Eigen、NumCpp、NumPy、CuPy运行时间对比

最新推荐文章于 2024-05-23 08:44:35 发布
最新推荐文章于 2024-05-23 08:44:35 发布
阅读量2.5k 收藏 7
点赞数 4
分类专栏： C++ python 文章标签： 算法 c++ python
原文链接：https://blog.csdn.net/u013468614/article/details/109383101
版权

对于矩阵运算，著名的C++库有Eigen与C++版本的Numpy——NumCpp，Python库有NumPy与带GPU加速的Numpy版本CuPy。博客园|Eigen vs Numpy时间对比中对Eigen与Numpy的效率进行了比较，但是比较时忽略了C++的编译模式[DeBug与Release]对结果的影响，对比出的结果是不够准确的。本篇对Eigen、NumCpp、NumPy以及CuPy四种主流矩阵运算库的计算效率进行比较，以备实战时作为考量的一点依据。

 

 

一、测试项目

本博客主要测试四种库对静态矩阵与动态矩阵运算的效率，主要分为两个项目：一）静态矩阵计算；二）动态矩阵计算。

二、静态矩阵计算效率测试

2.1 测试数据与函数

假定有向量集D = { d ⃗ 1 ∈ R M × 1 , . . . , d ⃗ N ∈ R M × 1 } D=\{\vec{d}_{1}\in R^{M\times 1},...,\vec{d}_{N}\in R^{M\times 1}\}D={d1​∈RM×1,...,dN​∈RM×1}与向量x ⃗ ∈ R M × 1 \vec{x}\in R^{M\times 1}x∈RM×1，测试计算以下函数为：

y i = exp ⁡ ( − ∣ ∣ x ⃗ − d ⃗ i ∣ ∣ 2 2 δ 2 ) ， i = 1 , 2 , . . . , 2000 y_{i}=\exp(\frac{-||\vec{x}-\vec{d}_{i}||^{2}}{2\delta^2})，i=1,2,...,2000yi​=exp(2δ2−∣∣x−di​∣∣2​)，i=1,2,...,2000

测试数据总共分成三类：
1）小型矩阵与向量计算，N = M = 20 N=M=20N=M=20；
2）中型矩阵与向量计算，N = M = 200 N=M=200N=M=200;
3) 大型矩阵与向量计算，N = M = 2000 N=M=2000N=M=2000。

2.2 测试程序

  • double型
<span style="color:#000000"><code class="language-cpp"><span style="color:#f1fa8c">#<span style="color:#ff79c6">include</span> <span style="color:#f1fa8c"><iostream></span></span>
<span style="color:#f1fa8c">#<span style="color:#ff79c6">include</span> <span style="color:#f1fa8c"><vector></span></span>
<span style="color:#f1fa8c">#<span style="color:#ff79c6">include</span> <span style="color:#f1fa8c"><fstream></span></span>
<span style="color:#f1fa8c">#<span style="color:#ff79c6">include</span> <span style="color:#f1fa8c"><cmath></span></span>
<span style="color:#f1fa8c">#<span style="color:#ff79c6">include</span> <span style="color:#f1fa8c"><sys/timeb.h></span></span>
<span style="color:#f1fa8c">#<span style="color:#ff79c6">include</span> <span style="color:#f1fa8c">"iomanip"</span></span>
<span style="color:#f1fa8c">#<span style="color:#ff79c6">include</span> <span style="color:#f1fa8c"><Eigen/Dense></span></span>

<span style="color:#f1fa8c">#<span style="color:#ff79c6">include</span> <span style="color:#f1fa8c">"NumCpp.hpp"</span></span>
<span style="color:#f1fa8c">#<span style="color:#ff79c6">include</span> <span style="color:#f1fa8c">"boost/filesystem.hpp"</span></span>

<span style="color:#ff79c6">using</span> <span style="color:#ff79c6">namespace</span> Eigen<span style="color:#999999">;</span>
<span style="color:#ff79c6">using</span> <span style="color:#ff79c6">namespace</span> std<span style="color:#999999">;</span>

<span style="color:#ff79c6">int</span> M <span style="color:#6272a4">=</span> <span style="color:#f1fa8c">2000</span><span style="color:#999999">;</span>  <span style="color:#6272a4">// 20 200 2000</span>
<span style="color:#ff79c6">int</span> N <span style="color:#6272a4">=</span> <span style="color:#f1fa8c">2000</span><span style="color:#999999">;</span>  <span style="color:#6272a4">// 20 200 2000</span>

