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

Add timer to basic_io example #321

Merged
merged 3 commits into from
Dec 4, 2023
Merged

Add timer to basic_io example #321

Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
c4dad0c
Add timer to basic_io example
yncxcw Nov 25, 2023
20516a2
Update Trimer class
yncxcw Nov 30, 2023
e27b111
Merge branch 'branch-24.02' into branch-23.12
madsbk Nov 30, 2023
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
42 changes: 41 additions & 1 deletion cpp/examples/basic_io.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -14,6 +14,7 @@
* limitations under the License.
*/

#include <chrono>
#include <iostream>

#include <cuda_runtime_api.h>
Expand All @@ -27,6 +28,25 @@

using namespace std;

class Timer {
public:
Timer() : start(std::chrono::high_resolution_clock::now()) {}

~Timer()
{
auto end = std::chrono::high_resolution_clock::now();
auto start_ms =
std::chrono::time_point_cast<std::chrono::microseconds>(start).time_since_epoch().count();
auto end_ms =
std::chrono::time_point_cast<std::chrono::microseconds>(end).time_since_epoch().count();

cout << "(" << end_ms - start_ms << " us)" << endl;
}

private:
std::chrono::time_point<std::chrono::high_resolution_clock> start;
};

void check(bool condition)
{
if (!condition) {
Expand Down Expand Up @@ -82,6 +102,8 @@ int main()
check(kvikio::is_host_memory(c_dev) == false);

{
cout << endl;
Timer timer;
kvikio::FileHandle f("/tmp/test-file", "w");
check(cudaMemcpy(a_dev, a, SIZE, cudaMemcpyHostToDevice) == cudaSuccess);
size_t written = f.pwrite(a_dev, SIZE, 0, 1).get();
Expand All @@ -90,6 +112,8 @@ int main()
cout << "Write: " << written << endl;
}
{
std::cout << std::endl;
Timer timer;
kvikio::FileHandle f("/tmp/test-file", "r");
size_t read = f.pread(b_dev, SIZE, 0, 1).get();
check(read == SIZE);
Expand All @@ -102,6 +126,8 @@ int main()
}
kvikio::defaults::thread_pool_nthreads_reset(16);
{
std::cout << std::endl;
Timer timer;
kvikio::FileHandle f("/tmp/test-file", "w");
size_t written = f.pwrite(a_dev, SIZE).get();
check(written == SIZE);
Expand All @@ -110,16 +136,20 @@ int main()
<< " threads): " << written << endl;
}
{
std::cout << std::endl;
Timer timer;
kvikio::FileHandle f("/tmp/test-file", "r");
size_t read = f.pread(b_dev, SIZE, 0).get();
cout << "Parallel write (" << kvikio::defaults::thread_pool_nthreads() << " threads): " << read
cout << "Parallel read (" << kvikio::defaults::thread_pool_nthreads() << " threads): " << read
<< endl;
check(cudaMemcpy(b, b_dev, SIZE, cudaMemcpyDeviceToHost) == cudaSuccess);
for (int i = 0; i < NELEM; ++i) {
check(a[i] == b[i]);
}
}
{
std::cout << std::endl;
Timer timer;
kvikio::FileHandle f("/tmp/test-file", "r+", kvikio::FileHandle::m644);
kvikio::buffer_register(c_dev, SIZE);
size_t read = f.pread(c_dev, SIZE).get();
Expand All @@ -129,6 +159,8 @@ int main()
cout << "Read buffer registered data: " << read << endl;
}
{
std::cout << std::endl;
Timer timer;
kvikio::FileHandle f("/tmp/test-file", "w");
size_t written = f.pwrite(a, SIZE).get();
check(written == SIZE);
Expand All @@ -137,6 +169,8 @@ int main()
<< " threads): " << written << endl;
}
{
std::cout << std::endl;
Timer timer;
kvikio::FileHandle f("/tmp/test-file", "r");
size_t read = f.pread(b, SIZE).get();
check(read == SIZE);
Expand All @@ -148,6 +182,8 @@ int main()
<< " threads): " << read << endl;
}
if (kvikio::is_batch_and_stream_available() && !kvikio::defaults::compat_mode()) {
std::cout << std::endl;
Timer timer;
// Here we use the batch API to read "/tmp/test-file" into `b_dev` by
// submitting 4 batch operations.
constexpr int num_ops_in_batch = 4;
Expand Down Expand Up @@ -198,6 +234,8 @@ int main()
cout << "The batch API isn't available, requires CUDA 12.2+" << endl;
}
{
std::cout << std::endl;
Timer timer;
cout << "Performing async I/O using by-reference arguments" << endl;
off_t f_off{0};
off_t d_off{0};
Expand Down Expand Up @@ -231,6 +269,8 @@ int main()
check(cudaFreeHost((void*)bytes_done_p) == cudaSuccess);
}
{
std::cout << std::endl;
Timer timer;
cout << "Performing async I/O using by-value arguments" << endl;

// Let's create a new stream and submit an async write
Expand Down