F07 Chapter Contents
F07 Chapter Introduction
NAG Library Manual

# NAG Library Routine DocumentF07ACF (DSGESV)

Note:  before using this routine, please read the Users' Note for your implementation to check the interpretation of bold italicised terms and other implementation-dependent details.

## 1  Purpose

F07ACF (DSGESV) computes the solution to a real system of linear equations
 $AX=B ,$
where $A$ is an $n$ by $n$ matrix and $X$ and $B$ are $n$ by $r$ matrices.

## 2  Specification

 SUBROUTINE F07ACF ( N, NRHS, A, LDA, IPIV, B, LDB, X, LDX, WORK, SWORK, ITER, INFO)
 INTEGER N, NRHS, LDA, IPIV(N), LDB, LDX, ITER, INFO REAL (KIND=nag_wp) A(LDA,*), B(LDB,*), X(LDX,*), WORK(N*NRHS) REAL (KIND=nag_rp) SWORK(N*(N+NRHS))
The routine may be called by its LAPACK name dsgesv.

## 3  Description

F07ACF (DSGESV) first attempts to factorize the matrix in single precision and use this factorization within an iterative refinement procedure to produce a solution with full double precision accuracy. If the approach fails the method switches to a double precision factorization and solve.
The iterative refinement process is stopped if
 $ITER>itermax ,$
where ITER is the number of iterations carried out thus far and $\mathit{itermax}$ is the maximum number of iterations allowed, which is fixed at $30$ iterations. The process is also stopped if for all right-hand sides we have
 $resid < N x A ε ,$
where $‖\mathit{resid}‖$ is the $\infty$-norm of the residual, $‖x‖$ is the $\infty$-norm of the solution, $‖A‖$ is the $\infty$-operator-norm of the matrix $A$ and $\epsilon$ is the machine precision returned by X02AJF.
The iterative refinement strategy used by F07ACF (DSGESV) can be more efficient than the corresponding direct full precision algorithm. Since this strategy must perform iterative refinement on each right-hand side, any efficiency gains will reduce as the number of right-hand sides increases. Conversely, as the matrix size increases the cost of these iterative refinements become less significant relative to the cost of factorization. Thus, any efficiency gains will be greatest for a very small number of right-hand sides and for large matrix sizes. The cut-off values for the number of right-hand sides and matrix size, for which the iterative refinement strategy performs better, depends on the relative performance of the reduced and full precision factorization and back-substitution. For now, F07ACF (DSGESV) always attempts the iterative refinement strategy first; you are advised to compare the performance of F07ACF (DSGESV) with that of its full precision counterpart F07AAF (DGESV) to determine whether this strategy is worthwhile for your particular problem dimensions.

## 4  References

Anderson E, Bai Z, Bischof C, Blackford S, Demmel J, Dongarra J J, Du Croz J J, Greenbaum A, Hammarling S, McKenney A and Sorensen D (1999) LAPACK Users' Guide (3rd Edition) SIAM, Philadelphia http://www.netlib.org/lapack/lug
Buttari A, Dongarra J, Langou J, Langou J, Luszczek P and Kurzak J (2007) Mixed precision iterative refinement techniques for the solution of dense linear systems International Journal of High Performance Computing Applications
Golub G H and Van Loan C F (1996) Matrix Computations (3rd Edition) Johns Hopkins University Press, Baltimore

