struct myfloat
{
float float_tmp;
};
struct myint
{
int int_tmp;
};
int game_(myint1,myfloat2)
myint **myint;
myfloat **myfloat2;
{
return(0);
}
Could you please clarify what you want to archive with this C code snippet? Without further information it will be difficult to give you any useful answer.
1 Like
Given that isnât actually valid C code, and you arenât telling us what itâs for or what the goal is, this is close(ish) to the given example C code, and an example of calling it.
typedef struct mf {
float float_tmp;
} myfloat;
typedef struct mi {
int int_tmp;
} myint;
int game(myint x, myfloat y) {
return 42;
}
program main
use iso_c_binding, only: c_int, c_float
implicit none
type, bind(c) :: myfloat
real(c_float) :: float_tmp
end type
type, bind(c) :: myint
integer(c_int) :: int_tmp
end type
interface
function game(x, y) bind(c, name="game")
import :: myfloat, myint, c_int
implicit none
type(myint), intent(in), value :: x
type(myfloat), intent(in), value :: y
integer(c_int) :: game
end function
end interface
type(myfloat) :: y
type(myint) :: x
print *, game(x, y)
end program
1 Like
Then hire me to work on the project, send me an NDA, show me the actual code and tell me how youâd like to be able to interact with it. Because you still havenât provided valid C code (the compiler errors are shown below), or told us what you want to be able to do with it from Fortran.
$ gcc -c stuff.c -o stuff.o (master)
stuff.c: In function âgame_â:
stuff.c:12:1: error: unknown type name âmyintâ; use âstructâ keyword to refer to the type
12 | myint **myint1;
| ^~~~~
| struct
stuff.c:13:1: error: unknown type name âmyfloatâ; use âstructâ keyword to refer to the type
13 | myfloat **myfloat2;
| ^~~~~~~
| struct
2 Likes
use, intrinsic :: iso_c_binding, only: c_int, c_float, c_ptr
type, bind(c) :: myfloat
real(c_float) :: float_tmp
end type
type, bind(c) :: myint
integer(c_int) :: int_tmp
end type
interface
function game(myint1, myfloat2) bind(c, name="game_")
import :: c_int, c_ptr
implicit none
type(c_ptr) :: myint1
type(c_ptr) :: myfloat2
integer(c_int) :: game
end function
end interface
Thatâs the interface to that C code. Note that due to Fortranâs âpass by referenceâ, a pointer to the pointer myint1 is what will be passed to C (as you suspected).
2 Likes
Convince management to bring somebody on that knows what theyâre doing?
I donât mean to offend, but it would take me an hour to write a coherent and comprehensive description of why what youâve done is wrong, what the right way to do it is, and why youâre going to have problems if you donât do it the right way. And even with that, Iâm not convinced you wouldnât come back with several variations on the same question, effectively having this community trying to do your work for you without being compensated.
3 Likes
When you write an expression like cptr1+j-1 in fortran, what exactly do you think is the result? Do you think it is the same as the same expression in C where cptr1 is a C pointer?
Friendly reminder that everyone is welcome to post questions and ask for help on Fortran Discourse as long as theyâre not disrespectful or otherwise violating the CoC. Itâs perfectly OK to not respond to the topic if you donât want to help.
3 Likes
Itâs only free consultancy if you choose to respond. Please use the flag feature if you think any post is inappropriate or somehow abusing the Fortran Discourse and weâll review it. Otherwise, please ignore the posts if you donât like them.
1 Like
There are several problem, first integer*8 is not in the Fortran standard even though it is a common extension that means an integer variable stored in 8 bytes.
The compiler is telling you that you cannot pass an integer variable to an opaque type as type(c_ptr).
An eight byte integer variable, from the point of view of the Fortran processor, is different from an opaque type equivalent to a C pointer like type(c_ptr). Like in C a long is different from a void* even if they were of the same length in bytes.
Now in some old Fortran program integer*8 is used to store C pointers on the assumption that the compiler will support that extension, you are on a 64 bit machine, and you donât want the compiler to make any check.
ISO_C_BINDING is far better as it helps you to restrict the places where you have to manually be sure that what is written in C has a correct correspondence in Fortran (in the interfaces of the C routines), letting the compiler to help you in checking that you are calling the C routines correctly. Moreover it is standard and it will work in old 32 bit machines and future strange machines as well.
Personally I always use iso_c_binding, for example when I need to call a Fortran routine from python through the ctypes, the Fortran routine will appear as a C routine and everything is fine.
Basically you are saying that you have an array and you pass elements of this array to the routine âgameâ or that you get back an array of objects. The only meaning (in C) to add a number to an address is when you want to get the addresses of successive array elements.
If so, well, use an array.
C_F_POINTER has a shape argument, so you can access successive array elements (if allocated inside the C function), this is the only equivalent way to add a number to the âinteger*8â as far as I may understand.
Or use an array of type(c_ptr), and pass each element of the array to the âgameâ routine, if that is what you want to do instead.
Look also at C_ASSOCIATED if it may help you.
As @milancurcic said, please do not respond to a post if you think it is not worth your time.
If you ask a question here, please be polite and respectful to people that try to respond to it. Also try to provide as many details as you can, the background and motivation of the question, to increase chances of getting good actionable answers.
1 Like