cory-master/BabelStream
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

all required operations implemented, errors are too large

Browse Source
This commit is contained in:
Peter Steinbach 2017-02-28 10:00:44 +01:00
parent f7af8ebc91
commit 383c5a3ae7
1 changed files with 74 additions and 34 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

108
HCStream.cpp
View File

@ -78,7 +78,49 @@ HCStream<T>::~HCStream()
template <class T> template <class T>
void HCStream<T>::init_arrays(T _a, T _b, T _c) void HCStream<T>::init_arrays(T _a, T _b, T _c)
{ {
std::cout << __FILE__ << ":" << __LINE__ << "\t" << "initializing arrays\n"; // hc::array_view<T,1> view_a(this->d_a);
// hc::array_view<T,1> view_b(this->d_b);
// hc::array_view<T,1> view_c(this->d_c);
// hc::completion_future future_a= hc::parallel_for_each(hc::extent<1>(array_size)
// , [=](hc::index<1> i) [[hc]] {
// view_a[i] = _a;
// });
// hc::completion_future future_b= hc::parallel_for_each(hc::extent<1>(array_size)
// , [=](hc::index<1> i) [[hc]] {
// view_b[i] = _b;
// });
// hc::completion_future future_c= hc::parallel_for_each(hc::extent<1>(array_size)
// , [=](hc::index<1> i) [[hc]] {
// view_c[i] = _c;
// });
// try{
// future_a.wait();
// }
// catch(std::exception& e){
// std::cout << __FILE__ << ":" << __LINE__ << "\t future_a " << e.what() << std::endl;
// throw;
// }
// try{
// future_b.wait();
// }
// catch(std::exception& e){
// std::cout << __FILE__ << ":" << __LINE__ << "\t future_b " << e.what() << std::endl;
// throw;
// }
// try{
// future_c.wait();
// }
// catch(std::exception& e){
// std::cout << __FILE__ << ":" << __LINE__ << "\t future_c " << e.what() << std::endl;
// throw;
// }
std::vector<T> temp(array_size,_a); std::vector<T> temp(array_size,_a);
hc::copy(temp.begin(), temp.end(),this->d_a); hc::copy(temp.begin(), temp.end(),this->d_a);
@ -93,8 +135,8 @@ void HCStream<T>::init_arrays(T _a, T _b, T _c)
template <class T> template <class T>
void HCStream<T>::read_arrays(std::vector<T>& a, std::vector<T>& b, std::vector<T>& c) void HCStream<T>::read_arrays(std::vector<T>& a, std::vector<T>& b, std::vector<T>& c)
{ {
// Copy device memory to host // Copy device memory to host
std::cout << __FILE__ << ":" << __LINE__ << "\t" << "read arrays\n";
hc::copy(d_a,a.begin()); hc::copy(d_a,a.begin());
hc::copy(d_b,b.begin()); hc::copy(d_b,b.begin());
hc::copy(d_c,c.begin()); hc::copy(d_c,c.begin());
@ -105,20 +147,19 @@ template <class T>
void HCStream<T>::copy() void HCStream<T>::copy()
{ {
std::cout << __FILE__ << ":" << __LINE__ << "\t" << "copy\n"; hc::array_view<T,1> view_a = this->d_a;
hc::array<T,1>& device_a = this->d_a; hc::array_view<T,1> view_c = this->d_c;
hc::array<T,1>& device_c = this->d_c;
try{ try{
// launch a GPU kernel to compute the saxpy in parallel // launch a GPU kernel to compute the saxpy in parallel
hc::completion_future future_kernel = hc::parallel_for_each(hc::extent<1>(array_size) hc::completion_future future_kernel = hc::parallel_for_each(hc::extent<1>(array_size)
, [&](hc::index<1> index) [[hc]] { , [=](hc::index<1> index) [[hc]] {
device_c[index] = device_a[index]; view_c[index] = view_a[index];
}); });
future_kernel.wait(); future_kernel.wait();
} }
catch(std::exception& e){ catch(std::exception& e){
std::cout << __FILE__ << ":" << __LINE__ << "\t" << e.what() << std::endl; std::cerr << __FILE__ << ":" << __LINE__ << "\t" << e.what() << std::endl;
throw; throw;
} }
} }
@ -126,21 +167,21 @@ void HCStream<T>::copy()
template <class T> template <class T>
void HCStream<T>::mul() void HCStream<T>::mul()
{ {
std::cout << __FILE__ << ":" << __LINE__ << "\t" << "mul\n";
const T scalar = 0.3; const T scalar = 0.3;
hc::array<T,1>& device_b = this->d_b; hc::array_view<T,1> view_b = this->d_b;
hc::array<T,1>& device_c = this->d_c; hc::array_view<T,1> view_c = this->d_c;
try{ try{
// launch a GPU kernel to compute the saxpy in parallel // launch a GPU kernel to compute the saxpy in parallel
hc::completion_future future_kernel = hc::parallel_for_each(hc::extent<1>(array_size) hc::completion_future future_kernel = hc::parallel_for_each(hc::extent<1>(array_size)
, [&](hc::index<1> i) [[hc]] { , [=](hc::index<1> i) [[hc]] {