Eigen<span style="color:#6272a4">::</span>RowVectorXd <span style="color:#8be9fd">construct_Phi2eigen</span><span style="color:#999999">(</span>Eigen<span style="color:#6272a4">::</span>Vector2d x<span style="color:#999999">,</span> Eigen<span style="color:#6272a4">::</span>Matrix2Xd ALD_mat<span style="color:#999999">)</span>
<span style="color:#999999">{</span>
	<span style="color:#ff79c6">double</span> delta <span style="color:#6272a4">=</span> <span style="color:#f1fa8c">1.2</span><span style="color:#999999">;</span>
	Eigen<span style="color:#6272a4">::</span>Matrix2Xd dmat <span style="color:#6272a4">=</span> ALD_mat<span style="color:#999999">.</span><span style="color:#8be9fd">colwise</span><span style="color:#999999">(</span><span style="color:#999999">)</span> <span style="color:#6272a4">-</span> x<span style="color:#999999">;</span>
	Eigen<span style="color:#6272a4">::</span>Matrix2Xd dmat2 <span style="color:#6272a4">=</span> dmat<span style="color:#999999">.</span><span style="color:#8be9fd">array</span><span style="color:#999999">(</span><span style="color:#999999">)</span><span style="color:#999999">.</span><span style="color:#8be9fd">pow</span><span style="color:#999999">(</span><span style="color:#f1fa8c">2</span><span style="color:#999999">)</span><span style="color:#999999">;</span>
	Eigen<span style="color:#6272a4">::</span>RowVectorXd dmat2sum <span style="color:#6272a4">=</span> <span style="color:#6272a4">-</span>dmat2<span style="color:#999999">.</span><span style="color:#8be9fd">colwise</span><span style="color:#999999">(</span><span style="color:#999999">)</span><span style="color:#999999">.</span><span style="color:#8be9fd">sum</span><span style="color:#999999">(</span><span style="color:#999999">)</span><span style="color:#6272a4">/</span><span style="color:#999999">(</span><span style="color:#f1fa8c">2.0</span><span style="color:#6272a4">*</span>delta<span style="color:#6272a4">*</span>delta<span style="color:#999999">)</span><span style="color:#999999">;</span>
	Eigen<span style="color:#6272a4">::</span>RowVectorXd y <span style="color:#6272a4">=</span> dmat2sum<span style="color:#999999">.</span><span style="color:#8be9fd">array</span><span style="color:#999999">(</span><span style="color:#999999">)</span><span style="color:#999999">.</span><span style="color:#8be9fd">exp</span><span style="color:#999999">(</span><span style="color:#999999">)</span><span style="color:#999999">;</span>
	<span style="color:#ff79c6">return</span> y<span style="color:#999999">;</span>
<span style="color:#999999">}</span>

nc<span style="color:#6272a4">::</span>NdArray<span style="color:#6272a4"><</span><span style="color:#ff79c6">double</span><span style="color:#6272a4">></span> <span style="color:#8be9fd">contruct_Phi2numcpp</span><span style="color:#999999">(</span>nc<span style="color:#6272a4">::</span>NdArray<span style="color:#6272a4"><</span><span style="color:#ff79c6">double</span><span style="color:#6272a4">></span> x<span style="color:#999999">,</span> nc<span style="color:#6272a4">::</span>NdArray<span style="color:#6272a4"><</span><span style="color:#ff79c6">double</span><span style="color:#6272a4">></span> ALD_buf<span style="color:#999999">)</span>
<span style="color:#999999">{</span>
	<span style="color:#ff79c6">double</span> delta <span style="color:#6272a4">=</span> <span style="color:#f1fa8c">1.2</span><span style="color:#999999">;</span>
	nc<span style="color:#6272a4">::</span>NdArray<span style="color:#6272a4"><</span><span style="color:#ff79c6">double</span><span style="color:#6272a4">></span> c <span style="color:#6272a4">=</span> ALD_buf <span style="color:#6272a4">-</span> nc<span style="color:#6272a4">::</span><span style="color:#8be9fd">tile</span><span style="color:#999999">(</span>x<span style="color:#999999">,</span> <span style="color:#999999">{</span>ALD_buf<span style="color:#999999">.</span><span style="color:#8be9fd">shape</span><span style="color:#999999">(</span><span style="color:#999999">)</span><span style="color:#999999">.</span>rows <span style="color:#999999">,</span> <span style="color:#f1fa8c">1</span><span style="color:#999999">}</span><span style="color:#999999">)</span><span style="color:#999999">;</span>
	nc<span style="color:#6272a4">::</span>NdArray<span style="color:#6272a4"><</span><span style="color:#ff79c6">double</span><span style="color:#6272a4">></span> e <span style="color:#6272a4">=</span> nc<span style="color:#6272a4">::</span><span style="color:#8be9fd">exp</span><span style="color:#999999">(</span><span style="color:#6272a4">-</span>nc<span style="color:#6272a4">::</span><span style="color:#8be9fd">sum</span><span style="color:#999999">(</span>nc<span style="color:#6272a4">::</span><span style="color:#8be9fd">power</span><span style="color:#999999">(</span>c<span style="color:#999999">,</span> <span style="color:#f1fa8c">2</span><span style="color:#999999">)</span><span style="color:#999999">,</span> nc<span style="color:#6272a4">::</span>Axis<span style="color:#6272a4">::</span>COL<span style="color:#999999">)</span><span style="color:#6272a4">/</span><span style="color:#999999">(</span><span style="color:#f1fa8c">2.0</span><span style="color:#6272a4">*</span>delta<span style="color:#6272a4">*</span>delta<span style="color:#999999">)</span><span style="color:#999999">)</span><span style="color:#999999">;</span>	
	<span style="color:#ff79c6">return</span> e<span style="color:#999999">;</span>
<span style="color:#999999">}</span>

