PGI Inter-languauge Questions
- How do you call pgf77-pgf90-pghpf routines from g++?
- what arguments can be passed from C++ to f90?
- How do you pass data between f77 and C/C++ routines.?
- How do you pass string data from HPF to an external C function?
- How do you pass data between f77 and f90?
- How about an example of hpf calling every type language routine?
- Visual C++ Interfacing
How do you call pgf77-pgf90-pghpf routines from g++?
The answer is straightforward but detailed.
- If you use pghpf, we recommend that the main routine be hpf.
- If you use pgf90 or pghpf, you need to include a call to pghpf_init.
-
If you are using a C++ main routine, be sure to save the argv, argc pointers so they are accessible to the PGI compiled routines. Also, the link line must contain the following fortran libraries:
pgf90: -lpgf90 -lpgf90_rpm1 -lpgf902 -lpgf90rtl -lpgftnrtl -lm -lpgc -lgcc pghpf: -lpghpf -lpghpf_rpm -lpghpf2 -lpgftnrtl -lm -lpgc -lgcc -lc -lgcc
main.c:
extern "C" void foo_ ( int * );
extern "C" void pghpf_init(int *);
int __argc_save;
char **__argv_save;
static int zz = 0;
main(int argc, char **argv)
{
int i;
__argc_save = argc;
__argv_save = argv;
pghpf_init(&zz);
i = 19;
foo_(&i);
}
foo.f90:
subroutine foo (i)
integer i
print *, i
end subroutine foo
pgf90 -c foo.f90
g++ main.c foo.o -L/usr/pgi/linux86/lib
-I/usr/pgi/linux86/include -lpgf90 -lpgf90_rpm1
-lpgf902 -lpgf90rtl -lpgftnrtl -lm -lpgc -lgcc
(assumes compilers installed in /usr/pgi)
In the case of code compiled with pgf77, the similar example is as follows:
main.c:
extern "C" void foo_ ( int * );
int __argc_save;
char **__argv_save;
main(int argc, char **argv)
{
int i;
__argc_save = argc;
__argv_save = argv;
i = 19;
foo_(&i);
}
foo.f:
subroutine foo (i)
integer i
print *, i
end
pgf77 -c foo.f
g++ main.c foo.o -L/usr/pgi/linux86/lib
-I/usr/pgi/linux86/include -lpgftnrtl -lpgc
The follow up question is
What arguments can be passed from C++ to f90?
Only the f77-style of passing arrays to an f90 routine from c++ will work. E.g., the dummy arrays of the f90 routine called from c/c++/f77 cannot be assumed shape or f90 pointer.
Example:
C/C++
float a[100];
int n = 100;
foo_(a,&n);
...
F90
subroutine foo(a,n)
real a(:) ! does not work - a cannot be assumed shape
...
subroutine foo(a,n)
real a(n) ! should work
What about passing command line arguments (argc, argv) from gcc programs to pgf77 compiled code?
Again remember that pgi compilers use different globals than gcc does.
% more cmain.c
int __argc_save;
char **__argv_save;
main(int argc, char **argv)
{
char bool1, letter1;
int numint1, numint2;
float numfloat1;
double numdoub1;
short numshor1;
extern void forts_ ();
int i;
__argc_save = argc;
__argv_save = argv;
for (i=0; i < argc; i++){
printf("main: command line arg %d is %s\n",i,argv[i]);
}
forts_(&bool1,&letter1,&numint1,&numint2,&numfloat1,
&numdoub1,&numshor1, 1);
printf("main: %s %c %d %d %3.1f %.0f %d\n",
bool1?"TRUE":"FALSE",letter1,numint1,
numint2, numfloat1, numdoub1, numshor1);
}
% more forts.f
subroutine forts ( bool1, letter1, numint1,
& numint2, numfloat1, numdoub1, numshor1)
character*(128) arg
logical*1 bool1
character letter1
integer numint1, numint2
double precision numdoub1
real numfloat1
integer*2 numshor1
integer i,j
bool1 = .true.
letter1 = "v"
numint1 = 11
numint2 = -44
numdoub1 = 902
numfloat1 = 39.6
numshor1 = 299
i = iargc()
print *,"forts: argc=",i
do j=0,i
call getarg(j,arg)
print *,"forts: arg ",j," is ", arg
end do
return
end
% gcc -c cmain.c
% pgf77 -o c2f -Mnomain cmain.o forts.f
Linking:
% c2f 123 456 momma
main: command line arg 0 is c2f
main: command line arg 1 is 123
main: command line arg 2 is 456
main: command line arg 3 is momma
forts: argc= 3
forts: arg 0 is c2f
forts: arg 1 is 123
forts: arg 2 is 456
forts: arg 3 is momma
main: TRUE v 11 -44 39.6 902 299
How do you pass data between f77 and C/C++ routines.?
