Contents
djadsm - Jagged-diagonal format triangular solve
SUBROUTINE DJADSM( TRANSA, M, N, UNITD, DV, ALPHA, DESCRA,
* VAL, INDX, PNTR, MAXNZ, IPERM,
* B, LDB, BETA, C, LDC, WORK, LWORK )
INTEGER TRANSA, M, N, UNITD, DESCRA(5), MAXNZ,
* LDB, LDC, LWORK
INTEGER INDX(NNZ), PNTR(MAXNZ+1), IPERM(M)
DOUBLE PRECISION ALPHA, BETA
DOUBLE PRECISION DV(M), VAL(NNZ), B(LDB,*), C(LDC,*), WORK(LWORK)
SUBROUTINE DJADSM_64( TRANSA, M, N, UNITD, DV, ALPHA, DESCRA,
* VAL, INDX, PNTR, MAXNZ, IPERM,
* B, LDB, BETA, C, LDC, WORK, LWORK )
INTEGER*8 TRANSA, M, N, UNITD, DESCRA(5), MAXNZ,
* LDB, LDC, LWORK
INTEGER*8 INDX(NNZ), PNTR(MAXNZ+1), IPERM(M)
DOUBLE PRECISION ALPHA, BETA
DOUBLE PRECISION DV(M), VAL(NNZ), B(LDB,*), C(LDC,*), WORK(LWORK)
where NNZ=PNTR(MAXNZ+1)-PNTR(1)+1 is the number of non-zero elements
F95 INTERFACE
SUBROUTINE JADSM(TRANSA, M, [N], UNITD, DV, ALPHA, DESCRA, VAL, INDX,
* PNTR, MAXNZ, IPERM, B, [LDB], BETA, C, [LDC], [WORK], [LWORK])
INTEGER TRANSA, M, MAXNZ
INTEGER, DIMENSION(:) :: DESCRA, INDX, PNTR, IPERM
DOUBLE PRECISION ALPHA, BETA
DOUBLE PRECISION, DIMENSION(:) :: VAL, DV
DOUBLE PRECISION, DIMENSION(:, :) :: B, C
SUBROUTINE JADSM_64(TRANSA, M, [N], UNITD, DV, ALPHA, DESCRA, VAL, INDX,
* PNTR, MAXNZ, IPERM, B, [LDB], BETA, C, [LDC], [WORK], [LWORK])
INTEGER*8 TRANSA, M, MAXNZ
INTEGER*8, DIMENSION(:) :: DESCRA, INDX, PNTR, IPERM
DOUBLE PRECISION ALPHA, BETA
DOUBLE PRECISION, DIMENSION(:) :: VAL, DV
DOUBLE PRECISION, DIMENSION(:, :) :: B, C
C INTERFACE
#include <sunperf.h>
void djadsm (int transa, int m, int n, int unitd, double
*dv, double alpha, int *descra, double *val, int *indx, int
*pntr, int maxnz, int *iperm, double *b, int ldb, double
beta, double *c, int ldc)
void djadsm_64(long transa, long m, long n, long unitd,
double *dv, double alpha, long *descra, double *val, long
*indx, long *pntr, long maxnz, long *iperm, double *b, long
ldb, double beta, double *c, long ldc);
djadsm performs one of the matrix-matrix operations
C <- alpha op(A) B + beta C, C <-alpha D op(A) B + beta C,
C <- alpha op(A) D B + beta C,
where alpha and beta are scalars, C and B are m by n dense matrices,
D is a diagonal scaling matrix, A is a sparse m by m unit, or non-unit,
upper or lower triangular matrix represented in the jagged-diagonal format
and op( A ) is one of
op( A ) = inv(A) or op( A ) = inv(A') or op( A ) =inv(conjg( A' ))
(inv denotes matrix inverse, ' indicates matrix transpose).
TRANSA(input) TRANSA specifies the form of op( A ) to be used in
the sparse matrix inverse as follows:
0 : operate with matrix
1 : operate with transpose matrix
2 : operate with the conjugate transpose of matrix.
2 is equivalent to 1 if matrix is real.
Unchanged on exit.
M(input) On entry, M specifies the number of rows in
the matrix A. Unchanged on exit.
N(input) On entry, N specifies the number of columns in
the matrix C. Unchanged on exit.
UNITD(input) On entry, UNITD specifies the type of scaling:
1 : Identity matrix (argument DV[] is ignored)
2 : Scale on left (row scaling)
3 : Scale on right (column scaling)
4 : Automatic row scaling (see section NOTES for
further details)
Unchanged on exit.
DV(input) On entry, DV is an array of length M consisting of the
diagonal entries of the diagonal scaling matrix D.
If UNITD is 4, DV contains diagonal matrix by which
the rows have been scaled (see section NOTES for further
details). Otherwise, unchanged on exit.
ALPHA(input) On entry, ALPHA specifies the scalar alpha. Unchanged on exit.
DESCRA (input) Descriptor argument. Five element integer array:
DESCRA(1) matrix structure
0 : general
1 : symmetric (A=A')
2 : Hermitian (A= CONJG(A'))
3 : Triangular
4 : Skew(Anti)-Symmetric (A=-A')
5 : Diagonal
6 : Skew-Hermitian (A= -CONJG(A'))
Note: For the routine, DESCRA(1)=3 is only supported.