<span style="color:#ff79c6">int</span> <span style="color:#8be9fd">main</span><span style="color:#999999">(</span><span style="color:#999999">)</span>
<span style="color:#999999">{</span>
	Eigen<span style="color:#6272a4">::</span>Matrix2Xd X_mat <span style="color:#6272a4">=</span> Eigen<span style="color:#6272a4">::</span>Matrix2Xd<span style="color:#6272a4">::</span><span style="color:#8be9fd">Random</span><span style="color:#999999">(</span>M<span style="color:#999999">,</span> N<span style="color:#999999">)</span><span style="color:#999999">;</span>
	Eigen<span style="color:#6272a4">::</span>Vector2d x <span style="color:#6272a4">=</span> Eigen<span style="color:#6272a4">::</span>Vector2d<span style="color:#6272a4">::</span><span style="color:#8be9fd">Random</span><span style="color:#999999">(</span>M<span style="color:#999999">,</span> <span style="color:#f1fa8c">1</span><span style="color:#999999">)</span><span style="color:#999999">;</span>	
	<span style="color:#ff79c6">struct</span> timeb startTime<span style="color:#999999">,</span> endTime<span style="color:#999999">;</span>
	
	std<span style="color:#6272a4">::</span>cout<span style="color:#6272a4"><<</span><span style="color:#f1fa8c">"test the time cost for Eigen with matrix"</span><span style="color:#6272a4"><<</span>std<span style="color:#6272a4">::</span>endl<span style="color:#999999">;</span>
	<span style="color:#8be9fd">ftime</span><span style="color:#999999">(</span><span style="color:#6272a4">&</span>startTime<span style="color:#999999">)</span><span style="color:#999999">;</span>  
	<span style="color:#ff79c6">for</span> <span style="color:#999999">(</span>size_t i <span style="color:#6272a4">=</span> <span style="color:#f1fa8c">0</span><span style="color:#999999">;</span> i <span style="color:#6272a4"><</span> <span style="color:#f1fa8c">10000</span><span style="color:#999999">;</span> i<span style="color:#6272a4">++</span><span style="color:#999999">)</span>
	<span style="color:#999999">{</span>
		<span style="color:#ff79c6">auto</span> phi2 <span style="color:#6272a4">=</span> <span style="color:#8be9fd">construct_Phi2eigen</span><span style="color:#999999">(</span>x<span style="color:#999999">,</span> X_mat<span style="color:#999999">)</span><span style="color:#999999">;</span>
	<span style="color:#999999">}</span>	
	<span style="color:#8be9fd">ftime</span><span style="color:#999999">(</span><span style="color:#6272a4">&</span>endTime<span style="color:#999999">)</span><span style="color:#999999">;</span>
	std<span style="color:#6272a4">::</span>cout <span style="color:#6272a4"><<</span> <span style="color:#f1fa8c">"time cost is："</span> <span style="color:#6272a4"><<</span> <span style="color:#999999">(</span>endTime<span style="color:#999999">.</span>time<span style="color:#6272a4">-</span>startTime<span style="color:#999999">.</span>time<span style="color:#999999">)</span><span style="color:#6272a4">*</span><span style="color:#f1fa8c">1000</span> <span style="color:#6272a4">+</span> <span style="color:#999999">(</span>endTime<span style="color:#999999">.</span>millitm <span style="color:#6272a4">-</span> startTime<span style="color:#999999">.</span>millitm<span style="color:#999999">)</span> <span style="color:#6272a4"><<</span> <span style="color:#f1fa8c">"ms"</span> <span style="color:#6272a4"><<</span> std<span style="color:#6272a4">::</span>endl<span style="color:#999999">;</span>


	
	nc<span style="color:#6272a4">::</span>NdArray<span style="color:#6272a4"><</span><span style="color:#ff79c6">double</span><span style="color:#6272a4">></span> D_arr <span style="color:#6272a4">=</span> nc<span style="color:#6272a4">::</span>random<span style="color:#6272a4">::</span>rand<span style="color:#6272a4"><</span><span style="color:#ff79c6">double</span><span style="color:#6272a4">></span><span style="color:#999999">(</span><span style="color:#999999">{</span>N<span style="color:#999999">,</span> M<span style="color:#999999">}</span><span style="color:#999999">)</span><span style="color:#999999">;</span>
	nc<span style="color:#6272a4">::</span>NdArray<span style="color:#6272a4"><</span><span style="color:#ff79c6">double</span><span style="color:#6272a4">></span> x1 <span style="color:#6272a4">=</span> nc<span style="color:#6272a4">::</span>random<span style="color:#6272a4">::</span>rand<span style="color:#6272a4"><</span><span style="color:#ff79c6">double</span><span style="color:#6272a4">></span><span style="color:#999999">(</span><span style="color:#999999">{</span><span style="color:#f1fa8c">1</span><span style="color:#999999">,</span> M<span style="color:#999999">}</span><span style="color:#999999">)</span><span style="color:#999999">;</span>


