parallel max
parent
611300636a
commit
70f4290b3c
@ -0,0 +1,93 @@
|
|||||||
|
#include <iostream>
|
||||||
|
#include <thread>
|
||||||
|
#include <chrono>
|
||||||
|
#include <vector>
|
||||||
|
#include <random>
|
||||||
|
#include <algorithm>
|
||||||
|
#include <cstdint>
|
||||||
|
#include <functional>
|
||||||
|
|
||||||
|
using std::cout, std::endl, std::size_t, std::vector;
|
||||||
|
using namespace std::chrono_literals;
|
||||||
|
|
||||||
|
uint64_t get_max(vector<uint64_t>::iterator start, vector<uint64_t>::iterator end)
|
||||||
|
{
|
||||||
|
uint64_t result = *start;
|
||||||
|
while (start++ != end)
|
||||||
|
if (*start > result)
|
||||||
|
result = *start;
|
||||||
|
|
||||||
|
return result;
|
||||||
|
}
|
||||||
|
|
||||||
|
struct max_block {
|
||||||
|
void operator() (vector<uint64_t>::iterator start, vector<uint64_t>::iterator end, uint64_t& res) {
|
||||||
|
res = get_max(start, end);
|
||||||
|
}
|
||||||
|
};
|
||||||
|
|
||||||
|
uint64_t parallel_max(vector<uint64_t>::iterator start, vector<uint64_t>::iterator end) {
|
||||||
|
int const length = std::distance(start, end);
|
||||||
|
int const min_per_thread = 25;
|
||||||
|
int const max_threads = (length + min_per_thread - 1) / min_per_thread;
|
||||||
|
int const hardware_threads = std::thread::hardware_concurrency();
|
||||||
|
int const num_threads = std::min(hardware_threads!=0?hardware_threads:2, max_threads);
|
||||||
|
int const block_size = length / num_threads;
|
||||||
|
#ifdef _DEBUG
|
||||||
|
cout << "num_threads = " << num_threads << endl;
|
||||||
|
#endif
|
||||||
|
vector<uint64_t> results(num_threads);
|
||||||
|
vector<std::thread> threads(num_threads - 1);
|
||||||
|
auto block_start = start;
|
||||||
|
|
||||||
|
for(auto i = 0; i < num_threads - 1; ++i) {
|
||||||
|
auto block_end = block_start;
|
||||||
|
std::advance(block_end, block_size);
|
||||||
|
threads[i] = std::thread(max_block(), block_start, block_end, std::ref(results[i]));
|
||||||
|
block_start = block_end;
|
||||||
|
}
|
||||||
|
|
||||||
|
max_block() (block_start, end, results[num_threads - 1]);
|
||||||
|
|
||||||
|
std::for_each(threads.begin(), threads.end(), std::mem_fn(&std::thread::join));
|
||||||
|
|
||||||
|
return get_max(results.begin(), results.end());
|
||||||
|
|
||||||
|
}
|
||||||
|
|
||||||
|
int main()
|
||||||
|
{
|
||||||
|
cout << "Generating numbers!" << endl;
|
||||||
|
|
||||||
|
#ifdef _DEBUG
|
||||||
|
std::vector<uint64_t> numbers(1000);
|
||||||
|
numbers[0] = 5;
|
||||||
|
for (size_t i = 1; i < numbers.size(); ++i)
|
||||||
|
{
|
||||||
|
numbers[i] = numbers[i - 1] * i + 1;
|
||||||
|
|
||||||
|
}
|
||||||
|
#else
|
||||||
|
std::vector<uint64_t> numbers(5e8);
|
||||||
|
numbers[0] = 123456789;
|
||||||
|
for (size_t i = 1; i < numbers.size(); ++i)
|
||||||
|
{
|
||||||
|
numbers[i] = numbers[i - 1] * i + 1;
|
||||||
|
}
|
||||||
|
cout << "Numbers generated!" << endl;
|
||||||
|
#endif
|
||||||
|
|
||||||
|
auto start1 = std::chrono::high_resolution_clock::now();
|
||||||
|
uint64_t m1 = get_max(numbers.begin(), numbers.end());
|
||||||
|
auto end1 = std::chrono::high_resolution_clock::now();
|
||||||
|
cout << "Maximum = " << m1 << endl;
|
||||||
|
cout << "Time to calculate max = " << std::chrono::duration_cast<std::chrono::milliseconds>(end1 - start1).count()
|
||||||
|
<< " milliseconds." << endl;
|
||||||
|
|
||||||
|
auto start2 = std::chrono::high_resolution_clock::now();
|
||||||
|
uint64_t m2 = parallel_max(numbers.begin(), numbers.end());
|
||||||
|
auto end2 = std::chrono::high_resolution_clock::now();
|
||||||
|
cout << "Maximum = " << m2 << endl;
|
||||||
|
cout << "Time to calculate max = " << std::chrono::duration_cast<std::chrono::milliseconds>(end2 - start2).count()
|
||||||
|
<< " milliseconds." << endl;
|
||||||
|
}
|
Reference in New Issue