Here are some examples. See Chapter 10 of the User's Guide for more information.
fortran calling C
% more fmain.f
program fort2c
logical*1 bool1
character letter1
integer*4 numint1, numint2
real numfloat1
double precision numdoub1
integer*2 numshor1
external cfunc
call cfunc (bool1, letter1, numint1, numint2,
& numfloat1, numdoub1, numshor1)
write( *, 100)
& bool1, letter1, numint1, numint2, numfloat1,
& numdoub1, numshor1
100 format(1x,"bool1 = ", L2,/,
& " letter1 = ", A2,/,
& " numint1 = ", I5,/,
& " numint2 = ", I5,/,
& " numfloat1 = ", F6.1,/,
& " numdoub1 = ", F6.1,/,
& " numshor1 = ", I5,/)
end
% more csub.c
#define TRUE 0xff
#define FALSE 0
void
cfunc_( bool1, letter1, numint1, numint2, numfloat1,\
numdoub1, numshor1, len_letter1)
char *bool1, *letter1;
int *numint1, *numint2;
float *numfloat1;
double *numdoub1;
short *numshor1;
int len_letter1;
{
*bool1 = TRUE;
*letter1 = 'v';
*numint1 = 11;
*numint2 = -44;
*numfloat1 = 39.6 ;
*numdoub1 = 39.2 ;
*numshor1 = 981;
}
%pgcc -c csub.c
%pgf77 fmain.f csub.o
% a.out
bool1 = T
letter1 = v
numint1 = 11
numint2 = -44
numfloat1 = 39.6
numdoub1 = 39.2
numshor1 = 981
C calling fortran
% more cmain.c
main ()
{
char bool1, letter1;
int numint1, numint2;
float numfloat1;
double numdoub1;
short numshor1;
extern void forts_ ();
forts_(&bool1,&letter1,&numint1,&numint2,&numfloat1,
&numdoub1,&numshor1, 1);
printf(" bool1 = %s\n letter1 = %c\n numint1 = %d \n numint2 = %d \n numfloat1 = %3.1f \n
numdoub1 = %.0f \n numshor1 = %d\n",
bool1?"TRUE":"FALSE",letter1,numint1,
numint2, numfloat1, numdoub1, numshor1);
}
% more fsub.f
subroutine forts ( bool1, letter1, numint1,
& numint2, numfloat1, numdoub1, numshor1)
logical*1 bool1
character letter1
integer numint1, numint2
double precision numdoub1
real numfloat1
integer*2 numshor1
bool1 = .true.
letter1 = "v"
numint1 = 11
numint2 = -44
numdoub1 = 902
numfloat1 = 39.6
numshor1 = 299
return
end
% pgcc -c cmain.c
% pgf77 -Mnomain cmain.o fsub.f
Linking:
% a.out
bool1 = TRUE
letter1 = v
numint1 = 11
numint2 = -44
numfloat1 = 39.6
numdoub1 = 902
numshor1 = 299
C++ calling fortran
% more cpmain.C
#include <iostream>
extern "C" {
extern void forts_ ( char *, char *, int *, int *, float *, double *, short * );
}
main ()
{
char bool1, letter1;
int numint1, numint2;
float numfloat1;
double numdoub1;
short numshor1;
forts_(&bool1, &letter1, &numint1, &numint2, &numfloat1, &numdoub1, &numshor1);
cout << " bool1 = ";
bool1?cout << "TRUE ":cout << "FALSE "; cout << endl;
cout << " letter1 = " << letter1 << endl;
cout << " numint1 = " << numint1 << endl;
cout << " numint2 = " << numint2 << endl;
cout << " numfloat1 = " << numfloat1 << endl;
cout << " numdoub1 = " << numdoub1 << endl;
cout << " numshor1 = " << numshor1 << endl ;
}
% more fsub.f
subroutine forts ( bool1, letter1, numint1,
& numint2, numfloat1, numdoub1, numshor1)
logical*1 bool1
character letter1
integer numint1, numint2
double precision numdoub1
real numfloat1
integer*2 numshor1
bool1 = .true.
