AnsweredAssumed Answered

VRf - Compiled functions in C and F77

Question asked by VRFuser on Aug 28, 1997

Like • Show 0 Likes0
Comment • 2

Hi VRf listeners,

I'm currently trying to pass an HP-VEE complex array to a compiled C
function. However, the type complex is not supported in C and VEE
doesn't like to pass complex to a compiled function. Does anybody
know how this could be done ?

Also, the HP-VEE documentation is not very clear on the requirements
for header files for FORTRAN compiled functions. Is there a better
definition somewhere ? Or does anybody have good examples ?

Regards

Michel Gagne
COM DEV Ltd
Cambridge, Canada
(519) 622-2300 x 575

Mike.gagne@comdev.ca

No one else has this question

Outcomes

2 Replies

VRFuser Aug 27, 1997 9:58 PM
Mark Correct
Correct Answer

------------------------
From: Mike Gagne <Mike.Gagne@comdev.ca>
Subject: VRf - Compiled functions in C and F77
Date: Thu, 28 Aug 1997 15:36:21 EST5EDT
To: hpvxd_xc@hpislsup.lvld.hp.com

> Hi VRf listeners,
>
> I'm currently trying to pass an HP-VEE complex array to a compiled C
> function. However, the type complex is not supported in C and VEE
> doesn't like to pass complex to a compiled function. Does anybody
> know how this could be done ?
>
> Also, the HP-VEE documentation is not very clear on the requirements
> for header files for FORTRAN compiled functions. Is there a better
> definition somewhere ? Or does anybody have good examples ?
>
> Regards
>
> Michel Gagne
> COM DEV Ltd
> Cambridge, Canada
> (519) 622-2300 x 575
>
> Mike.gagne@comdev.ca
>

---------------End of Original Message-----------------

Michel,
I usually do an Unbuild Complex and pass 2 arrays to C. When calling vee from C there are functions to extract arrays from a VEE complex that I
have been told will also work when calling C from vee but I have never tried it. Look at the function vdcGetComplex1DArray in veeData.h for more
info.

--------------------------------------------------------
Name: Tom Sanders
E-mail: Tom Sanders <toms@symca.com>
Date: 08/28/97 Time: 12:58:03
--------------------------------------------------------
Like • Show 0 Likes0
Actions
VRFuser Aug 27, 1997 10:55 PM
Mark Correct
Correct Answer
from: Greg Goebel / HP-MXD
      gvg@lvld.hp.com / 970-679-3030 / FAX 970-679-5971
      website: ftp://fcext3.external.hp.com/dist/mxd/index.html
to:   VRf-Michael Gagne
date: Thursday, 28 August 1997 1355 MDT

> Hi VRf listeners,
>
> I'm currently trying to pass an HP-VEE complex array to a compiled C
> function. However, the type complex is not supported in C and VEE
> doesn't like to pass complex to a compiled function. Does anybody
> know how this could be done ?

Sure, convert the complex values into a two-dimensional array.

> Also, the HP-VEE documentation is not very clear on the requirements
> for header files for FORTRAN compiled functions. Is there a better
> definition somewhere ? Or does anybody have good examples ?

Please see below.

>
> Regards
>
> Michel Gagne
> COM DEV Ltd
> Cambridge, Canada
> (519) 622-2300 x 575
>
> Mike.gagne@comdev.ca

--------------------------------- cut here ----------------------------------

[VEE] Building FORTRAN Functions For VEE

* It is possible to build FORTRAN functions for use with VEE, but you have to
configure them to resemble C functions as far as their interface to VEE is
concerned.

The following example is a FORTRAN function -- written into a file named
"add5.f" -- that accepts a floating-point array and an integer. The function
adds the value 5 to each element in the array; the integer represents the
number of elements.

   integer*4 function add5(wave, size)

   real*8 wave
   dimension wave(1:*)
   integer*4 size

   integer*4 i

   do 10 i = 1, size
      wave(i) = wave(i) + 5
10 continue

   add5 = size

   return
   end

The function prototype is very similar to the equivalent C function. We
declare the return type, followed by the function name and its parameters.
The parameter types are declared after the function declaration and before
the function body.

Notice the array declaration:

   real*8 wave
   dimension wave(1:*)

We declare a variable of type "real*8", which is a 64-bit floating-point
real, or the equivalent of a double. We then dimension the variable to be an
"assumed size" array, using the "(1:8)" syntax. In essence, this statement
says: "Give me an array of reals, but don't tell it how many elements it has
in it." FORTRAN, unless you tell it otherwise, assumes all array bounds
start at 1. By also passing in the size of the array, we can use the
variable named "size" to check the bounds on our array subscript.

FORTRAN assumes all parameters passed to a function are passed by reference.
This means that the address of the variable, as opposed to the value that the
variable holds, is given to the called function. This makes our coding
task easier, because we can declare variables without the need for pointer
dereferencing.

Here is the make file for the routine (named, say, "add5.mak"):

   add5.sl:   add5.o
              ld -b -o add5.sl add5.o
   add5.o:    add5.f
              f77 -c +z add5.f

You can then compile the function using the command: "make +f add5.mak" --
and then you get a library file named "add5.sl".

To actually run the function, you need to create the VEE definition file
(named, say, "add5.vh") to tell VEE parameters the function uses:

   long add5 (double *wave, long *size)

A VEE program to run the library is as follows:

                   +---------------------------------------+
                   |             Import Library            |
                   +---------------------------------------+
                   | Library Type    [ Compiled Function ] |
                   | Library Name    [        fl         ] |
                   | File Name       [      add5.sl      ] |
                   | Definition File [      add5.vh      ] |
                   +-------------------+-------------------+
+------------+                       |
|    Real    |      +----------------+-----------------+   +-----------+
+------------+      |           Call Function          |   |   Alpha   |
| 0000: 1    |      +------+---------------------------+   +-----------+
| 0000: 10   +----->| wave | Function Name | Ret Value |   | 0: 6      |
| 0000: 100 |      |      |               |   wave    +-->| 1: 15     |
| 0000: 1000 | +-->| size | [ fl.add5 ] |   size    |   | 2: 105    |
+------------+ |   +------+---------+-----+-----------+   | 3: 1O05   |
                  |                    |                     +-----------+
+------------+ |        +-----------+-----------+
| Integer   | |        |    Delete Library     |
+------------+ |        +-----------------------+
| 4          +--+        | Library Name [ fl ] |
+------------+           +-----------------------+

The library is loaded under the name "fl", which is arbitrary ... you should
set up the Import Library and then execute it before adding the other
objects, so when you add the Call Function, you can do a select function from
its object menu and install the function.

[<>]
Like • Show 0 Likes0
Actions

VRf - Compiled functions in C and F77

Outcomes

Related Content

Recommended Content