	std<span style="color:#6272a4">::</span>cout<span style="color:#6272a4"><<</span><span style="color:#f1fa8c">"test the time cost for numcpp"</span><span style="color:#6272a4"><<</span>std<span style="color:#6272a4">::</span>endl<span style="color:#999999">;</span>
	<span style="color:#8be9fd">ftime</span><span style="color:#999999">(</span><span style="color:#6272a4">&</span>startTime<span style="color:#999999">)</span><span style="color:#999999">;</span>  
	<span style="color:#ff79c6">for</span> <span style="color:#999999">(</span>size_t i <span style="color:#6272a4">=</span> <span style="color:#f1fa8c">0</span><span style="color:#999999">;</span> i <span style="color:#6272a4"><</span> <span style="color:#f1fa8c">10000</span><span style="color:#999999">;</span> i<span style="color:#6272a4">++</span><span style="color:#999999">)</span>
	<span style="color:#999999">{</span>
		<span style="color:#ff79c6">auto</span> phi3 <span style="color:#6272a4">=</span> <span style="color:#8be9fd">contruct_Phi2numcpp</span><span style="color:#999999">(</span>x1<span style="color:#999999">,</span> D_arr<span style="color:#999999">)</span><span style="color:#999999">;</span>
		<span style="color:#6272a4">// std::cout<<phi3<<std::endl;</span>
	<span style="color:#999999">}</span>
	<span style="color:#8be9fd">ftime</span><span style="color:#999999">(</span><span style="color:#6272a4">&</span>endTime<span style="color:#999999">)</span><span style="color:#999999">;</span>
	std<span style="color:#6272a4">::</span>cout <span style="color:#6272a4"><<</span> <span style="color:#f1fa8c">"time cost is："</span> <span style="color:#6272a4"><<</span> <span style="color:#999999">(</span>endTime<span style="color:#999999">.</span>time<span style="color:#6272a4">-</span>startTime<span style="color:#999999">.</span>time<span style="color:#999999">)</span><span style="color:#6272a4">*</span><span style="color:#f1fa8c">1000</span> <span style="color:#6272a4">+</span> <span style="color:#999999">(</span>endTime<span style="color:#999999">.</span>millitm <span style="color:#6272a4">-</span> startTime<span style="color:#999999">.</span>millitm<span style="color:#999999">)</span> <span style="color:#6272a4"><<</span> <span style="color:#f1fa8c">"ms"</span> <span style="color:#6272a4"><<</span> std<span style="color:#6272a4">::</span>endl<span style="color:#999999">;</span>
<span style="color:#999999">}</span>
</code></span>
  • 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
  • float型
<span style="color:#000000"><code class="language-cpp"><span style="color:#f1fa8c">#<span style="color:#ff79c6">include</span> <span style="color:#f1fa8c"><iostream></span></span>
<span style="color:#f1fa8c">#<span style="color:#ff79c6">include</span> <span style="color:#f1fa8c"><vector></span></span>
<span style="color:#f1fa8c">#<span style="color:#ff79c6">include</span> <span style="color:#f1fa8c"><fstream></span></span>
<span style="color:#f1fa8c">#<span style="color:#ff79c6">include</span> <span style="color:#f1fa8c"><cmath></span></span>
<span style="color:#f1fa8c">#<span style="color:#ff79c6">include</span> <span style="color:#f1fa8c"><sys/timeb.h></span></span>
<span style="color:#f1fa8c">#<span style="color:#ff79c6">include</span><span style="color:#f1fa8c">"iomanip"</span></span>
<span style="color:#f1fa8c">#<span style="color:#ff79c6">include</span> <span style="color:#f1fa8c"><Eigen/Dense></span></span>

<span style="color:#f1fa8c">#<span style="color:#ff79c6">include</span><span style="color:#f1fa8c">"NumCpp.hpp"</span></span>
<span style="color:#f1fa8c">#<span style="color:#ff79c6">include</span><span style="color:#f1fa8c">"boost/filesystem.hpp"</span></span>

<span style="color:#ff79c6">using</span> <span style="color:#ff79c6">namespace</span> Eigen<span style="color:#999999">;</span>
<span style="color:#ff79c6">using</span> <span style="color:#ff79c6">namespace</span> std<span style="color:#999999">;</span>

<span style="color:#ff79c6">int</span> M <span style="color:#6272a4">=</span> <span style="color:#f1fa8c">2</span><span style="color:#999999">;</span>  <span style="color:#6272a4">// 2 20 200</span>
<span style="color:#ff79c6">int</span> N <span style="color:#6272a4">=</span> <span style="color:#f1fa8c">2</span><span style="color:#999999">;</span>  <span style="color:#6272a4">// 2 20 200</span>