letter1 = "v"
numint1 = 11
numint2 = -44
numdoub1 = 902
numfloat1 = 39.6
numshor1 = 299
return
% pgf77 -c fsub.f
% pgCC cpmain.C fsub.o
cpmain.C:
Linking:
% a.out
bool1 = TRUE
letter1 = v
numint1 = 11
numint2 = -44
numfloat1 = 39.6
numdoub1 = 902
numshor1 = 299
f77 calling C++
% more fmain.f
logical*1 bool1
character letter1
integer*4 numint1, numint2
real numfloat1
double precision numdoub1
integer*2 numshor1
external cfunc
call cplus_func (bool1, letter1, numint1,
& numint2, numfloat1, numdoub1, numshor1)
write( *, 100)
& bool1, letter1, numint1, numint2, numfloat1,
& numdoub1, numshor1
100 format(1x,"bool1 = ", L2,/,
& " letter1 = ", A2,/,
& " numint1 = ", I5,/,
& " numint2 = ", I5,/,
& " numfloat1 = ", F6.1,/,
& " numdoub1 = ", F6.1,/,
& " numshor1 = ", I5,/)
end
% more cpsub.C
#define TRUE 0xff
#define FALSE 0
extern "C" {
extern void cplus_func_ (
char *bool1,
char *letter1,
int *numint1,
int *numint2,
float *numfloat1,
double *numdoub1,
short *numshort1,
int len_letter1) {
*bool1= TRUE;
*letter1 = 'v';
*numint1 = 11;
*numint2 = -44;
*numfloat1 = 39.6;
*numdoub1 = 39.2;
*numshort1 = 981;
}
}
% pgCC -c cpsub.C
cpsub.C:
% pgf77 fmain.f cpsub.o
Linking:
% a.out
bool1 = T
letter1 = v
numint1 = 11
numint2 = -44
numfloat1 = 39.6
numdoub1 = 39.2
numshor1 = 981
C calling C++
% more cmain.c
extern void func2_cplus(int a, int b, int *c);
#include <stdio.h>
main()
{
int a,b,c;
a=8;
b=2;
printf("main: a = %d b = %d ptr c = %x\n",a,b,&c);
func2_cplus(a,b,&c);
printf("main: res = %d\n",c);
}
% more cpsub.C
#include <iostream>
extern "C" void func2_cplus(int num1,int num2,int *res)
{
cout << "sub: a = " << num1 << " b = " << num2 << " ptr c = " << res << endl << endl;
*res=num1/num2;
cout << "sub: res = " << *res << endl ;
}
% pgcc -c cmain.c
% pgCC cmain.o cpsub.C
cpsub.C:
Linking:
% a.out
main: a = 8 b = 2 ptr c = bffffca4
sub: a = 8 b = 2 ptr c = 0xbffffca4
sub: res = 4
main: res = 4
C++ calling C
% more cpmain.C
extern "C" void func1_cplus(int n, int m, int *p);
#include <iostream>
main()
{
int a,b,c;
a=8;
b=2;
cout << "main: a = " << a << " b = " << b << " ptr c = " << &c << endl;
func1_cplus(a,b,&c);
cout << "main: res = " << c << endl;
}
% more csub.c
#include <stdio.h>
void func1_cplus(num1, num2, res)
int num1, num2, *res;
{
printf("sub: a = %d b = %d ptr c = %x\n",num1,num2,res);
*res=num1/num2;
printf("\nsub: res = %d\n",*res);
}
% pgcc -c csub.c
% pgCC cpmain.C csub.o
cpmain.C:
Linking:
% a.out
main: a = 8 b = 2 ptr c = 0xbffffca4
sub: a = 8 b = 2 ptr c = bffffca4
sub: res = 4
main: res = 4
How do you pass string data from HPF to an external C function?
In summary, you need a function prototype in your fortran program if the C function returns a value. Next you need to explicitly pass the length of the string to the C function. In the example below we pass in the string followed by its length using the len_trim intrinsic. By default values are passed by reference, so the C function receives the addresses of the actual arguments...
