dsgesv()

subroutine dsgesv	(	integer	N,
		integer	NRHS,
		double precision, dimension( lda, * )	A,
		integer	LDA,
		integer, dimension( * )	IPIV,
		double precision, dimension( ldb, * )	B,
		integer	LDB,
		double precision, dimension( ldx, * )	X,
		integer	LDX,
		double precision, dimension( n, * )	WORK,
		real, dimension( * )	SWORK,
		integer	ITER,
		integer	INFO
	)

DSGESV computes the solution to system of linear equations A * X = B for GE matrices (mixed precision with iterative refinement)

Download DSGESV + dependencies [TGZ] [ZIP] [TXT]

Purpose:

 DSGESV computes the solution to a real system of linear equations
    A * X = B,
 where A is an N-by-N matrix and X and B are N-by-NRHS matrices.

 DSGESV first attempts to factorize the matrix in SINGLE PRECISION
 and use this factorization within an iterative refinement procedure
 to produce a solution with DOUBLE PRECISION normwise backward error
 quality (see below). If the approach fails the method switches to a
 DOUBLE PRECISION factorization and solve.

 The iterative refinement is not going to be a winning strategy if
 the ratio SINGLE PRECISION performance over DOUBLE PRECISION
 performance is too small. A reasonable strategy should take the
 number of right-hand sides and the size of the matrix into account.
 This might be done with a call to ILAENV in the future. Up to now, we
 always try iterative refinement.

 The iterative refinement process is stopped if
     ITER > ITERMAX
 or for all the RHS we have:
     RNRM < SQRT(N)*XNRM*ANRM*EPS*BWDMAX
 where
     o ITER is the number of the current iteration in the iterative
       refinement process
     o RNRM is the infinity-norm of the residual
     o XNRM is the infinity-norm of the solution
     o ANRM is the infinity-operator-norm of the matrix A
     o EPS is the machine epsilon returned by DLAMCH('Epsilon')
 The value ITERMAX and BWDMAX are fixed to 30 and 1.0D+00
 respectively.

Parameters

[in]	N	N is INTEGER The number of linear equations, i.e., the order of the matrix A. N >= 0.
[in]	NRHS	NRHS is INTEGER The number of right hand sides, i.e., the number of columns of the matrix B. NRHS >= 0.
[in,out]	A	A is DOUBLE PRECISION array, dimension (LDA,N) On entry, the N-by-N coefficient matrix A. On exit, if iterative refinement has been successfully used (INFO = 0 and ITER >= 0, see description below), then A is unchanged, if double precision factorization has been used (INFO = 0 and ITER < 0, see description below), then the array A contains the factors L and U from the factorization A = P*L*U; the unit diagonal elements of L are not stored.
[in]	LDA	LDA is INTEGER The leading dimension of the array A. LDA >= max(1,N).
[out]	IPIV	IPIV is INTEGER array, dimension (N) The pivot indices that define the permutation matrix P; row i of the matrix was interchanged with row IPIV(i). Corresponds either to the single precision factorization (if INFO = 0 and ITER >= 0) or the double precision factorization (if INFO = 0 and ITER < 0).
[in]	B	B is DOUBLE PRECISION array, dimension (LDB,NRHS) The N-by-NRHS right hand side matrix B.
[in]	LDB	LDB is INTEGER The leading dimension of the array B. LDB >= max(1,N).
[out]	X	X is DOUBLE PRECISION array, dimension (LDX,NRHS) If INFO = 0, the N-by-NRHS solution matrix X.
[in]	LDX	LDX is INTEGER The leading dimension of the array X. LDX >= max(1,N).
[out]	WORK	WORK is DOUBLE PRECISION array, dimension (N,NRHS) This array is used to hold the residual vectors.
[out]	SWORK	SWORK is REAL array, dimension (N*(N+NRHS)) This array is used to use the single precision matrix and the right-hand sides or solutions in single precision.
[out]	ITER	ITER is INTEGER < 0: iterative refinement has failed, double precision factorization has been performed -1 : the routine fell back to full precision for implementation- or machine-specific reasons -2 : narrowing the precision induced an overflow, the routine fell back to full precision -3 : failure of SGETRF -31: stop the iterative refinement after the 30th iterations > 0: iterative refinement has been successfully used. Returns the number of iterations
[out]	INFO	INFO is INTEGER = 0: successful exit < 0: if INFO = -i, the i-th argument had an illegal value > 0: if INFO = i, U(i,i) computed in DOUBLE PRECISION is exactly zero. The factorization has been completed, but the factor U is exactly singular, so the solution could not be computed.

