57 lines
1.3 KiB
C++
57 lines
1.3 KiB
C++
|
#include "binary_tree.hpp"
|
||
|
|
#include <tbb/global_control.h>
|
||
|
|
#include <tbb/flow_graph.h>
|
||
|
|
|
||
|
|
// binary_tree_tbb
|
||
|
|
void binary_tree_tbb(size_t num_layers, unsigned num_threads) {
|
||
|
|
|
||
|
|
using namespace tbb;
|
||
|
|
using namespace tbb::flow;
|
||
|
|
|
||
|
|
tbb::global_control c(
|
||
|
|
tbb::global_control::max_allowed_parallelism, num_threads
|
||
|
|
);
|
||
|
|
|
||
|
|
std::atomic<size_t> counter {0};
|
||
|
|
|
||
|
|
graph g;
|
||
|
|
|
||
|
|
std::vector<continue_node<continue_msg>*> tasks(1 << num_layers);
|
||
|
|
|
||
|
|
for(unsigned i=1; i<tasks.size(); i++) {
|
||
|
|
tasks[i] = new continue_node<continue_msg>(g,
|
||
|
|
[&]( const continue_msg& ) {
|
||
|
|
counter.fetch_add(1, std::memory_order_relaxed);
|
||
|
|
}
|
||
|
|
);
|
||
|
|
}
|
||
|
|
|
||
|
|
for(unsigned i=1; i<tasks.size(); i++) {
|
||
|
|
unsigned l = i << 1;
|
||
|
|
unsigned r = l + 1;
|
||
|
|
if(l < tasks.size() && r < tasks.size()) {
|
||
|
|
make_edge(*tasks[i], *tasks[l]);
|
||
|
|
make_edge(*tasks[i], *tasks[r]);
|
||
|
|
}
|
||
|
|
}
|
||
|
|
|
||
|
|
tasks[1]->try_put(continue_msg());
|
||
|
|
g.wait_for_all();
|
||
|
|
|
||
|
|
for(auto& task : tasks) {
|
||
|
|
delete task;
|
||
|
|
}
|
||
|
|
|
||
|
|
assert(counter + 1 == tasks.size());
|
||
|
|
}
|
||
|
|
|
||
|
|
std::chrono::microseconds measure_time_tbb(
|
||
|
|
size_t num_layers,
|
||
|
|
unsigned num_threads
|
||
|
|
) {
|
||
|
|
auto beg = std::chrono::high_resolution_clock::now();
|
||
|
|
binary_tree_tbb(num_layers, num_threads);
|
||
|
|
auto end = std::chrono::high_resolution_clock::now();
|
||
|
|
return std::chrono::duration_cast<std::chrono::microseconds>(end - beg);
|
||
|
|
}
|