First the HPF:
program p
integer*4 call_c ! C function prototype
character(len=128) :: my_string
my_string = 'this is a test 1234567890'
ierr=call_c(my_string,len_trim(my_string))
print *, ierr
end
Here's the C:
int call_c_(char *string, unsigned *len){
char * c_string;
c_string = (char*) malloc(sizeof(char)*((*len)+1));
memcpy((void*) c_string, (const void*) string, *len);
c_string[*len] = '\0';
printf("string = %s\n",c_string);
free(c_string);
return 0;
}
How do you pass data between f77 and f90?
f77 calling f90
% more fmain.f
program fmain
integer i
i= 19
call f90sub(i)
end
% more f90sub.f90
subroutine f90sub (i)
include 'lib3f.h'
character(32) argstring
integer i
print *, i
print *, "iargc = ", iargc()
do j=0,iargc()
call getarg(j,argstring)
print *,"arg",j," is ",argstring
end do
end subroutine f90sub
% pgf77 -c fmain.f
% pgf90 -o f2f90 f90sub.f90 fmain.o
f90sub.f90:
Linking:
% f2f90 is a good call 123
19
iargc = 5
arg 0 is f2f90
arg 1 is is
arg 2 is a
arg 3 is good
arg 4 is call
arg 5 is 123
f90 calling f77
% more f90main.f90
program f90main
interface
extrinsic (f77_local) subroutine f77sub(i)
integer , intent(in) :: i
end subroutine f77sub
end interface
integer i
i= 19
call f77sub(i)
end
% more f77sub.f
subroutine f77sub (i)
character*32 argstring
integer i
print *, i
print *, "iargc = ", iargc()
do j=0,iargc()
call getarg(j,argstring)
print *,"arg",j," is ",argstring
end do
end
% pgf77 -c f77sub.f
% pgf90 -o f902f f90main.f90 f77sub.o
f90main.f90:
Linking:
% f902f is a good call also 123
19
iargc = 6
arg 0 is f902f
arg 1 is is
arg 2 is a
arg 3 is good
arg 4 is call
arg 5 is also
arg 6 is 123
How about an example of hpf calling every type language routine?
Here is an example of an hpf program calling a c, f77, f90, hpf, and C++, program, and even a subroutine compiled.
pghpf -Mf90 -c
% more hpfmain.f
program hpf2all
logical*1 bool1
character letter1
integer*4 numint1, numint2
real numfloat1
double precision numdoub1
integer*2 numshor1
interface
extrinsic (f77_local) subroutine csub(
& bool1, letter1, numint1, numint2,
& numfloat1, numdoub1, numshor1)
logical*1, intent(out) :: bool1
character, intent(out) :: letter1
integer*4, intent(out) :: numint1, numint2
real, intent(out) :: numfloat1
double precision, intent(out) :: numdoub1
integer*2, intent(out) :: numshor1
end subroutine csub
end interface
interface
extrinsic (f77_local) subroutine f77sub(
& bool1, letter1, numint1, numint2,
& numfloat1, numdoub1, numshor1)
logical*1, intent(out) :: bool1
character, intent(out) :: letter1
integer*4, intent(out) :: numint1, numint2
real, intent(out) :: numfloat1
double precision, intent(out) :: numdoub1
integer*2, intent(out) :: numshor1
end subroutine f77sub
end interface
interface
subroutine hpfsub(
& bool1, letter1, numint1, numint2,
& numfloat1, numdoub1, numshor1)
logical*1, intent(out) :: bool1
character, intent(out) :: letter1
integer*4, intent(out) :: numint1, numint2
real, intent(out) :: numfloat1
double precision, intent(out) :: numdoub1
integer*2, intent(out) :: numshor1
end subroutine hpfsub
end interface
interface
extrinsic (f90_local) subroutine hpf90sub(
& bool1, letter1, numint1, numint2,
& numfloat1, numdoub1, numshor1)
logical*1, intent(out) :: bool1
character, intent(out) :: letter1
integer*4, intent(out) :: numint1, numint2
real, intent(out) :: numfloat1
double precision, intent(out) :: numdoub1
integer*2, intent(out) :: numshor1
end subroutine hpf90sub
end interface
interface
extrinsic (f90_local) subroutine f90sub(
