I tried to add parallel execution to my c++ project.
Thus, based on this example I defined my app.cpp as follows:
#include <chrono>
#include <iostream>
#include <omp.h>
int sum_serial(int n) {
int sum = 0;
for (int i = 0; i <= n; ++i) {
sum += i;
}
return sum;
}
// Parallel programming function
int sum_parallel(int n) {
int sum = 0;
#pragma omp parallel for reduction(+ : sum)
for (int i = 0; i <= n; ++i) {
sum += i;
}
return sum;
}
int main(int argc, char* argv[]) {
// Beginning of parallel region
#pragma omp parallel
{ printf("Hello World... from thread = %d\n", omp_get_thread_num()); }
// Set threads number.
#if defined(_OPENMP)
omp_set_num_threads(2);
#endif
{
const int n = 100000000;
auto start_time = std::chrono::high_resolution_clock::now();
int result_serial = sum_serial(n);
auto end_time = std::chrono::high_resolution_clock::now();
std::chrono::duration<double> serial_duration = end_time - start_time;
start_time = std::chrono::high_resolution_clock::now();
int result_parallel = sum_parallel(n);
end_time = std::chrono::high_resolution_clock::now();
std::chrono::duration<double> parallel_duration = end_time - start_time;
std::cout << "Serial result: " << result_serial << std::endl;
std::cout << "Parallel result: " << result_parallel << std::endl;
std::cout << "Serial duration: " << serial_duration.count() << " seconds" << std::endl;
std::cout << "Parallel duration: " << parallel_duration.count() << " seconds" << std::endl;
std::cout << "Speedup: " << serial_duration.count() / parallel_duration.count()
<< std::endl;
}
return 0;
}
What surprised me is that there is no speedup, in fact, the parallel execution is quite a lot slower. My output is:
Serial result: 987459712
Parallel result: 987459712
Serial duration: 0.132073 seconds
Parallel duration: 0.645815 seconds
Speedup: 0.204507
Note that my cmake is:
add_executable(wingdesigner app.cpp)
target_compile_features(app PRIVATE cxx_std_17)
add_compile_options(-Wall -O3 -fopenmp)
target_link_libraries(app PUBLIC gomp)
I must be doing something wrong. I get it that large number of threads can result in barely observable speedup but in this particular case 2 threads compared to 1 should be faster, right? I assume my compiling is wrong but I can't figure out what the problem would be.
I tried your example and noticed that only one thread was used, even with OMP.
As a matter of fact, it seems that your CMake file is wrong (should use target_compile_options instead of add_compile_options). I used the following and observed some speedup:
project (foobar)
add_executable(app app.cpp)
target_compile_features(app PRIVATE cxx_std_17)
target_compile_options (app PUBLIC -Wall -O3 -fopenmp)
target_link_libraries (app PUBLIC gomp)
I also used the following snippet (using sin) in order to have "more" work to do:
#include <chrono>
#include <iostream>
#include <omp.h>
#include <cmath>
auto sum_serial(int n) {
double sum = 0;
for (int i = 0; i <= n; ++i) {
sum += sin(i);
}
return sum;
}
// Parallel programming function
auto sum_parallel(int n) {
double sum = 0;
#pragma omp parallel for reduction(+ : sum)
for (int i = 0; i <= n; ++i) {
sum += sin(i);
}
return sum;
}
int main(int argc, char* argv[]) {
// Beginning of parallel region
#pragma omp parallel
{ printf("Hello World... from thread = %d\n", omp_get_thread_num()); }
// Set threads number.
#if defined(_OPENMP)
omp_set_num_threads(2);
#endif
{
const int n = 100000000;
auto start_time = std::chrono::high_resolution_clock::now();
auto result_serial = sum_serial(n);
auto end_time = std::chrono::high_resolution_clock::now();
std::chrono::duration<double> serial_duration = end_time - start_time;
start_time = std::chrono::high_resolution_clock::now();
auto result_parallel = sum_parallel(n);
end_time = std::chrono::high_resolution_clock::now();
std::chrono::duration<double> parallel_duration = end_time - start_time;
std::cout << "Serial result : " << result_serial << std::endl;
std::cout << "Parallel result : " << result_parallel << std::endl;
std::cout << "Serial duration : " << serial_duration.count() << " seconds" << std::endl;
std::cout << "Parallel duration: " << parallel_duration.count() << " seconds" << std::endl;
std::cout << "Speedup : " << serial_duration.count() / parallel_duration.count() << std::endl;
}
return 0;
}
I got the following output:
Serial result : 1.71365
Parallel result : 1.71365
Serial duration : 1.08041 seconds
Parallel duration: 0.546919 seconds
Speedup : 1.97545