device_b[i] = scalar*device_c[i]; view_b[i] = scalar*view_c[i];
}); });
future_kernel.wait(); future_kernel.wait();
} }
catch(std::exception& e){ catch(std::exception& e){
std::cout << __FILE__ << ":" << __LINE__ << "\t" << e.what() << std::endl; std::cerr << __FILE__ << ":" << __LINE__ << "\t" << e.what() << std::endl;
throw; throw;
} }
} }
@ -148,21 +189,22 @@ void HCStream<T>::mul()
template <class T> template <class T>
void HCStream<T>::add() void HCStream<T>::add()
{ {
std::cout << __FILE__ << ":" << __LINE__ << "\t" << "add\n";
hc::array<T,1>& device_a = this->d_a;
hc::array<T,1>& device_b = this->d_b; hc::array_view<T,1> view_a(this->d_a);
hc::array<T,1>& device_c = this->d_c; hc::array_view<T,1> view_b(this->d_b);
hc::array_view<T,1> view_c(this->d_c);
try{ try{
// launch a GPU kernel to compute the saxpy in parallel // launch a GPU kernel to compute the saxpy in parallel
hc::completion_future future_kernel = hc::parallel_for_each(hc::extent<1>(array_size) hc::completion_future future_kernel = hc::parallel_for_each(hc::extent<1>(array_size)
, [&](hc::index<1> i) [[hc]] { , [=](hc::index<1> i) [[hc]] {
device_c[i] = device_a[i]+device_b[i]; view_c[i] = view_a[i]+view_b[i];
}); });
future_kernel.wait(); future_kernel.wait();
} }
catch(std::exception& e){ catch(std::exception& e){
std::cout << __FILE__ << ":" << __LINE__ << "\t" << e.what() << std::endl; std::cerr << __FILE__ << ":" << __LINE__ << "\t" << e.what() << std::endl;
throw; throw;
} }
} }
@ -170,22 +212,22 @@ void HCStream<T>::add()
template <class T> template <class T>
void HCStream<T>::triad() void HCStream<T>::triad()
{ {
std::cout << __FILE__ << ":" << __LINE__ << "\t" << "triad\n";
const T scalar = 0.3; const T scalar = 0.3;
hc::array<T,1>& device_a = this->d_a; hc::array_view<T,1> view_a(this->d_a);
hc::array<T,1>& device_b = this->d_b; hc::array_view<T,1> view_b(this->d_b);
hc::array<T,1>& device_c = this->d_c; hc::array_view<T,1> view_c(this->d_c);
try{ try{
// launch a GPU kernel to compute the saxpy in parallel // launch a GPU kernel to compute the saxpy in parallel
hc::completion_future future_kernel = hc::parallel_for_each(hc::extent<1>(array_size) hc::completion_future future_kernel = hc::parallel_for_each(hc::extent<1>(array_size)
, [&](hc::index<1> i) [[hc]] { , [=](hc::index<1> i) [[hc]] {
device_a[i] = device_b[i] + scalar*device_c[i]; view_a[i] = view_b[i] + scalar*view_c[i];
}); });
future_kernel.wait(); future_kernel.wait();
} }
catch(std::exception& e){ catch(std::exception& e){
std::cout << __FILE__ << ":" << __LINE__ << "\t" << e.what() << std::endl; std::cerr << __FILE__ << ":" << __LINE__ << "\t" << e.what() << std::endl;
throw; throw;
} }
} }
@ -193,7 +235,7 @@ void HCStream<T>::triad()
template <class T> template <class T>
T HCStream<T>::dot() T HCStream<T>::dot()
{ {
std::cout << __FILE__ << ":" << __LINE__ << "\t" << "dot\n";
hc::array_view<T,1> view_a(this->d_a); hc::array_view<T,1> view_a(this->d_a);
hc::array_view<T,1> view_p(this->d_b); hc::array_view<T,1> view_p(this->d_b);
@ -202,22 +244,20 @@ T HCStream<T>::dot()
try{ try{
// launch a GPU kernel to compute the saxpy in parallel // launch a GPU kernel to compute the saxpy in parallel
hc::completion_future future_kernel = hc::parallel_for_each(view_a.get_extent(), hc::completion_future future_kernel = hc::parallel_for_each(view_a.get_extent(),
[&](hc::index<1> i) [[hc]] { [=](hc::index<1> i) [[hc]] {
view_p[i] = view_p[i]*view_a[i]; view_p[i] = view_p[i]*view_a[i];
}); });
future_kernel.wait(); future_kernel.wait();
} }
catch(std::exception& e){ catch(std::exception& e){
std::cout << __FILE__ << ":" << __LINE__ << "\t" << e.what() << std::endl; std::cerr << __FILE__ << ":" << __LINE__ << "\t" << e.what() << std::endl;
throw; throw;
} }
std::cout << __FILE__ << ":" << __LINE__ << "\t" << "dot - for_each done\n";
std::vector<T> h_product(array_size,sum); std::vector<T> h_product(array_size,sum);
hc::copy(view_p,h_product.begin()); hc::copy(view_p,h_product.begin());
std::cout << __FILE__ << ":" << __LINE__ << "\t" << "dot - copy-out done\n";
sum = std::accumulate(h_product.begin(), h_product.end(),sum); sum = std::accumulate(h_product.begin(), h_product.end(),sum);
return sum; return sum;