& bool1, letter1, numint1, numint2,
& numfloat1, numdoub1, numshor1)
logical*1, intent(out) :: bool1
character, intent(out) :: letter1
integer*4, intent(out) :: numint1, numint2
real, intent(out) :: numfloat1
double precision, intent(out) :: numdoub1
integer*2, intent(out) :: numshor1
end subroutine f90sub
end interface
interface
extrinsic (f77_local) subroutine cplus(
& bool1, letter1, numint1, numint2,
& numfloat1, numdoub1, numshor1)
logical*1, intent(out) :: bool1
character, intent(out) :: letter1
integer*4, intent(out) :: numint1, numint2
real, intent(out) :: numfloat1
double precision, intent(out) :: numdoub1
integer*2, intent(out) :: numshor1
end subroutine cplus
end interface
call csub (bool1, letter1, numint1, numint2,
& numfloat1, numdoub1, numshor1)
write( *, 100)
& bool1, letter1, numint1, numint2, numfloat1,
& numdoub1, numshor1
100 format(1x,"csub: ",/,
& " bool1 = ", L2,/,
& " letter1 = ", A2,/,
& " numint1 = ", I5,/,
& " numint2 = ", I5,/,
& " numfloat1 = ", F9.1,/,
& " numdoub1 = ", F9.1,/,
& " numshor1 = ", I5,/)
call f77sub (bool1, letter1, numint1, numint2,
& numfloat1, numdoub1, numshor1)
write( *, 200)
& bool1, letter1, numint1, numint2, numfloat1,
& numdoub1, numshor1
200 format(1x,"f77sub: ",/,
& " bool1 = ", L2,/,
& " letter1 = ", A2,/,
& " numint1 = ", I5,/,
& " numint2 = ", I5,/,
& " numfloat1 = ", F9.1,/,
& " numdoub1 = ", F9.1,/,
& " numshor1 = ", I5,/)
call hpfsub (bool1, letter1, numint1, numint2,
& numfloat1, numdoub1, numshor1)
write( *, 400)
& bool1, letter1, numint1, numint2, numfloat1,
& numdoub1, numshor1
400 format(1x,"hpfsub: ",/,
& " bool1 = ", L2,/,
& " letter1 = ", A2,/,
& " numint1 = ", I5,/,
& " numint2 = ", I5,/,
& " numfloat1 = ", F9.1,/,
& " numdoub1 = ", F9.1,/,
& " numshor1 = ", I5,/)
call hpf90sub (bool1, letter1, numint1, numint2,
& numfloat1, numdoub1, numshor1)
write( *, 500)
& bool1, letter1, numint1, numint2, numfloat1,
& numdoub1, numshor1
500 format(1x,"hpf90sub: ",/,
& " bool1 = ", L2,/,
& " letter1 = ", A2,/,
& " numint1 = ", I5,/,
& " numint2 = ", I5,/,
& " numfloat1 = ", F9.1,/,
& " numdoub1 = ", F9.1,/,
& " numshor1 = ", I5,/)
call f90sub (bool1, letter1, numint1, numint2,
& numfloat1, numdoub1, numshor1)
write( *, 600)
& bool1, letter1, numint1, numint2, numfloat1,
& numdoub1, numshor1
600 format(1x,"f90sub: ",/,
& " bool1 = ", L2,/,
& " letter1 = ", A2,/,
& " numint1 = ", I5,/,
& " numint2 = ", I5,/,
& " numfloat1 = ", F9.1,/,
& " numdoub1 = ", F9.1,/,
& " numshor1 = ", I5,/)
call cplus (bool1, letter1, numint1, numint2,
& numfloat1, numdoub1, numshor1)
write( *, 700)
& bool1, letter1, numint1, numint2, numfloat1,
& numdoub1, numshor1
700 format(1x,"cplus: ",/,
& " bool1 = ", L2,/,
& " letter1 = ", A2,/,
& " numint1 = ", I5,/,
& " numint2 = ", I5,/,
& " numfloat1 = ", F9.1,/,
& " numdoub1 = ", F9.1,/,
& " numshor1 = ", I5,/)
end program hpf2all
% more csub.c
#define TRUE 0xff
#define FALSE 0
void
csub_( bool1, letter1, numint1, numint2, numfloat1,\
numdoub1, numshor1, len_letter1)
char *bool1, *letter1;
int *numint1, *numint2;
float *numfloat1;
double *numdoub1;
short *numshor1;
int len_letter1;
{
*bool1 = TRUE;
*letter1 = 'a';
*numint1 = 11;
*numint2 = -11;
*numfloat1 =10123.4;
*numdoub1 = 20123.4;
*numshor1 = 111;
}
% more f77sub.f
subroutine f77sub ( bool1, letter1, numint1,
& numint2, numfloat1, numdoub1, numshor1)
logical*1 bool1
character letter1
integer numint1, numint2
double precision numdoub1
real numfloat1
integer*2 numshor1
bool1 = .false.
