https://gcc.gnu.org/bugzilla/show_bug.cgi?id=69858
Bug ID: 69858
Summary: OpenACC no free memory
Product: gcc
Version: 6.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: libgomp
Assignee: unassigned at gcc dot gnu.org
Reporter: oliveriandrea at gmail dot com
CC: jakub at gcc dot gnu.org
Target Milestone: ---
Hi,
i have compiled GCC SNAPSHOT 20160214 with nvpt-offload support.
i have tested it with this code
/***********************************************/
#include <stdio.h>
#define N 2000000000
#define vl 1024
int main(void) {
double pi = 0.0f;
long long i;
double t;
#pragma acc data copyout(pi)
{
#pragma acc parallel vector_length(vl)
#pragma acc loop gang worker vector reduction(+:pi) private(t)
for (i=0; i<N; i++) {
t= (double)((i+0.5)/N);
pi +=4.0/(1.0+t*t);
}
}
printf("pi=%11.10f\n",pi/N);
return 0;
}
/************************************************/
the result is correct pi=3,14....
but if i run again the program the output is
pi =6.28.....
and if i run again
pi=9.....
it seems that the GPU memory "remembers" the result...
if i substitute #pragma acc data copyout(pi) with #pragma acc data copy(pi)
the program works correctly.
