Parameter passing mechanisms in Fortran

Parameter passing mechamism in F90

Parameters in (non-recursive) functions in F90 are always passed by reference
To illustrate this fact, I will use C to illustrate this fact.

NOTE:

I do not C++ because C++ needs to modify the function names to avoid conflicts.
In C++, a function is identified by:
To avoid name collision at the assembler language level , C++ pre-pend and append unique strings to the function name

Name of a Fortran Function in Assembler code

Consider the following Fortran program:

PROGRAM main IMPLICIT NONE INTEGER i interface subroutine myprint(x) integer, intent(in) :: i end subroutine end interface i = 123 CALL myprint(i) END

Example Program: (Demo above code)
- Prog file: click here

You can discover how the Fortran compiler translate the program by using the -S option

Example:

f90 -S link-1-f90.f90

Result: (look for "CALL myprint(i)")

! 15 CALL myprint(i) call myprint_

In other words:

Name of a C function in Assembler code

Consider the following C program:

#include <stdio.h> void myprint(int *i) { printf("%d", *i); }

Example Program: (Demo above code)
- Prog file: click here

You can also discover how the C compiler translate the program by using the -S option

Example:

cc -S link-1-C.c

Result: (look for "myprint:")

.type myprint,#function myprint: save %sp,-96,%sp ! block 1 .L94: ! File link-1-C.c: ! 1 #include ! 2 ! 3 void myprint(int *i) ! 4 { ! 5 printf("%d", *i);

In other words:

From the above experiments, we can conclude that:

The F90 function call
will invoke the assembler function myprint_

The C function

void myprint_(....) { ... }

will translate into the assembler function myprint_

We have a way to invoke a C function from within a F90 program

NOTE: try:
See if you can figure out what name in the assembler code is mapped to "myprint()"...

Fortran program passing a parameter to a C function

Now we can demonstrate that Fortran passes parameters by reference

Example: Fortran calls C function

Fortran main program:

PROGRAM main IMPLICIT NONE INTEGER i interface subroutine myprint(x) integer, intent(in) :: i end subroutine end interface i = 123 CALL myprint(i) END

C function:

#include <stdio.h> void myprint_(int *i) { printf("i = %u", i); printf("*i = %d", *i); }

Example Program: (Demo above code)
- The Fortran main Prog file: click here
- The C function file: click here
Compile using:
Because the expression
obtains the parameter passed, Fortran must have passed the parameter by reference

C program passing parameter to a Fortran function

We can apply the above and arrive at the following programming rules for a C program to pass a parameter to a Fortran program

The C program must pass the address of a variable to a Fortran subroutine
A function name used in the function call in C must be
where functionName is the name of the subroutine in Fortran
(That is because the Fortran compiler f90 will translate functionName into assembler code as functionName_)

Example: C calls Fortran function

The C main program:

void myprint_(int *); int main(int argc, char argv[]) { int i = 1234; myprint_(&i); }

The Fortran subroutine:

SUBROUTINE myprint(n) IMPLICIT NONE INTEGER, INTENT(IN) :: n print n END SUBROUTINE

Example Program: (Demo above code)
- The C main program with "pass-by-reference": click here
- The Fortran myprint function: click here

The meainig of INTENT(IN)
- Recall that a variable passed by-reference can be modified by the callee function (See: click here )
- Since Fortran passes input variables with intent(in) using by reference , you may wonder if you can modified the variable in the function
  The answer is NO
- A variable specified as intent(in) , cannot be assigned a value
  Example:
- Example Program: (Demo above code)
  - Prog file: click here
  Compile with:
  You will get a compiler error...