Author

Univ. of Tennessee

Univ. of California Berkeley

Univ. of Colorado Denver

NAG Ltd.

Date

June 2016

Definition at line 197 of file dsgesv.f.

*
*  -- LAPACK driver routine (version 3.8.0) --
*  -- LAPACK is a software package provided by Univ. of Tennessee,    --
*  -- Univ. of California Berkeley, Univ. of Colorado Denver and NAG Ltd..--
*     June 2016
*
*     .. Scalar Arguments ..
      INTEGER            INFO, ITER, LDA, LDB, LDX, N, NRHS
*     ..
*     .. Array Arguments ..
      INTEGER            IPIV( * )
      REAL               SWORK( * )
      DOUBLE PRECISION   A( LDA, * ), B( LDB, * ), WORK( N, * ),
     $                   X( LDX, * )
*     ..
*
*  =====================================================================
*
*     .. Parameters ..
      LOGICAL            DOITREF
      parameter( doitref = .true. )
*
      INTEGER            ITERMAX
      parameter( itermax = 30 )
*
      DOUBLE PRECISION   BWDMAX
      parameter( bwdmax = 1.0e+00 )
*
      DOUBLE PRECISION   NEGONE, ONE
      parameter( negone = -1.0d+0, one = 1.0d+0 )
*
*     .. Local Scalars ..
      INTEGER            I, IITER, PTSA, PTSX
      DOUBLE PRECISION   ANRM, CTE, EPS, RNRM, XNRM
*
*     .. External Subroutines ..
      EXTERNAL           daxpy, dgemm, dlacpy, dlag2s, dgetrf, dgetrs,
     $                   sgetrf, sgetrs, slag2d, xerbla
*     ..
*     .. External Functions ..
      INTEGER            IDAMAX
      DOUBLE PRECISION   DLAMCH, DLANGE
      EXTERNAL           idamax, dlamch, dlange
*     ..
*     .. Intrinsic Functions ..
      INTRINSIC          abs, dble, max, sqrt
*     ..
*     .. Executable Statements ..
*
      info = 0
      iter = 0
*
*     Test the input parameters.
*
      IF( n.LT.0 ) THEN
         info = -1
      ELSE IF( nrhs.LT.0 ) THEN
         info = -2
      ELSE IF( lda.LT.max( 1, n ) ) THEN
         info = -4
      ELSE IF( ldb.LT.max( 1, n ) ) THEN
         info = -7
      ELSE IF( ldx.LT.max( 1, n ) ) THEN
         info = -9
      END IF
      IF( info.NE.0 ) THEN
         CALL xerbla( 'DSGESV', -info )
         RETURN
      END IF
*
*     Quick return if (N.EQ.0).
*
      IF( n.EQ.0 )
     $   RETURN
*
*     Skip single precision iterative refinement if a priori slower
*     than double precision factorization.
*
      IF( .NOT.doitref ) THEN
         iter = -1
         GO TO 40
      END IF
*
*     Compute some constants.
*
      anrm = dlange( 'I', n, n, a, lda, work )
      eps = dlamch( 'Epsilon' )
      cte = anrm*eps*sqrt( dble( n ) )*bwdmax
*
*     Set the indices PTSA, PTSX for referencing SA and SX in SWORK.
*
      ptsa = 1
      ptsx = ptsa + n*n
*
*     Convert B from double precision to single precision and store the
*     result in SX.
*
      CALL dlag2s( n, nrhs, b, ldb, swork( ptsx ), n, info )
*
      IF( info.NE.0 ) THEN
         iter = -2
         GO TO 40
      END IF
*
*     Convert A from double precision to single precision and store the
*     result in SA.
*
      CALL dlag2s( n, n, a, lda, swork( ptsa ), n, info )
*
      IF( info.NE.0 ) THEN
         iter = -2
         GO TO 40
      END IF
*
*     Compute the LU factorization of SA.
*
      CALL sgetrf( n, n, swork( ptsa ), n, ipiv, info )
*
      IF( info.NE.0 ) THEN
         iter = -3
         GO TO 40