Eigen<span style="color:#6272a4">::</span>RowVectorXf <span style="color:#8be9fd">construct_Phi2eigen</span><span style="color:#999999">(</span>Eigen<span style="color:#6272a4">::</span>VectorXf x<span style="color:#999999">,</span> Eigen<span style="color:#6272a4">::</span>MatrixXf ALD_mat<span style="color:#999999">)</span>
<span style="color:#999999">{</span>
	<span style="color:#ff79c6">float</span> delta <span style="color:#6272a4">=</span> <span style="color:#f1fa8c">1.2</span><span style="color:#999999">;</span>
	Eigen<span style="color:#6272a4">::</span>MatrixXf dmat <span style="color:#6272a4">=</span> ALD_mat<span style="color:#999999">.</span><span style="color:#8be9fd">colwise</span><span style="color:#999999">(</span><span style="color:#999999">)</span> <span style="color:#6272a4">-</span> x<span style="color:#999999">;</span>
	Eigen<span style="color:#6272a4">::</span>MatrixXf dmat2 <span style="color:#6272a4">=</span> dmat<span style="color:#999999">.</span><span style="color:#8be9fd">array</span><span style="color:#999999">(</span><span style="color:#999999">)</span><span style="color:#999999">.</span><span style="color:#8be9fd">pow</span><span style="color:#999999">(</span><span style="color:#f1fa8c">2</span><span style="color:#999999">)</span><span style="color:#999999">;</span>
	Eigen<span style="color:#6272a4">::</span>RowVectorXf dmat2sum <span style="color:#6272a4">=</span> <span style="color:#6272a4">-</span>dmat2<span style="color:#999999">.</span><span style="color:#8be9fd">colwise</span><span style="color:#999999">(</span><span style="color:#999999">)</span><span style="color:#999999">.</span><span style="color:#8be9fd">sum</span><span style="color:#999999">(</span><span style="color:#999999">)</span><span style="color:#6272a4">/</span><span style="color:#999999">(</span><span style="color:#f1fa8c">2.0f</span><span style="color:#6272a4">*</span>delta<span style="color:#6272a4">*</span>delta<span style="color:#999999">)</span><span style="color:#999999">;</span>
	Eigen<span style="color:#6272a4">::</span>RowVectorXf y <span style="color:#6272a4">=</span> dmat2sum<span style="color:#999999">.</span><span style="color:#8be9fd">array</span><span style="color:#999999">(</span><span style="color:#999999">)</span><span style="color:#999999">.</span><span style="color:#8be9fd">exp</span><span style="color:#999999">(</span><span style="color:#999999">)</span><span style="color:#999999">;</span>
	<span style="color:#ff79c6">return</span> y<span style="color:#999999">;</span>
<span style="color:#999999">}</span>

nc<span style="color:#6272a4">::</span>NdArray<span style="color:#6272a4"><</span><span style="color:#ff79c6">float</span><span style="color:#6272a4">></span> <span style="color:#8be9fd">contruct_Phi2numcpp</span><span style="color:#999999">(</span>nc<span style="color:#6272a4">::</span>NdArray<span style="color:#6272a4"><</span><span style="color:#ff79c6">float</span><span style="color:#6272a4">></span> x<span style="color:#999999">,</span> nc<span style="color:#6272a4">::</span>NdArray<span style="color:#6272a4"><</span><span style="color:#ff79c6">float</span><span style="color:#6272a4">></span> ALD_buf<span style="color:#999999">)</span>
<span style="color:#999999">{</span>
	<span style="color:#ff79c6">float</span> delta <span style="color:#6272a4">=</span> <span style="color:#f1fa8c">1.2</span><span style="color:#999999">;</span>
	nc<span style="color:#6272a4">::</span>NdArray<span style="color:#6272a4"><</span><span style="color:#ff79c6">float</span><span style="color:#6272a4">></span> c <span style="color:#6272a4">=</span> ALD_buf <span style="color:#6272a4">-</span> nc<span style="color:#6272a4">::</span><span style="color:#8be9fd">tile</span><span style="color:#999999">(</span>x<span style="color:#999999">,</span> <span style="color:#999999">{</span>ALD_buf<span style="color:#999999">.</span><span style="color:#8be9fd">shape</span><span style="color:#999999">(</span><span style="color:#999999">)</span><span style="color:#999999">.</span>rows <span style="color:#999999">,</span> <span style="color:#f1fa8c">1</span><span style="color:#999999">}</span><span style="color:#999999">)</span><span style="color:#999999">;</span>
	nc<span style="color:#6272a4">::</span>NdArray<span style="color:#6272a4"><</span><span style="color:#ff79c6">float</span><span style="color:#6272a4">></span> e <span style="color:#6272a4">=</span> nc<span style="color:#6272a4">::</span><span style="color:#8be9fd">exp</span><span style="color:#999999">(</span><span style="color:#6272a4">-</span>nc<span style="color:#6272a4">::</span><span style="color:#8be9fd">sum</span><span style="color:#999999">(</span>nc<span style="color:#6272a4">::</span><span style="color:#8be9fd">power</span><span style="color:#999999">(</span>c<span style="color:#999999">,</span> <span style="color:#f1fa8c">2</span><span style="color:#999999">)</span><span style="color:#999999">,</span> nc<span style="color:#6272a4">::</span>Axis<span style="color:#6272a4">::</span>COL<span style="color:#999999">)</span><span style="color:#6272a4">/</span><span style="color:#999999">(</span><span style="color:#f1fa8c">2.0f</span><span style="color:#6272a4">*</span>delta<span style="color:#6272a4">*</span>delta<span style="color:#999999">)</span><span style="color:#999999">)</span><span style="color:#999999">;</span>	
	<span style="color:#ff79c6">return</span> e<span style="color:#999999">;</span>
<span style="color:#999999">}</span>

