使用两个C++向量作为参数的“复杂”操作函数的最佳方法

html5 • • 问答 • 167 阅读

为了简单起见，请问用以下 Python 代码在 C++ 中编写以下代码的最佳方法是什么

import numpy as np

X = np.array([1,2,3,...])
Y = np.array([10,11,12,...])

def funct(X,Y):
  return 0.5 * X + 0.5 * 3 * Y

或者，在没有 Python 的情况下进行解释，有没有办法：

#include <vector>

std::vector<double> Z {};
std::vector<double> X {1,2,3,4};
std::vector<double> Y {5,6,7,8};

Z = 0.5 * X + 0.5 * 3 * Y; // something like this

回答

也许valarray是你需要的：

#include <iostream>
#include <valarray>

int main() {
    std::valarray<double> Z {};
    std::valarray<double> X {1,2,3,4};
    std::valarray<double> Y {5,6,7,8};
    Z = 0.5 * X + 0.5 * 3 * Y;

    for(auto v : Z) std::cout << v << 'n';
}

输出

8
10
12
14

请注意std::valarray：“当两个参数是大小不同的 valarray 时，行为未定义。 ”

  • Great to see a suggestion to use valarray! I rarely see it being used. It's really old, but how mature is it? I mean, is it properly supported by optimizers? Last time I tried to use a vallarray in ms vs2019, it turned out it was actually broken in the ms stl (its fixed now).
  • OK, found my mistake. Wrong constructor for std::valarray. Anyhow, [it would seen that valarray is 3.3x faster then vector/algo](https://quick-bench.com/q/G3X-qT_QMz--r8aGrTbMA6LlmKM) . But as valarray uses expression trees, I'm not sure thats actually the case. [I did a second bench where I accumulate the output](https://quick-bench.com/q/GFMW0tUmkmThxv3s3VbTgITkpNk), which gives a different result (~1.5x) but the accumulate is also in the bench loop, so the bench is not isolated. Still valarray seems faster! noice!
以上是使用两个C++向量作为参数的“复杂”操作函数的最佳方法的全部内容。
THE END
分享
二维码
< <上一篇
下一篇>>