      END IF
*
*     Solve the system SA*SX = SB.
*
      CALL sgetrs( 'No transpose', n, nrhs, swork( ptsa ), n, ipiv,
     $             swork( ptsx ), n, info )
*
*     Convert SX back to double precision
*
      CALL slag2d( n, nrhs, swork( ptsx ), n, x, ldx, info )
*
*     Compute R = B - AX (R is WORK).
*
      CALL dlacpy( 'All', n, nrhs, b, ldb, work, n )
*
      CALL dgemm( 'No Transpose', 'No Transpose', n, nrhs, n, negone, a,
     $            lda, x, ldx, one, work, n )
*
*     Check whether the NRHS normwise backward errors satisfy the
*     stopping criterion. If yes, set ITER=0 and return.
*
      DO i = 1, nrhs
         xnrm = abs( x( idamax( n, x( 1, i ), 1 ), i ) )
         rnrm = abs( work( idamax( n, work( 1, i ), 1 ), i ) )
         IF( rnrm.GT.xnrm*cte )
     $      GO TO 10
      END DO
*
*     If we are here, the NRHS normwise backward errors satisfy the
*     stopping criterion. We are good to exit.
*
      iter = 0
      RETURN
*
   10 CONTINUE
*
      DO 30 iiter = 1, itermax
*
*        Convert R (in WORK) from double precision to single precision
*        and store the result in SX.
*
         CALL dlag2s( n, nrhs, work, n, swork( ptsx ), n, info )
*
         IF( info.NE.0 ) THEN
            iter = -2
            GO TO 40
         END IF
*
*        Solve the system SA*SX = SR.
*
         CALL sgetrs( 'No transpose', n, nrhs, swork( ptsa ), n, ipiv,
     $                swork( ptsx ), n, info )
*
*        Convert SX back to double precision and update the current
*        iterate.
*
         CALL slag2d( n, nrhs, swork( ptsx ), n, work, n, info )
*
         DO i = 1, nrhs
            CALL daxpy( n, one, work( 1, i ), 1, x( 1, i ), 1 )
         END DO
*
*        Compute R = B - AX (R is WORK).
*
         CALL dlacpy( 'All', n, nrhs, b, ldb, work, n )
*
         CALL dgemm( 'No Transpose', 'No Transpose', n, nrhs, n, negone,
     $               a, lda, x, ldx, one, work, n )
*
*        Check whether the NRHS normwise backward errors satisfy the
*        stopping criterion. If yes, set ITER=IITER>0 and return.
*
         DO i = 1, nrhs
            xnrm = abs( x( idamax( n, x( 1, i ), 1 ), i ) )
            rnrm = abs( work( idamax( n, work( 1, i ), 1 ), i ) )
            IF( rnrm.GT.xnrm*cte )
     $         GO TO 20
         END DO
*
*        If we are here, the NRHS normwise backward errors satisfy the
*        stopping criterion, we are good to exit.
*
         iter = iiter
*
         RETURN
*
   20    CONTINUE
*
   30 CONTINUE
*
*     If we are at this place of the code, this is because we have
*     performed ITER=ITERMAX iterations and never satisfied the
*     stopping criterion, set up the ITER flag accordingly and follow up
*     on double precision routine.
*
      iter = -itermax - 1
*
   40 CONTINUE
*
*     Single-precision iterative refinement failed to converge to a
*     satisfactory solution, so we resort to double precision.
*
      CALL dgetrf( n, n, a, lda, ipiv, info )
*
      IF( info.NE.0 )
     $   RETURN
*
      CALL dlacpy( 'All', n, nrhs, b, ldb, x, ldx )
      CALL dgetrs( 'No transpose', n, nrhs, a, lda, ipiv, x, ldx,
     $             info )
*
      RETURN
*
*     End of DSGESV.
*

Here is the call graph for this function:

Here is the caller graph for this function: