Uses the LU factorization of the
nmatrix computed by
?getc2and computes a contribution to the reciprocal Dif-estimate.
LUfactorization of the
?getc2and computes a contribution to the reciprocal Dif-estimate by solving
x, and choosing the right-hand side
bsuch that the norm of
xis as large as possible. On entry
holds the contribution from earlier solved sub-systems, and on return
The factorization of
?getc2has the form
Qare permutation matrices.
Lis lower triangular with unit diagonal elements and
Uis upper triangular.
- INTEGER.: First compute an approximative null-vectorijob= 2eofZusing?gecon,eis normalized, and solve forwith the sign giving the greater value of 2-norm(-Z*x=±efx). This option is about 5 times as expensive as default.(default): Local look ahead strategy where all entries of the right-hand sideijob≠2bis chosen as either +1 or -1 .
- INTEGER. The number of columns of the matrixZ.
- REALforslatdf/clatdfDOUBLE PRECISIONfordlatdf/zlatdf.Array,DIMENSION(ldz,n)On entry, theLUpart of the factorization of then-by-nmatrixZcomputed by?getc2:.Z=P*L*U*Q
- INTEGER. The leading dimension of the arrayZ..lda≥max(1,n)
- REALforslatdf/clatdfDOUBLE PRECISIONfordlatdf/zlatdf.Array,DIMENSION(n).On entry,rhscontains contributions from other subsystems.
- REALforslatdf/clatdfDOUBLE PRECISIONfordlatdf/zlatdf.On entry, the sum of squares of computed contributions to theDif-estimate under computation by?tgsyL, where the scaling factorrdscalhas been factored out. If,trans='T'rdsumis not touched.Note thatrdsumonly makes sense when?tgsy2is called by?tgsyL.