I'm trying to adapt a normal code to a parallel one. When I make a for loop parallel, which has some variables declared inside of it, are those variables private or shared?
Should I define each of them as private when I define the pragma?
By the way, one last question. I can use for-pragma with an initial parameter of an iterator too right? like for(iter=xlist.begin() ; ... )
I use latest version of codeblocks and mingw.
Anything declared inside a parallel region (whether it's a loop or not) is private, with exceptions as listed below in @ejd 's comment. You can't list it as private on the #pragma
line as the variable doesn't exist yet.
So for instance in the below, even with default(none)
we don't need to specify the sharing of tid
even if you could; it's inside the parallel section, so it is private to each thread. (Note too that you don't really need to specify i as private, as the loop index of an omp for is necessarily private.)
$ more foo.c
#include <stdio.h>
#include <omp.h>
int main(int argc, char *argv[]) {
int i;
#pragma omp parallel for default(none) private(i)
for(i=0;i<10;i++){
int tid = omp_get_thread_num();
printf("Thread %d gets iteration %d\n", tid, i);
}
return 0;
}
gpc-f103n084-$ !g
gcc -o foo foo.c -fopenmp
$ ./foo
Thread 1 gets iteration 2
Thread 1 gets iteration 3
Thread 3 gets iteration 6
Thread 3 gets iteration 7
Thread 0 gets iteration 0
Thread 0 gets iteration 1
Thread 4 gets iteration 8
Thread 4 gets iteration 9
Thread 2 gets iteration 4
Thread 2 gets iteration 5
If you love us? You can donate to us via Paypal or buy me a coffee so we can maintain and grow! Thank you!
Donate Us With