// An HSA version of C++AMP program
int main ()
{
// Removed until linking/alloc qualifier issue is solved
auto XmallocFlag = newInit.XmallocFlag;
auto mallocFlag = newInit.mallocFlag;
auto ptr_a = newInit.ptr_a; // pointer to Xmalloc syscall numbers
auto ptr_b = newInit.ptr_b; // pointer to Xmalloc syscall parameters
auto ptr_c = newInit.ptr_c; // pointer to Xmalloc test results
auto ptr_x = newInit.ptr_x; // pointer to Xfree/free/malloc syscall numbers
auto ptr_y = newInit.ptr_y; // pointer to Xfree/free/malloc syscall parameters
auto ptr_z = newInit.ptr_z; // pointer to Xfree/free/malloc test results
// define inputs and output
const int vecSize = 16;
const int tileSize = 4;
const int tileCount = vecSize / tileSize;
// launch kernel
unsigned long int sumCPP[vecSize];
Concurrency::array_view<unsigned long int, 1> sum(vecSize, sumCPP);
parallel_for_each(
Concurrency::extent<1>(vecSize).tile<tileSize>(),
[=](Concurrency::tiled_index<tileSize> tidx) restrict(amp) {
// Removed until linking/alloc qualifier issue is solved
putXmallocFlag(XmallocFlag);
putMallocFlag(mallocFlag);
put_ptr_a(ptr_a);
put_ptr_b(ptr_b);
put_ptr_c(ptr_c);
put_ptr_x(ptr_x);
put_ptr_y(ptr_y);
put_ptr_z(ptr_z);
int local = tidx.local[0];
unsigned int *p = NULL;
if (local % 2 == 0) p = new unsigned int(local * 2);
sum[tidx.global[0]] = (unsigned long int)p;
delete p;
});
// An HSA version of C++AMP program
int main ()
{
// Removed until linking/alloc qualifier issue is solved
auto XmallocFlag = newInit.XmallocFlag;
auto mallocFlag = newInit.mallocFlag;
auto ptr_a = newInit.ptr_a; // pointer to Xmalloc syscall numbers
auto ptr_b = newInit.ptr_b; // pointer to Xmalloc syscall parameters
auto ptr_c = newInit.ptr_c; // pointer to Xmalloc test results
auto ptr_x = newInit.ptr_x; // pointer to Xfree/free/malloc syscall numbers
auto ptr_y = newInit.ptr_y; // pointer to Xfree/free/malloc syscall parameters
auto ptr_z = newInit.ptr_z; // pointer to Xfree/free/malloc test results
// define inputs and output
const int vecSize = 16;
const int tileSize = 4;
const int tileCount = vecSize / tileSize;
// launch kernel
unsigned long int sumCPP[vecSize];
Concurrency::array_view<unsigned long int, 1> sum(vecSize, sumCPP);
parallel_for_each(
Concurrency::extent<1>(vecSize),
[=](Concurrency::index<1> idx) restrict(amp) {
// Removed until linking/alloc qualifier issue is solved
putXmallocFlag(XmallocFlag);
putMallocFlag(mallocFlag);
put_ptr_a(ptr_a);
put_ptr_b(ptr_b);
put_ptr_c(ptr_c);
put_ptr_x(ptr_x);
put_ptr_y(ptr_y);
put_ptr_z(ptr_z);
sum[idx[0]] = (unsigned long int)new Point[2]();
});