<span style="color:#ff79c6">int</span> <span style="color:#8be9fd">main</span><span style="color:#999999">(</span><span style="color:#999999">)</span>
<span style="color:#999999">{</span>
	Eigen<span style="color:#6272a4">::</span>MatrixXf X_mat <span style="color:#6272a4">=</span> Eigen<span style="color:#6272a4">::</span>MatrixXf<span style="color:#6272a4">::</span><span style="color:#8be9fd">Random</span><span style="color:#999999">(</span>M<span style="color:#999999">,</span> N<span style="color:#999999">)</span><span style="color:#999999">;</span>
	Eigen<span style="color:#6272a4">::</span>VectorXf x <span style="color:#6272a4">=</span> Eigen<span style="color:#6272a4">::</span>VectorXf<span style="color:#6272a4">::</span><span style="color:#8be9fd">Random</span><span style="color:#999999">(</span>M<span style="color:#999999">,</span> <span style="color:#f1fa8c">1</span><span style="color:#999999">)</span><span style="color:#999999">;</span>

	<span style="color:#ff79c6">struct</span> timeb startTime<span style="color:#999999">,</span> endTime<span style="color:#999999">;</span>

	std<span style="color:#6272a4">::</span>cout<span style="color:#6272a4"><<</span><span style="color:#f1fa8c">"test the time cost for Eigen with matrix"</span><span style="color:#6272a4"><<</span>std<span style="color:#6272a4">::</span>endl<span style="color:#999999">;</span>
	<span style="color:#8be9fd">ftime</span><span style="color:#999999">(</span><span style="color:#6272a4">&</span>startTime<span style="color:#999999">)</span><span style="color:#999999">;</span>  
	<span style="color:#ff79c6">for</span> <span style="color:#999999">(</span>size_t i <span style="color:#6272a4">=</span> <span style="color:#f1fa8c">0</span><span style="color:#999999">;</span> i <span style="color:#6272a4"><</span> <span style="color:#f1fa8c">10000</span><span style="color:#999999">;</span> i<span style="color:#6272a4">++</span><span style="color:#999999">)</span>
	<span style="color:#999999">{</span>
		<span style="color:#ff79c6">auto</span> phi2 <span style="color:#6272a4">=</span> <span style="color:#8be9fd">construct_Phi2eigen</span><span style="color:#999999">(</span>x<span style="color:#999999">,</span> X_mat<span style="color:#999999">)</span><span style="color:#999999">;</span>
	<span style="color:#999999">}</span>	
	<span style="color:#8be9fd">ftime</span><span style="color:#999999">(</span><span style="color:#6272a4">&</span>endTime<span style="color:#999999">)</span><span style="color:#999999">;</span>
	std<span style="color:#6272a4">::</span>cout <span style="color:#6272a4"><<</span> <span style="color:#f1fa8c">"time cost is："</span> <span style="color:#6272a4"><<</span> <span style="color:#999999">(</span>endTime<span style="color:#999999">.</span>time<span style="color:#6272a4">-</span>startTime<span style="color:#999999">.</span>time<span style="color:#999999">)</span><span style="color:#6272a4">*</span><span style="color:#f1fa8c">1000</span> <span style="color:#6272a4">+</span> <span style="color:#999999">(</span>endTime<span style="color:#999999">.</span>millitm <span style="color:#6272a4">-</span> startTime<span style="color:#999999">.</span>millitm<span style="color:#999999">)</span> <span style="color:#6272a4"><<</span> <span style="color:#f1fa8c">"ms"</span> <span style="color:#6272a4"><<</span> std<span style="color:#6272a4">::</span>endl<span style="color:#999999">;</span>


