Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

hcc - support for multiple devices #1

Open
wants to merge 3 commits into
base: hcc
Choose a base branch
from
Open

hcc - support for multiple devices #1

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
89f12ec
allocate array and launch kernel on the selected device
scchan Aug 16, 2016
2c73ad1
changing references hc::array to local references in order to avoid t…
scchan Aug 29, 2016
f16f7bd
remove debug code and messages
scchan Sep 2, 2016
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
41 changes: 29 additions & 12 deletions HCCStream.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -8,6 +8,7 @@
#include <codecvt>
#include <vector>
#include <locale>
#include <cstdio>

#include "HCCStream.h"

Expand Down Expand Up @@ -44,14 +45,13 @@ void listDevices(void)
}
}



template <class T>
HCCStream<T>::HCCStream(const unsigned int ARRAY_SIZE, const int device_index):
array_size(ARRAY_SIZE),
d_a(ARRAY_SIZE),
d_b(ARRAY_SIZE),
d_c(ARRAY_SIZE)
accelerator(hc::accelerator::get_all()[device_index]),
d_a(ARRAY_SIZE,accelerator.get_default_view()),
d_b(ARRAY_SIZE,accelerator.get_default_view()),
d_c(ARRAY_SIZE,accelerator.get_default_view())
{

// The array size must be divisible by TBSIZE for kernel launches
Expand All @@ -67,7 +67,6 @@ HCCStream<T>::HCCStream(const unsigned int ARRAY_SIZE, const int device_index):
auto current = accs[device_index];

std::cout << "Using HCC device " << getDeviceName(current) << std::endl;


// Check buffers fit on the device
// TODO: unclear how to do that!!
Expand Down Expand Up @@ -96,15 +95,19 @@ void HCCStream<T>::read_arrays(std::vector<T>& a, std::vector<T>& b, std::vector
hc::copy(d_c,c.begin());
}


template <class T>
void HCCStream<T>::copy()
{
hc::array<T> &d_a = this->d_a;
hc::array<T> &d_c = this->d_c;

try{
// 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(accelerator.get_default_view()
, hc::extent<1>(array_size)
, [&](hc::index<1> i) __attribute((hc)) {
d_c[i] = d_a[i];

d_c[i] = d_a[i];
});
future_kernel.wait();
}
Expand All @@ -117,10 +120,14 @@ void HCCStream<T>::copy()
template <class T>
void HCCStream<T>::mul()
{
hc::array<T> &d_b = this->d_b;
hc::array<T> &d_c = this->d_c;

const T scalar = 0.3;
try{
// 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(accelerator.get_default_view()
, hc::extent<1>(array_size)
, [&](hc::index<1> i) __attribute((hc)) {
d_b[i] = scalar*d_c[i];
});
Expand All @@ -135,9 +142,14 @@ void HCCStream<T>::mul()
template <class T>
void HCCStream<T>::add()
{
hc::array<T> &d_a = this->d_a;
hc::array<T> &d_b = this->d_b;
hc::array<T> &d_c = this->d_c;

try{
// 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(accelerator.get_default_view()
, hc::extent<1>(array_size)
, [&](hc::index<1> i) __attribute((hc)) {
d_c[i] = d_a[i]+d_b[i];
});
Expand All @@ -152,10 +164,15 @@ void HCCStream<T>::add()
template <class T>
void HCCStream<T>::triad()
{
hc::array<T> &d_a = this->d_a;
hc::array<T> &d_b = this->d_b;
hc::array<T> &d_c = this->d_c;

const T scalar = 0.3;
try{
// 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(accelerator.get_default_view()
, hc::extent<1>(array_size)
, [&](hc::index<1> i) __attribute((hc)) {
d_a[i] = d_b[i] + scalar*d_c[i];
});
Expand Down
3 changes: 3 additions & 0 deletions HCCStream.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -23,6 +23,9 @@ class HCCStream : public Stream<T>
// Size of arrays
unsigned int array_size;

// Selected device
hc::accelerator accelerator;

// Device side pointers to arrays
hc::array<T,1> d_a;
hc::array<T,1> d_b;
Expand Down
5 changes: 2 additions & 3 deletions main.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -40,6 +40,7 @@

// Default size of 2^25
unsigned int ARRAY_SIZE = 33554432;

unsigned int num_times = 100;
unsigned int deviceIndex = 0;
bool use_float = false;
Expand Down Expand Up @@ -133,7 +134,7 @@ void run()
#elif defined(HCC)
// Use the "reference" OpenMP 3 implementation
stream = new HCCStream<T>(ARRAY_SIZE, deviceIndex);

#endif

stream->write_arrays(a, b, c);
Expand Down Expand Up @@ -213,9 +214,7 @@ void run()
<< std::endl;

}

delete stream;

}

template <typename T>
Expand Down