pass object pointer

gistfile1.txt

#include <amp.h>
#include <vector>
#include <stdlib.h>
using namespace concurrency;
using namespace std;
const static int DOMAIN_SIZE = 1024;
const static int BLOCK_SIZE = 16;
typedef int TY;

template<typename type, int rank>
struct s1
{
    s1(Concurrency::array<type, rank> &a) __GPU : av_a(a) {}
    s1(Concurrency::extent<rank> e, type *pa) __GPU : av_a(e, pa){}
    ~s1() __GPU {}

    int placeholder1;
    array_view<type, rank> av_a;
    float placeholder2;
};

void bar0(index<1> idx, s1<TY, 1> *p) restrict(amp) {
  int * pa = &p->av_a[idx];
  *pa = 1; 
}

void bar(tiled_index<BLOCK_SIZE> idx, s1<TY, 1> *p) __GPU {
  int * pa = &p->av_a[idx];
  *pa = 1; 
}

void foo() {
    Concurrency::extent<1> g(DOMAIN_SIZE);
    vector<int> a(DOMAIN_SIZE);
    Concurrency::array<int, 1> a_a(g);
    
//    parallel_for_each(a_a.get_extent().template tile<BLOCK_SIZE>(), [&] (tiled_index<BLOCK_SIZE> idx) restrict(amp) {
    parallel_for_each(a_a.get_extent(), [&] (index<1> idx) restrict(amp) {
      s1<TY,1> o(a_a);
      bar0(idx, &o);
//        a_a[idx] = 1;
    });

    a = a_a;

    for (int i=0; i<DOMAIN_SIZE; ++i) {
      if (a[i] != 1) {
        fprintf(stderr, "wrong answer idx = %d bad value= %d\n", i, a[i]);
        exit(-1);
      }
    }
} 

int main() {
foo();
}

[1105123.203229] VM fault (0x01, vmid 1) at page 268404840, read from TC (72)
[1105138.940925] radeon 0000:01:00.0: GPU fault detected: 147 0x0da20401
[1105138.940928] radeon 0000:01:00.0: VM_CONTEXT1_PROTECTION_FAULT_ADDR 0x0FF02F6D
[1105138.940929] radeon 0000:01:00.0: VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x02004001
[1105138.940931] VM fault (0x01, vmid 1) at page 267399021, read from TC (4)