NAME

blas_dpermute - permutes a real (double precision) array in terms of the permutation vector P, output by dsortv

SYNOPSIS

  SUBROUTINE BLAS_DPERMUTE( N, P, INCP, X, INCX)
  INTEGER N, INCP, INCX
  INTEGER P(*)
  DOUBLE PRECISION X(*)

  SUBROUTINE BLAS_DPERMUTE_64( N, P, INCP, X, INCX)
  INTEGER*8 N, INCP, INCX
  INTEGER*8 P(*)
  DOUBLE PRECISION X(*)

  SUBROUTINE PERMUTE( [N], P, [INCP], X, [INCX])
  INTEGER :: N, INCP, INCX
  INTEGER, DIMENSION(:) :: P
  REAL(8), DIMENSION(:) :: X

  SUBROUTINE PERMUTE_64( [N], P, [INCP], X, [INCX])
  INTEGER(8) :: N, INCP, INCX
  INTEGER(8), DIMENSION(:) :: P
  REAL(8), DIMENSION(:) :: X

#include <sunperf.h>

void blas_dpermute(int n, int *p, int incp, double *x, int incx);

void blas_dpermute_64(long n, long *p, long incp, double *x, long incx);

N (input)
If N < = 1, the subroutine returns without trying to permute X.
P (input)
vector defined follows the same conventions as that for DTYPE SORTV. It records the details of the interchanges of the elements of X during sorting. That is X = P*X. In current implementation, P contains the index of sorted X.

INCP (input)
INCP must not be zero. INCP could be negative. If INCP < 0, the permutation is applied in the opposite direction. That is

If INCP > 0,

  if INCX  > 0,

    sorted X((i-1)*INCX+1)  = X(P((i-1)*INCP+1)),

  if INCX  < 0,

    sorted X((N-i)*|INCX|+1)  = X(P((i-1)*INCP+1));

If INCP < 0,

  if INCX  > 0,

    sorted X((i-1)*INCX+1)  = X(P((N-i)*|INCP|+1)).

  if INCX  < 0,

    sorted X((N-i)*|INCX|+1)

          = X(P((N-i)*|INCP|+1)).

X (input/output)
be permuted. Minimum size (N-1)*|INCX|+1 is required
INCX (input)
INCX must not be zero. INCX could be negative. If INCX < 0, X will be permuted in a reverse way (see the description for INCP above).