Skip to content

hw06/tbb #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
wants to merge 2 commits into
base: main
Choose a base branch
from
Open

hw06/tbb #2

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
f5c7c24
hw06/tbb
Tigerrr07 Jan 19, 2022
70a63a2
+ -> *
Tigerrr07 Jan 19, 2022
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
167 changes: 139 additions & 28 deletions main.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5,76 +5,187 @@
#include <numeric>
#include <algorithm>
#include "ticktock.h"

#include <tbb/parallel_for.h>
#include <tbb/parallel_reduce.h>
#include "pod.h"
#include <atomic>
#include <tbb/parallel_scan.h>
// TODO: 并行化所有这些 for 循环
// 使用任务域指定使用的线程数4

template <class T, class Func>
std::vector<T> fill(std::vector<T> &arr, Func const &func) {
TICK(fill);
for (size_t i = 0; i < arr.size(); i++) {
arr[i] = func(i);
}
// for (size_t i = 0; i < arr.size(); i++) {
// arr[i] = func(i);
// }

// fiil(x) 0.951s -> 0.258s 提高3.69倍
// fill(y) 0.978s -> 0.253s 提高3.86倍
tbb::task_arena ta(4);
ta.execute([&] {
tbb::parallel_for(tbb::blocked_range<size_t>(0, arr.size()),
[&] (tbb::blocked_range<size_t> r) {
for (size_t i = r.begin(); i < r.end(); i++) {
arr[i] = func(i);
}
});
});
TOCK(fill);
return arr;
}

template <class T>
void saxpy(T a, std::vector<T> &x, std::vector<T> const &y) {
TICK(saxpy);
for (size_t i = 0; i < x.size(); i++) {
x[i] = a * x[i] + y[i];
}
// for (size_t i = 0; i < x.size(); i++) {
// x[i] = a * x[i] + y[i];
// }
// 0.042s -> 0.021s 提高2倍
tbb::task_arena ta(4);
ta.execute([&] {
tbb::parallel_for(tbb::blocked_range<size_t>(0, x.size()),
[&] (tbb::blocked_range<size_t> r) {
for (size_t i = r.begin(); i < r.end(); i++) {
x[i] = a * x[i] + y[i];
}
}, tbb::auto_partitioner{});
});
TOCK(saxpy);
}

template <class T>
T sqrtdot(std::vector<T> const &x, std::vector<T> const &y) {
TICK(sqrtdot);
// 串行reduce float类型会出现浮点误差
// T ret = 0;
// for (size_t i = 0; i < std::min(x.size(), y.size()); i++) {
// ret += x[i] * y[i];
// }

// 0.0916s -> 0.0214s 提高3.88倍
T ret = 0;
for (size_t i = 0; i < std::min(x.size(), y.size()); i++) {
ret += x[i] * y[i];
}
tbb::task_arena ta(4);

ta.execute([&] {
ret = tbb::parallel_reduce(tbb::blocked_range<size_t>(0, std::min(x.size(), y.size())), T(0),
[&] (tbb::blocked_range<size_t> r, T local_ret) {
for (size_t i = r.begin(); i < r.end(); i++) {
local_ret += x[i] * y[i];
}
return local_ret;
}, [] (T x, T y) {
return x + y;
});
});
ret = std::sqrt(ret);
TOCK(sqrtdot);
return ret;
}

// 求最小值也是reduce
template <class T>
T minvalue(std::vector<T> const &x) {
TICK(minvalue);
// T ret = x[0];
// for (size_t i = 1; i < x.size(); i++) {
// if (x[i] < ret)
// ret = x[i];
// }

// 0.092s -> 0.023s 提高4倍
T ret = x[0];
for (size_t i = 1; i < x.size(); i++) {
if (x[i] < ret)
ret = x[i];
}
tbb::task_arena ta(4);

ta.execute([&] {
ret = tbb::parallel_reduce(tbb::blocked_range<size_t>(0, x.size()), x[1],
[&] (tbb::blocked_range<size_t> r, T local_ret) {
for (size_t i = r.begin(); i < r.end(); i++) {
if (x[i] < local_ret)
local_ret = x[i];
}
return local_ret;
}, [] (T x, T y) {
return std::min(x, y);
});
});
TOCK(minvalue);
return ret;
}

// 这里使用了小彭老师写的pod.h 使用resize不初始化其中的值
template <class T>
std::vector<T> magicfilter(std::vector<T> const &x, std::vector<T> const &y) {
auto magicfilter(std::vector<T> const &x, std::vector<T> const &y) {
TICK(magicfilter);
std::vector<T> res;
for (size_t i = 0; i < std::min(x.size(), y.size()); i++) {
if (x[i] > y[i]) {
res.push_back(x[i]);
} else if (y[i] > x[i] && y[i] > 0.5f) {
res.push_back(y[i]);
res.push_back(x[i] * y[i]);
}
}
// std::vector<T> res;
// for (size_t i = 0; i < std::min(x.size(), y.size()); i++) {
// if (x[i] > y[i]) {
// res.push_back(x[i]);
// } else if (y[i] > x[i] && y[i] > 0.5f) {
// res.push_back(y[i]);
// res.push_back(x[i] * y[i]);
// }
// }

// 0.41s -> 0.05s 提高8倍
std::vector<pod<T>> res; // 避免初始化
std::atomic<size_t> res_size = 0;
size_t n = std::min(x.size(), y.size());
res.resize(n);

tbb::task_arena ta(4);

ta.execute([&] {
tbb::parallel_for(tbb::blocked_range<size_t>(0, n),
[&] (tbb::blocked_range<size_t> r) {
std::vector<pod<T>> local_res(r.size());
size_t lrsize = 0;
for (size_t i = r.begin(); i < r.end(); i++) {
if (x[i] > y[i]) {
local_res[lrsize++] = x[i];
} else if (y[i] > x[i] && y[i] > 0.5f) {
local_res[lrsize++] = y[i];
local_res[lrsize++] = x[i] * y[i];
}
}
size_t base = res_size.fetch_add(lrsize);
for (size_t i = 0; i < lrsize; i++) {
res[base + i] = local_res[i];
}
});
});
res.resize(res_size);
TOCK(magicfilter);
return res;
}

template <class T>
T scanner(std::vector<T> &x) {
TICK(scanner);
// T ret = 0;
// for (size_t i = 0; i < x.size(); i++) {
// ret += x[i];
// x[i] = ret;
// }

// 0.09s -> 0.047s 提高1.91倍
T ret = 0;
for (size_t i = 0; i < x.size(); i++) {
ret += x[i];
x[i] = ret;
}
tbb::task_arena ta(4);

ta.execute([&] {
ret = tbb::parallel_scan(tbb::blocked_range<size_t>(0, x.size()), T(0),
[&] (tbb::blocked_range<size_t> r, T local_ret, auto is_final) {
for (size_t i = r.begin(); i < r.end(); i++) {
local_ret += x[i];
if (is_final) {
x[i] = local_ret;
}
}
return local_ret;
}, [] (T x, T y) {
return x + y;
});
});
TOCK(scanner);
return ret;
}
Expand Down