Fortran C interoperability: Accessing constant variables (parameter) from modules

[Petros]

Hi everybody,

I've been using the Fortran-C interoperability features for some time now but always for accessing variables. I cannot figure out how to access a constant from inside a module. I have the following example:

module testing
   integer, parameter ::  const_int=160
   integer :: var_int
end module testing

I compile with:

ifort -traceback -fPIC -module obj/Release/ -c testing_mod.f90 -o obj/Release/testing_mod.o

If I execute objdump -x testing_mod.o, I get:

SYMBOL TABLE:
0000000000000000 l    df *ABS*  0000000000000000 testing_mod.f90
0000000000000000 l    d  .data  0000000000000000 .data
0000000000000000 l    d  .bss   0000000000000000 .bss
0000000000000000 l    d  .text  0000000000000000 .text
0000000000000000 l    d  .note.GNU-stack        0000000000000000 .note.GNU-stack
0000000000000000 l    d  .comment       0000000000000000 .comment
0000000000000000 l    d  .eh_frame      0000000000000000 .eh_frame
0000000000000000 l    d  .trace 0000000000000000 .trace
0000000000000000 g     F .text  0000000000000010 testing._
0000000000000004       O *COM*  0000000000000008 testing_mp_var_int_

I cannot see the const_int and I cannot access it from C. I can bind the variable (var_int) and select the name but when I try to bind the constant by adding bind(C, name="const_int"), I get:

testing_mod.f90(3): error #6406: Conflicting attributes or multiple declaration of name.   [CONST_INT]
   integer, parameter, bind(C, name="const_int") ::  const_int=160
-----------------------------------------------------^
compilation aborted for testing_mod.f90 (code 1)

Any idea on how to access it? I work under Debian Linux 7 and my compiler is: Version 14.0.2.144 Build 20140120

Thanks in advance,

Petros

[Petros]

I found my answer... Really sorry for the spamming. I was searching with wrong keywords. --> http://software.intel.com/en-us/forums/topic/291657

Found a workaround:

module testing
   use iso_c_binding

   integer, parameter ::  const_int=160
   integer, protected, bind(C, name="const_int") ::  const_int_c
   data const_int_c/const_int/

   integer :: var_int
end module testing

Seems to work fine. Thanks again!

[FortranFan]

p3tris wrote:

I found my answer... Really sorry for the spamming. I was searching with wrong keywords. --> http://software.intel.com/en-us/forums/topic/291657

Found a workaround:

module testing
   use iso_c_binding

   integer, parameter ::  const_int=160
   integer, protected, bind(C, name="const_int") ::  const_int_c
   data const_int_c/const_int/

   integer :: var_int
end module testing

Seems to work fine. Thanks again!

Why can't you simply add an accessor method in your module instead?

[Petros]

FortranFan wrote:

Quote:

p3tris wrote:

I found my answer... Really sorry for the spamming. I was searching with wrong keywords. --> http://software.intel.com/en-us/forums/topic/291657

Found a workaround:

module testing
   use iso_c_binding

   integer, parameter ::  const_int=160
   integer, protected, bind(C, name="const_int") ::  const_int_c
   data const_int_c/const_int/

   integer :: var_int
end module testing

Seems to work fine. Thanks again!

Why can't you simply add an accessor method in your module instead?

Yes, that was my other option. But as I have more than one contstants to access, I prefered this one. I guess both work fine.

[FortranFan]

p3tris wrote:

Quote:

FortranFan wrote:

Why can't you simply add an accessor method in your module instead?

Yes, that was my other option. But as I have more than one contstants to access, I prefered this one. I guess both work fine.

Re: "both work fine" - some comments:

• As one's program gets bigger and bigger and more coders work on it, I usually find such direct access to data is fraught with danger, especially in mixed-language environments.
• The accessor method provides good data encapsulation methodology which is one of the benefits of using modules.
• I'm not a fan of DATA statement and have "deprecated" its use in my code.  It can interfere with thread safety and hinder parallelization efforts

My 2 cents,

[Petros]

FortranFan wrote:

Quote:

p3tris wrote:

Quote:

FortranFan wrote:

Why can't you simply add an accessor method in your module instead?

Yes, that was my other option. But as I have more than one contstants to access, I prefered this one. I guess both work fine.

Re: "both work fine" - some comments:

• As one's program gets bigger and bigger and more coders work on it, I usually find such direct access to data is fraught with danger, especially in mixed-language environments.
• The accessor method provides good data encapsulation methodology which is one of the benefits of using modules.
• I'm not a fan of DATA statement and have "deprecated" its use in my code.  It can interfere with thread safety and hinder parallelization efforts

My 2 cents,

Thanks! I think I'll create some accessors as you suggest. However, the DATA statement, I use it in other situations not relating to the C interface. Could you give me some source of information about its interference with parallelization? Or is it a personal observation?

[TimP]

DATA implies SAVE so you would need to jump through more hoops to achieve thread safety.