## 5  Parameters

1:     N – INTEGERInput
On entry: $n$, the number of linear equations, i.e., the order of the matrix $A$.
Constraint: ${\mathbf{N}}\ge 0$.
2:     NRHS – INTEGERInput
On entry: $r$, the number of right-hand sides, i.e., the number of columns of the matrix $B$.
Constraint: ${\mathbf{NRHS}}\ge 0$.
3:     A(LDA,$*$) – REAL (KIND=nag_wp) arrayInput/Output
Note: the second dimension of the array A must be at least $\mathrm{max}\phantom{\rule{0.125em}{0ex}}\left(1,{\mathbf{N}}\right)$.
On entry: the $n$ by $n$ coefficient matrix $A$.
On exit: if iterative refinement has been successfully used (i.e., if ${\mathbf{INFO}}={\mathbf{0}}$ and ${\mathbf{ITER}}\ge 0$), then $A$ is unchanged. If double precision factorization has been used (when ${\mathbf{INFO}}={\mathbf{0}}$ and ${\mathbf{ITER}}<0$), $A$ contains the factors $L$ and $U$ from the factorization $A=PLU$; the unit diagonal elements of $L$ are not stored.
4:     LDA – INTEGERInput
On entry: the first dimension of the array A as declared in the (sub)program from which F07ACF (DSGESV) is called.
Constraint: ${\mathbf{LDA}}\ge \mathrm{max}\phantom{\rule{0.125em}{0ex}}\left(1,{\mathbf{N}}\right)$.
5:     IPIV(N) – INTEGER arrayOutput
On exit: if no constraints are violated, the pivot indices that define the permutation matrix $P$; at the $i$th step row $i$ of the matrix was interchanged with row ${\mathbf{IPIV}}\left(i\right)$. ${\mathbf{IPIV}}\left(i\right)=i$ indicates a row interchange was not required. ${\mathbf{IPIV}}$ corresponds either to the single precision factorization (if ${\mathbf{INFO}}={\mathbf{0}}$ and ${\mathbf{ITER}}\ge 0$) or to the double precision factorization (if ${\mathbf{INFO}}={\mathbf{0}}$ and ${\mathbf{ITER}}<0$).
6:     B(LDB,$*$) – REAL (KIND=nag_wp) arrayInput
Note: the second dimension of the array B must be at least $\mathrm{max}\phantom{\rule{0.125em}{0ex}}\left(1,{\mathbf{NRHS}}\right)$.
On entry: the $n$ by $r$ right-hand side matrix $B$.
7:     LDB – INTEGERInput
On entry: the first dimension of the array B as declared in the (sub)program from which F07ACF (DSGESV) is called.
Constraint: ${\mathbf{LDB}}\ge \mathrm{max}\phantom{\rule{0.125em}{0ex}}\left(1,{\mathbf{N}}\right)$.
8:     X(LDX,$*$) – REAL (KIND=nag_wp) arrayOutput
Note: the second dimension of the array X must be at least $\mathrm{max}\phantom{\rule{0.125em}{0ex}}\left(1,{\mathbf{NRHS}}\right)$.
On exit: if ${\mathbf{INFO}}={\mathbf{0}}$, the $n$ by $r$ solution matrix $X$.
9:     LDX – INTEGERInput
On entry: the first dimension of the array X as declared in the (sub)program from which F07ACF (DSGESV) is called.
Constraint: ${\mathbf{LDX}}\ge \mathrm{max}\phantom{\rule{0.125em}{0ex}}\left(1,{\mathbf{N}}\right)$.
10:   WORK(${\mathbf{N}}*{\mathbf{NRHS}}$) – REAL (KIND=nag_wp) arrayWorkspace
11:   SWORK(${\mathbf{N}}×\left({\mathbf{N}}+{\mathbf{NRHS}}\right)$) – REAL (KIND=nag_rp) arrayWorkspace
Note: this array is utilized in the reduced precision computation, consequently its type nag_rp reflects this usage.
12:   ITER – INTEGEROutput
On exit: if ${\mathbf{ITER}}>0$, iterative refinement has been successfully used and ITER is the number of iterations carried out.
If ${\mathbf{ITER}}<0$, iterative refinement has failed for one of the reasons given below and double precision factorization has been carried out instead.
${\mathbf{ITER}}=-1$
Taking into account machine parameters, and the values of N and NRHS, it is not worth working in single precision.
${\mathbf{ITER}}=-2$
Overflow of an entry occurred when moving from double to single precision.
${\mathbf{ITER}}=-3$
An intermediate single precision factorization failed.
${\mathbf{ITER}}=-31$
The maximum permitted number of iterations was exceeded.
13:   INFO – INTEGEROutput
On exit: ${\mathbf{INFO}}=0$ unless the routine detects an error (see Section 6).

## 6  Error Indicators and Warnings

Errors or warnings detected by the routine:
${\mathbf{INFO}}<0$
If ${\mathbf{INFO}}=-i$, the $i$th argument had an illegal value. An explanatory message is output, and execution of the program is terminated.
${\mathbf{INFO}}>0$
If ${\mathbf{INFO}}=i$, ${u}_{ii}$ is exactly zero. The factorization has been completed, but the factor $U$ is exactly singular, so the solution could not be computed.

## 7  Accuracy

The computed solution for a single right-hand side, $\stackrel{^}{x}$, satisfies the equation of the form
 $A+E x^=b ,$
where
 $E1 = Oε A1$
and $\epsilon$ is the machine precision. An approximate error bound for the computed solution is given by
 $x^ - x 1 x 1 ≤ κA E 1 A 1$
where $\kappa \left(A\right)={‖{A}^{-1}‖}_{1}{‖A‖}_{1}$, the condition number of $A$ with respect to the solution of the linear equations. See Section 4.4 of Anderson et al. (1999) for further details.

The complex analogue of this routine is F07AQF (ZCGESV).

## 9  Example

This example solves the equations
 $Ax = b ,$
where $A$ is the general matrix
 $A = 1.80 2.88 2.05 -0.89 5.25 -2.95 -0.95 -3.80 1.58 -2.69 -2.90 -1.04 -1.11 -0.66 -0.59 0.80 and b = 9.52 24.35 0.77 -6.22 .$

### 9.1  Program Text

Program Text (f07acfe.f90)

### 9.2  Program Data

Program Data (f07acfe.d)

### 9.3  Program Results

Program Results (f07acfe.r)