矢量化扩展

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66

#include <array>
#include <chrono>
#include <iostream>
#include <string>
#include <vector>
#include "boost/core/noncopyable.hpp"

using std::chrono::duration_cast;
using std::chrono::high_resolution_clock;
class BenchTimer : boost::noncopyable {
  high_resolution_clock::time_point tp;
  std::string name;
  std::vector<std::string> rd;
  std::vector<double> dt;

  public:
  inline void st(const std::string &name) {
    this->name = name;
    tp = high_resolution_clock::now();
  }
  inline void end() {
    dt.emplace_back(duration_cast<std::chrono::nanoseconds>(high_resolution_clock::now() - tp).count());
    rd.emplace_back(this->name);
  }
  void showBenchTest() {
    for (int i = 0, size = static_cast<int>(rd.size()); i < size; i++) {
      std::cout << rd[i] << ":" << dt[i] / 1000 << "\n";
    }
  }
};

int main() {
  std::vector<int> arr_a(256, 0);
  std::vector<int> arr_b(256, 100);
  typedef int int256 __attribute__((vector_size(256 * sizeof(int))));
  int256 a;
  int256 b;

  for (int i = 0; i < 256; i++) {
    arr_a[i] = i;
    a[i] = i;
    arr_b[i] = 256 - i;
    b[i] = 256 - i;
  }
  int t = 10000;
  auto bench = BenchTimer();
  #pragma clang optimize off
  bench.st("foreach");
  while (t > 0) {
    for (auto i = 0; i < 256; i++) {
      arr_a[i] = 2 * arr_b[i];
    }
    t--;
  }
  bench.end();
  t = 10000;
  bench.st("vec");
  while (t > 0) {
    a = 2 * b;
    t--;
  }
  bench.end();
  #pragma clang optimize on
  bench.showBenchTest();
  std::cout << "2*b[0]==" << a[0] << std::endl;
}