letter1 = "b"
numint1 = 22
numint2 = -22
numfloat1 = 30123.4
numdoub1 = 40123.4
numshor1 = 222
return
end
% more hpfsub.f
subroutine hpfsub ( bool1, letter1, numint1,
& numint2, numfloat1, numdoub1, numshor1)
logical*1 bool1
character letter1
integer numint1, numint2
double precision numdoub1
real numfloat1
integer*2 numshor1
bool1 = .true.
letter1 = "c"
numint1 = 33
numint2 = -33
numfloat1 = 50123.4
numdoub1 = 60123.4
numshor1 = 333
return
end subroutine hpfsub
% more hpf90sub.f
subroutine hpf90sub ( bool1, letter1, numint1,
& numint2, numfloat1, numdoub1, numshor1)
logical*1 bool1
character letter1
integer numint1, numint2
double precision numdoub1
real numfloat1
integer*2 numshor1
bool1 = .false.
letter1 = "d"
numint1 = 44
numint2 = -44
numfloat1 = 70123.4
numdoub1 = 80123.4
numshor1 = 444
return
end subroutine hpf90sub
% more f90sub.f
subroutine f90sub ( bool1, letter1, numint1,
& numint2, numfloat1, numdoub1, numshor1)
logical*1 bool1
character letter1
integer numint1, numint2
double precision numdoub1
real numfloat1
integer*2 numshor1
bool1 = .true.
letter1 = "e"
numint1 = 55
numint2 = -55
numfloat1 = 90123.4
numdoub1 = 100123.4
numshor1 = 555
return
end subroutine f90sub
% more cplus.C
#define TRUE 0xff
#define FALSE 0
extern "C" {
extern void cplus_ (
char *bool1,
char *letter1,
int *numint1,
int *numint2,
float *numfloat1,
double *numdoub1,
short *numshort1,
int len_letter1) {
*bool1= FALSE;
*letter1 = 'f';
*numint1 = 66;
*numint2 = -66;
*numfloat1 = 110123.4;
*numdoub1 = 120123.4;
*numshort1 = 666;
}
}
% pgcc -c csub.c
% pgf77 -c f77sub.f
% pghpf -c hpfsub.f
% pghpf -c -Mf90 hpf90sub.f
% pgf90 -c -Mx,50,2 f90sub.f
% pgCC -c cplus.C
% pghpf -o hpf2all hpfmain.f csub.o f77sub.o hpfsub.o hpf90sub.o f90sub.o cplus.o
% hpf2all
csub:
bool1 = T
letter1 = a
numint1 = 11
numint2 = -11
numfloat1 = 10123.4
numdoub1 = 20123.4
numshor1 = 111
f77sub:
bool1 = F
letter1 = b
numint1 = 22
numint2 = -22
numfloat1 = 30123.4
numdoub1 = 40123.4
numshor1 = 222
hpfsub:
bool1 = T
letter1 = c
numint1 = 33
numint2 = -33
numfloat1 = 50123.4
numdoub1 = 60123.4
numshor1 = 333
hpf90sub:
bool1 = F
letter1 = d
numint1 = 44
numint2 = -44
numfloat1 = 70123.4
numdoub1 = 80123.4
numshor1 = 444
f90sub:
bool1 = T
letter1 = e
numint1 = 55
numint2 = -55
numfloat1 = 90123.4
numdoub1 = 100123.4
numshor1 = 555
cplus:
bool1 = F
letter1 = f
numint1 = 66
numint2 = -66
numfloat1 = 110123.4
numdoub1 = 120123.4
numshor1 = 666
Visual C++ Interfacing
How do you build and link pgf77-pgf90 routines from Visual C++?
