx, to appear.

5.8 Example

Suppose we wish to factorize the matrix

and then solve for the right-hand side

The following code may be used.

This produces the following output: