Fortran: help with arguments to a pre-built software

brydustin

There are two functions that I need help with:

http://www.maths.uq.edu.au/expokit/fortran/dgpadm.f
http://www.maths.uq.edu.au/expokit/fortran/dgchbv.f

The problem I'm having is that the arguments on the line DGPADM(....) don't match identically with the explanation. For example, H, ldh and m vs. m and H(ldh,m) not sure why there is a parenthesis. If someone could give me a simple example of acceptable parameters that I could try out,that would be greatly appreciated.

I understand what to put for ideg, m, t, but not so sure about wsp, ipiv , H or the others. For example, H is a matrix, so I guess it needs a matrix as an argument..... regardless could someone help me out here?! Thanks!

In the first case:
subroutine DGPADM( ideg,m,t,H,ldh,wsp,lwsp,ipiv,iexph,ns,iflag )

* ideg : (input) the degre of the diagonal Pade to be used.
* a value of 6 is generally satisfactory.
*
* m : (input) order of H.
*
* H(ldh,m) : (input) argument matrix.
*
* t : (input) time-scale (can be < 0).
*
* wsp(lwsp) : (workspace/output) lwsp .ge. 4*m*m+ideg+1.
*
* ipiv(m) : (workspace)
*
*>>>> iexph : (output) number such that wsp(iexph) points to exp(tH)
* i.e., exp(tH) is located at wsp(iexph ... iexph+m*m-1)
* ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
* NOTE: if the routine was called with wsp(iptr),
* then exp(tH) will start at wsp(iptr+iexph-1).
*
* ns : (output) number of scaling-squaring used.
*
* iflag : (output) exit flag.
* 0 - no problem
* <0 - problem


Second case:
subroutine DGCHBV( m, t, H,ldh, y, wsp, iwsp, iflag )

* m : (input) order of the matrix H
*
* t : (input) time-scaling factor (can be < 0).
*
* H(ldh,m): (input) argument matrix.
*
* y(m) : (input/output) on input the operand vector,
* on output the resulting vector exp(t*H)*y.
*
* iwsp(m) : (workspace)
*
* wsp : (workspace). Observe that a double precision vector of
* length 2*m*(m+2) can be used as well when calling this
* routine (thus avoiding an idle complex array elsewhere)

Mark44

Mod note: moved to Prog & CS section

Mark44

I'll take a stab at the first one.

H - the name of a two-dimensional matrix - the matrix contains double precision numbers.
ldh - # of rows in matrix H - an integer.
m - # of cols in matrix H - an integer.
t - timescale. This is the constant t in e^tH -- double precision.
wsp - workspace - an output array with lwsp elements, each of type double precision.
lwsp - # of elements of the wsp array - an integer. It needs to be > 4m² + ideg + 1.
ipiv - an output array with m integer elements.
iexph - an index into the wsp array - an integer.
The computed elements of e^tH are located in the wsp array. They start at wsp(iexph) and end at wsp(iexph + m² - 1). This section of the array contains all of the elements of e^tH.
ns - no clue - integer.
iflag - a flag that indicates whether the calculation was successful or not. Negative value indicates some kind of error.

I would start with a very simple matrix, like
$$H = \begin{bmatrix} 0 & 1 \\ 0 & 0 \end{bmatrix}$$
and a value of 1.0 for t.

You should end up with
$$e^{tH} = \begin{bmatrix} 1 & 1 \\ 0 & 1 \end{bmatrix}$$

The variables that are identified as being output don't need to be initialized before calling this routine. The other variables need to be initialized before the call.