	nc<span style="color:#6272a4">::</span>NdArray<span style="color:#6272a4"><</span><span style="color:#ff79c6">float</span><span style="color:#6272a4">></span> x1 <span style="color:#6272a4">=</span> nc<span style="color:#6272a4">::</span>random<span style="color:#6272a4">::</span>rand<span style="color:#6272a4"><</span><span style="color:#ff79c6">float</span><span style="color:#6272a4">></span><span style="color:#999999">(</span><span style="color:#999999">{</span><span style="color:#f1fa8c">1</span><span style="color:#999999">,</span> M<span style="color:#999999">}</span><span style="color:#999999">)</span><span style="color:#999999">;</span>
	nc<span style="color:#6272a4">::</span>NdArray<span style="color:#6272a4"><</span><span style="color:#ff79c6">float</span><span style="color:#6272a4">></span> D_arr <span style="color:#6272a4">=</span> nc<span style="color:#6272a4">::</span>random<span style="color:#6272a4">::</span>rand<span style="color:#6272a4"><</span><span style="color:#ff79c6">float</span><span style="color:#6272a4">></span><span style="color:#999999">(</span><span style="color:#999999">{</span>N<span style="color:#999999">,</span> M<span style="color:#999999">}</span><span style="color:#999999">)</span><span style="color:#999999">;</span>


	std<span style="color:#6272a4">::</span>cout<span style="color:#6272a4"><<</span><span style="color:#f1fa8c">"test the time cost for numcpp"</span><span style="color:#6272a4"><<</span>std<span style="color:#6272a4">::</span>endl<span style="color:#999999">;</span>
	<span style="color:#8be9fd">ftime</span><span style="color:#999999">(</span><span style="color:#6272a4">&</span>startTime<span style="color:#999999">)</span><span style="color:#999999">;</span>  
	<span style="color:#ff79c6">for</span> <span style="color:#999999">(</span>size_t i <span style="color:#6272a4">=</span> <span style="color:#f1fa8c">0</span><span style="color:#999999">;</span> i <span style="color:#6272a4"><</span> <span style="color:#f1fa8c">10000</span><span style="color:#999999">;</span> i<span style="color:#6272a4">++</span><span style="color:#999999">)</span>
	<span style="color:#999999">{</span>
		<span style="color:#ff79c6">auto</span> phi3 <span style="color:#6272a4">=</span> <span style="color:#8be9fd">contruct_Phi2numcpp</span><span style="color:#999999">(</span>x1<span style="color:#999999">,</span> D_arr<span style="color:#999999">)</span><span style="color:#999999">;</span>
		<span style="color:#6272a4">// std::cout<<phi3<<std::endl;</span>
	<span style="color:#999999">}</span>
	<span style="color:#8be9fd">ftime</span><span style="color:#999999">(</span><span style="color:#6272a4">&</span>endTime<span style="color:#999999">)</span><span style="color:#999999">;</span>
	std<span style="color:#6272a4">::</span>cout <span style="color:#6272a4"><<</span> <span style="color:#f1fa8c">"time cost is："</span> <span style="color:#6272a4"><<</span> <span style="color:#999999">(</span>endTime<span style="color:#999999">.</span>time<span style="color:#6272a4">-</span>startTime<span style="color:#999999">.</span>time<span style="color:#999999">)</span><span style="color:#6272a4">*</span><span style="color:#f1fa8c">1000</span> <span style="color:#6272a4">+</span> <span style="color:#999999">(</span>endTime<span style="color:#999999">.</span>millitm <span style="color:#6272a4">-</span> startTime<span style="color:#999999">.</span>millitm<span style="color:#999999">)</span> <span style="color:#6272a4"><<</span> <span style="color:#f1fa8c">"ms"</span> <span style="color:#6272a4"><<</span> std<span style="color:#6272a4">::</span>endl<span style="color:#999999">;</span>
<span style="color:#999999">}</span>
</code></span>
  • 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
  • python for numpy and cupy
<span style="color:#000000"><code class="language-python"><span style="color:#ff79c6">import</span> cupy <span style="color:#ff79c6">as</span> cp
<span style="color:#ff79c6">import</span> numpy <span style="color:#ff79c6">as</span> np
<span style="color:#ff79c6">import</span> time

delta <span style="color:#6272a4">=</span> <span style="color:#f1fa8c">1.2</span>

M <span style="color:#6272a4">=</span> <span style="color:#f1fa8c">2</span> <span style="color:#6272a4"># 2 20 200</span>
N <span style="color:#6272a4">=</span> <span style="color:#f1fa8c">2</span> <span style="color:#6272a4"># 2 20 200</span>

x_cp <span style="color:#6272a4">=</span> cp<span style="color:#999999">.</span>random<span style="color:#999999">.</span>rand<span style="color:#999999">(</span><span style="color:#f1fa8c">1</span><span style="color:#999999">,</span>M<span style="color:#999999">)</span>
D_arr_cp <span style="color:#6272a4">=</span> cp<span style="color:#999999">.</span>random<span style="color:#999999">.</span>rand<span style="color:#999999">(</span>N<span style="color:#999999">,</span> M<span style="color:#999999">)</span>

x_np <span style="color:#6272a4">=</span> np<span style="color:#999999">.</span>random<span style="color:#999999">.</span>rand<span style="color:#999999">(</span><span style="color:#f1fa8c">1</span><span style="color:#999999">,</span>M<span style="color:#999999">)</span>
D_arr_np <span style="color:#6272a4">=</span> np<span style="color:#999999">.</span>random<span style="color:#999999">.</span>rand<span style="color:#999999">(</span>N<span style="color:#999999">,</span> M<span style="color:#999999">)</span>