The answer is pretty straight forward. You link the same f77/f90 libraries you would link under pgCC, using the Visual C++ linker.
To build an executable using the following source on a linux86 and NT system using pgCC or VC++ and pgf77 or pgf90.
c.C
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#ifdef __PGI
extern "C" {
void f_(float *, char *, int *, int *, int);
void c2_(int *, float *);
}
#else
void f_(float *, char *, int *, int *, int);
void c2_(int *, float *);
#endif
void main (void) {
int i, nchar;
char b[80];
char *c;
float a[3];
i = 3;
a[0] = 10.0;
a[1] = 20.0;
a[2] = 30.0;
sprintf(b, "a.dat");
nchar = strlen(b);
printf("string length = %d\n", nchar);
c = (char *) malloc (nchar*sizeof(char));
strncpy(c,b,nchar);
f_(a, c, &i, &nchar, nchar);
}
void c2_(int *b, float *a){
printf("CALLBACK %d %f %f %f\n",*b, a[0], a[1], a[2]);
}
The following fortran file was used
f.f
subroutine f(a,b,i,j)
character *(*) b
dimension a(*)
print*, i, a(1), a(2), a(3)
print*, j
print*, "String in F77=",b(1:j)
open(11,file=b(1:j))
write(11,*) "writing in f77"
close(11)
call c2(i,a)
return
end
The program 'c' was built and executed on a linux x86 system as follows
% pgf77/f90 -c f.f
% pgCC -o c c.C f.o -lpgftnrtl -lpgc
% pgCC -o c90 c.C
f.o -lpgf90 -lpgf90_rpm1 -lpgf902 -lpgf90rtl -lpgftnrtl -lm -lpgc
% c
string length = 5
3 10.00000 20.00000 30.00000
5
String in F77=a.dat
CALLBACK 3 10.000000 20.000000 30.000000
On an NT system, under PGI Workstation, create the following object files using pgf77/f90.
PGI> pgf77 -c -Munix f.f PGI> pgf90 -o f90.o -c -Munix f.f
Under VC++ The file c.C opened as a C++ source file. It was compiled w/o errors.
The object file c.obj was copied to the pgi working area my_dir milesto c.o.
Linking on NT using pgf90/pgf77 (not necessary to recompile, but okay to)
PGI> pgf90 -o c90 c.o f.f -Munix -Mnomain -msvcrt -mslibs PGI> pgf77 -o c c.o f.f -Munix -Mnomain -msvcrt -mslibs
Linking on NT using VC++
Under Project>Settings
In the Link page, add the following objects and libraries to the 'Object/library modules' window, at the beginning:
for f.o
C:\PGI\nt86\lib\pgstdinit-mslibs.o my_dir\f.o C:\PGI\nt86\lib\libpgftnrtl.a C:\PGI\nt86\lib\libpgc.a libc.lib oldnames.lib
for f90.o
C:\PGI\nt86\lib\pgstdinit-mslibs.o my_dir\f90.o C:\PGI\nt86\lib\libpgf90.a C:\PGI\nt86\lib\libpgf90_rpm1.a C:\PGI\nt86\lib\libpgf902.a C:\PGI\nt86\lib\libpgf90rtl.a C:\PGI\nt86\lib\libpgftnrtl.a C:\PGI\nt86\lib\libpgc.a libc.lib oldnames.lib
set the flag 'Ignore all default libraries'
The executables run as on linux86
How do you build and link pgf77-pgf90 OpenMP routines from Visual C++?
The big question is not whether you can or not, but whether the resulting application is truly running parallel. This example assumes a version of NT (service pack 4.0 or greater?) that supports SMP.
Here is a simple program that wastes cycles but scales well. It times the process so that a successful execution on two processors will take half the time of one processor.