DESCRA(2) upper/lower triangular indicator
1 : lower
2 : upper
DESCRA(3) main diagonal type
0 : non-unit
1 : unit
DESCRA(4) Array base (NOT IMPLEMENTED)
0 : C/C++ compatible
1 : Fortran compatible
DESCRA(5) repeated indices? (NOT IMPLEMENTED)
0 : unknown
1 : no repeated indices
VAL(input) On entry, VAL is a scalar array of length
NNZ=PNTR(MAXNZ+1)-PNTR(1)+1 consisting of entries of A.
VAL can be viewed as a column major ordering of a
row permutation of the Ellpack representation of A,
where the Ellpack representation is permuted so that
the rows are non-increasing in the number of nonzero
entries. Values added for padding in Ellpack are
not included in the Jagged-Diagonal format.
Unchanged on exit if UNITD is not equal to 4.
INDX(input) On entry, INDX is an integer array of length
NNZ=PNTR(MAXNZ+1)-PNTR(1)+1 consisting of the column
indices of the corresponding entries in VAL.
Unchanged on exit.
PNTR(input) On entry, PNTR is an integer array of length
MAXNZ+1, where PNTR(I)-PNTR(1)+1 points to
the location in VAL of the first element
in the row-permuted Ellpack represenation of A.
Unchanged on exit.
MAXNZ(input) On entry, MAXNZ specifies the max number of
nonzeros elements per row. Unchanged on exit.
IPERM(input) On entry, IPERM is an integer array of length M
such that I = IPERM(I'), where row I in the
original Ellpack representation corresponds
to row I' in the permuted representation.
If IPERM(1) = 0, it is assumed by convention that
IPERM(I) = I. IPERM is used to determine the order
in which rows of C are updated. Unchanged on exit.
B (input) Array of DIMENSION ( LDB, N ).
On entry, the leading m by n part of the array B
must contain the matrix B. Unchanged on exit.
LDB (input) On entry, LDB specifies the first dimension of B as declared
in the calling (sub) program. Unchanged on exit.
BETA (input) On entry, BETA specifies the scalar beta. Unchanged on exit.
C(input/output) Array of DIMENSION ( LDC, N ).
On entry, the leading m by n part of the array C
must contain the matrix C. On exit, the array C is
overwritten.
LDC (input) On entry, LDC specifies the first dimension of C as declared
in the calling (sub) program. Unchanged on exit.
WORK(workspace) Scratch array of length LWORK.
On exit, if LWORK= -1, WORK(1) returns the optimum size
of LWORK.
LWORK (input) On entry, LWORK specifies the length of WORK array. LWORK
should be at least 2*M.
For good performance, LWORK should generally be larger.
For optimum performance on multiple processors, LWORK
>=2*M*N_CPUS where N_CPUS is the maximum number of
processors available to the program.
If LWORK=0, the routine is to allocate workspace needed.
If LWORK = -1, then a workspace query is assumed; the
routine only calculates the optimum size of the WORK array,
returns this value as the first entry of the WORK array,
and no error message related to LWORK is issued by XERBLA.
Libsunperf SPARSE BLAS is parallelized with the help of OPENMP and it is
fully compatible with NIST FORTRAN Sparse Blas but the sources are different.
Libsunperf SPARSE BLAS is free of bugs found in NIST FORTRAN Sparse Blas.
Besides several new features and routines are implemented.
NIST FORTRAN Sparse Blas User's Guide available at:
http://math.nist.gov/mcsd/Staff/KRemington/fspblas/
Based on the standard proposed in
"Document for the Basic Linear Algebra Subprograms (BLAS)
Standard", University of Tennessee, Knoxville, Tennessee, 1996:
http://www.netlib.org/utk/papers/sparse.ps
NOTES/BUGS
1. No test for singularity or near-singularity is included
in this routine. Such tests must be performed before calling
this routine.
2. If UNITD =4, the routine scales the rows of the sparse
matrix A such that their 2-norms are one. The scaling may
improve the accuracy of the computed solution. Corresponding
entries of VAL are changed only in the particular case. On
return DV matrix stored as a vector contains the diagonal
matrix by which the rows have been scaled. UNITD=2 should be
used for the next calls to the routine with overwritten VAL
and DV.
WORK(1)=0 on return if the scaling has been completed
successfully, otherwise WORK(1) = - i where i is the row
number which 2-norm is exactly zero.
3. If DESCRA(3)=1 and UNITD < 4, the diagonal entries are
each used with the mathematical value 1. The entries of the
main diagonal in the JAD representation of a sparse matrix
do not need to be 1.0 in this usage. They are not used by
the routine in these cases. But if UNITD=4, the unit
diagonal elements MUST be referenced in the JAD
representation.
4. The routine is designed so that it checks the validity of
each sparse entry given in the sparse blas representation.
Entries with incorrect indices are not used and no error
message related to the entries is issued.
The feature also provides a possibility to use the sparse
matrix representation of a general matrix A for solving
triangular systems with the upper or lower triangle of A.
But DESCRA(1) MUST be equal to 3 even in this case.
Assume that there is the sparse matrix representation a
general matrix A decomposed in the form
A = L + D + U
where L is the strictly lower triangle of A, U is the
strictly upper triangle of A, D is the diagonal matrix.
Let's I denotes the identity matrix.
Then the correspondence between the first three values of
DESCRA and the result matrix for the sparse representation
of A is
DESCRA(1) DESCRA(2) DESCRA(3) RESULT
3 1 1 alpha*op(L+I)*B+beta*C
3 1 0 alpha*op(L+D)*B+beta*C
3 2 1 alpha*op(U+I)*B+beta*C
3 2 0 alpha*op(U+D)*B+beta*C