Skip to content

Thrust: add kernel, async CUDA #2

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

Open
siboehm wants to merge 1 commit into
base: master
Choose a base branch
from
Open

Thrust: add kernel, async CUDA #2

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
53 changes: 36 additions & 17 deletions vector_add/vector_add_thrust.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,22 +1,26 @@
#include "vector_add/util.hpp"
#include <chrono>
#include <cstdint>
#include <iostream>
#include <chrono>
#include "vector_add/util.hpp"
#include <thrust/host_vector.h>
#include <thrust/device_vector.h>
#include <thrust/execution_policy.h>
#include <thrust/host_vector.h>
#include <thrust/sequence.h>

constexpr size_t NUM = 32384 * 32384;

int main()
{
std::cout << "Thrust backend: ";
__device__ float add(float a, float b) { return a + b; }

int main() {
std::cout << "Thrust backend: ";
#if THRUST_DEVICE_SYSTEM == 0
std::cout << "undefined" << std::endl;
#elif THRUST_DEVICE_SYSTEM == THRUST_DEVICE_SYSTEM_CUDA
std::cout << "CUDA" << std::endl;
auto executionPolicy = thrust::cuda::par_nosync;
#elif THRUST_DEVICE_SYSTEM == THRUST_DEVICE_SYSTEM_TBB
std::cout << "TBB" << std::endl;
auto executionPolicy = thrust::device;
#endif

auto start = std::chrono::system_clock::now();
Expand All @@ -29,32 +33,47 @@ std::cout << "Thrust backend: ";
thrust::device_vector<float> c_device(a_device.size());
print_elapsed(&start, "initialize input memory");

thrust::transform(a_device.begin(), a_device.end(), b_device.begin(), c_device.begin(), thrust::plus<float>());
thrust::for_each(executionPolicy,
thrust::make_zip_iterator(thrust::make_tuple(
a_device.begin(), b_device.begin(), c_device.begin())),
thrust::make_zip_iterator(thrust::make_tuple(
a_device.end(), b_device.end(), c_device.end())),
[] __device__(thrust::tuple<float &, float &, float &> t) {
float &a = thrust::get<0>(t);
float &b = thrust::get<1>(t);
float &c = thrust::get<2>(t);
c = add(a, b);
});

#if THRUST_DEVICE_SYSTEM == THRUST_DEVICE_SYSTEM_CUDA
cudaDeviceSynchronize();
print_elapsed(&start, "run kernel");
// copy data from GPU to CPU
thrust::host_vector<float> a = a_device;
thrust::host_vector<float> b = b_device;
thrust::host_vector<float> c = c_device;
#else
thrust::device_vector<float>& a = a_device;
thrust::device_vector<float>& b = b_device;
thrust::device_vector<float>& c = c_device;
print_elapsed(&start, "run kernel");
thrust::device_vector<float> &a = a_device;
thrust::device_vector<float> &b = b_device;
thrust::device_vector<float> &c = c_device;
#endif
print_elapsed(&start, "run kernel and copy from device memory");
print_elapsed(&start, "copy from device memory");

// check the results
int errors = 0;
for (size_t i = 0; i < NUM; i++) {
if (c[i] != (a[i] + b[i])) {
errors++;
std::cout << "Error at index " << i << ": Expected " << a[i] + b[i] << ", got " << c[i] << "\n";
}
if (c[i] != (a[i] + b[i])) {
errors++;
std::cout << "Error at index " << i << ": Expected " << a[i] + b[i]
<< ", got " << c[i] << "\n";
}
}
print_elapsed(&start, "check results");
if (errors) {
printf("FAILED: %d errors\n", errors);
printf("FAILED: %d errors\n", errors);
} else {
printf("PASSED!\n");
printf("PASSED!\n");
}
return errors;
}