Here is a timer we use, dclock. It is written in assembler.
dclock.s
.file "dclock-pentium.s"
.align 8
.data
# .clock: .double 0.0000000075 # 133MHz
# .clock: .double 0.00000000666667 # 150 MHz
# .clock: .double 0.0000000060 # 166 MHz
.clock: .double 0.0000000050 # 200 MHz
# .clock: .double 0d4.28571429183673480000e-09 # 233.3333 MHz
# .clock: .double 0.00000000333333 # 300 MHz
# .clock: .double 0.00000000222223 # 450 MHz
# .clock: .double 0.000000002 # 500 MHz
# .clock: .double 0.0000000018181818 # 550 MHz
# .clock: .double 0.0000000016666667 # 600 MHz
# .clock: .double 0.0000000015384615 # 650 MHz
# .clock: .double 0.0000000014285714 # 700 MHz
# .clock: .double 0.0000000013658095 # 733 MHz
# .clock: .double 0.0000000013333333 # 750 MHz
# .clock: .double 0.0000000012500000 # 800 MHz
# .clock: .double 0.000000001 # 1.0 GHz
# .clock: .double 0.000000000666 # 1.5 GHz
# .clock: .double 0.00000000059 # 1.7 GHz
# .clock: .double 0.0000000005 # 2.0 GHz
.low: .long 0x00000000
.high: .long 0x00000000
.text
.align 4
.globl _DCLOCK@0, dclock, _dclock, _dclock_, dclock_
_DCLOCK@0:
dclock:
_dclock:
_dclock_:
dclock_:
.byte 0x0f, 0x31
movl %eax, .low
movl %edx, .high
fildll .low
fmull .clock
ret
it is set for a 200 MHz cpu, so use the .clock val that applies to your machine.
Here is a c main program that you will compile with VC++
ex4main.c
#include <stdio.h>
#include <stdlib.h>
void ex4_(int *);
void main(int argc, char** argv)
{
int i;
if(argc == 1) exit(0);
i = atoi(argv[1]);
printf( "(main) calling ex4 with i = %d \n", i );
ex4_(&i);
}
Here is a fortran OpenMP routine that will use 2 cpus if you have them. Compile it.
pgf77 -mp -c -Munix ex4,f
ex4.f
subroutine ex4(i)
integer i,k
integer omp_get_num_procs, omp_get_max_threads
real*8 dclock, time1, time2
print *,"num procs =", omp_get_num_procs()
print *,"max threads =", omp_get_max_threads()
call omp_set_num_threads(1)
time1 = dclock()
!$omp parallel
!$omp do
do k = 1, 2
call delay(i)
enddo
!$omp end parallel
time2 = dclock() - time1
print *, " 1 cpu test - delay value doubled =", i*2,
* " time =", time2, " seconds"
if (omp_get_num_procs() .ge. 2) then
call omp_set_num_threads(2)
print *, " more than 1 cpu, trying parallel"
else
print *, " only 1 cpu, program exiting"
return
endif
time1 = dclock()
!$omp parallel
!$omp do
do k = 1, 2
call delay(i)
enddo
!$omp end parallel
time2 = dclock() - time1
print *, " 2 cpu test - delay value doubled =", i*2,
* " time =", time2, " seconds"
return
end
Here is a simple delay program that wastes cycles w/o reading memory, so it should scale well.
subroutine delay(n)
integer n
integer i
do i=1,n
call abc()
end do
return
end
subroutine abc()
integer i
do i=1,1000000
call def()
end do
return
end
subroutine def()
return
end
To link using pgf77, do the following
pgf77 -Mnomain -Munix -mp -msvcrt -mslibs -o ex4 ex4main.o ex4.f delay.f dclock.s
Note that ex4main.obj was renamed ex4main.o (not necessary) Also note that the -mslibs switch causes the MS linker to be used.
To link using VC++
pgf77 -c -Munix -mp ex4.f pgf77 -c -Munix dclock.s pgf77 -c -Munix delay.f
in VC++ project settings, add the following files to the object/library module section of Link page in the front (you may also want to set the nodefault libs flag)
C:\PGI\nt86\lib\pgstdinit-mslibs.o my_dir\ex4.o my_dir\delay.o my_dir\dclock.o C:\PGI\nt86\lib\libpgmp.a <---- only needed when compiling -mp C:\PGI\nt86\lib\libpgftnrtl.a C:\PGI\nt86\lib\libpgc.a libc.lib oldnames.lib
On a machine configured with dual cpus, it should execute as follows-- (note that we execute the program with a value that will delay some amount - 113 for a 200 MHz machine was ~4 seconds)
PGI$ ex4main.exe 113 (main) calling ex4 with i = 113 num procs = 2 max threads = 1 1 cpu test - delay value doubled = 226 time = 8.014712320000399 seconds more than 1 cpu, trying parallel 2 cpu test - delay value doubled = 226 time = 4.028287394990912 seconds