<span style="color:#ff79c6">def</span> <span style="color:#8be9fd">construct_Phi2_cp</span><span style="color:#999999">(</span>x<span style="color:#999999">,</span> ALD_buf<span style="color:#999999">)</span><span style="color:#999999">:</span>
    ALD_arr <span style="color:#6272a4">=</span> cp<span style="color:#999999">.</span>array<span style="color:#999999">(</span>ALD_buf<span style="color:#999999">)</span>
    x <span style="color:#6272a4">=</span> cp<span style="color:#999999">.</span>tile<span style="color:#999999">(</span>x<span style="color:#999999">,</span> <span style="color:#999999">(</span><span style="color:#f1fa8c">len</span><span style="color:#999999">(</span>ALD_buf<span style="color:#999999">)</span><span style="color:#999999">,</span> <span style="color:#f1fa8c">1</span><span style="color:#999999">)</span><span style="color:#999999">)</span>
    dX <span style="color:#6272a4">=</span> ALD_arr <span style="color:#6272a4">-</span> x
    <span style="color:#ff79c6">return</span> cp<span style="color:#999999">.</span>exp<span style="color:#999999">(</span><span style="color:#6272a4">-</span>cp<span style="color:#999999">.</span><span style="color:#f1fa8c">sum</span><span style="color:#999999">(</span>cp<span style="color:#999999">.</span>power<span style="color:#999999">(</span>dX<span style="color:#999999">,</span> <span style="color:#f1fa8c">2</span><span style="color:#999999">)</span><span style="color:#999999">,</span> axis<span style="color:#6272a4">=</span><span style="color:#f1fa8c">1</span><span style="color:#999999">)</span><span style="color:#6272a4">/</span><span style="color:#f1fa8c">2</span><span style="color:#6272a4">*</span>delta<span style="color:#6272a4">*</span>delta<span style="color:#999999">)</span>
        
<span style="color:#ff79c6">def</span> <span style="color:#8be9fd">construct_Phi2_np</span><span style="color:#999999">(</span>x<span style="color:#999999">,</span> ALD_buf<span style="color:#999999">)</span><span style="color:#999999">:</span>
    ALD_arr <span style="color:#6272a4">=</span> np<span style="color:#999999">.</span>array<span style="color:#999999">(</span>ALD_buf<span style="color:#999999">)</span>
    x <span style="color:#6272a4">=</span> np<span style="color:#999999">.</span>tile<span style="color:#999999">(</span>x<span style="color:#999999">,</span> <span style="color:#999999">(</span><span style="color:#f1fa8c">len</span><span style="color:#999999">(</span>ALD_buf<span style="color:#999999">)</span><span style="color:#999999">,</span> <span style="color:#f1fa8c">1</span><span style="color:#999999">)</span><span style="color:#999999">)</span>
    dX <span style="color:#6272a4">=</span> ALD_arr <span style="color:#6272a4">-</span> x
    <span style="color:#ff79c6">return</span> np<span style="color:#999999">.</span>exp<span style="color:#999999">(</span><span style="color:#6272a4">-</span>np<span style="color:#999999">.</span><span style="color:#f1fa8c">sum</span><span style="color:#999999">(</span>np<span style="color:#999999">.</span>power<span style="color:#999999">(</span>dX<span style="color:#999999">,</span> <span style="color:#f1fa8c">2</span><span style="color:#999999">)</span><span style="color:#999999">,</span> axis<span style="color:#6272a4">=</span><span style="color:#f1fa8c">1</span><span style="color:#999999">)</span><span style="color:#6272a4">/</span><span style="color:#f1fa8c">2</span><span style="color:#6272a4">*</span>delta<span style="color:#6272a4">*</span>delta<span style="color:#999999">)</span>
    
    
<span style="color:#ff79c6">def</span> <span style="color:#8be9fd">testtimecp</span><span style="color:#999999">(</span><span style="color:#999999">)</span><span style="color:#999999">:</span>
    <span style="color:#ff79c6">for</span> i <span style="color:#ff79c6">in</span> <span style="color:#f1fa8c">range</span><span style="color:#999999">(</span><span style="color:#f1fa8c">1000</span><span style="color:#999999">)</span><span style="color:#999999">:</span>
        construct_Phi2_cp<span style="color:#999999">(</span>x_cp<span style="color:#999999">,</span> D_arr_cp<span style="color:#999999">)</span>
        
        
<span style="color:#ff79c6">def</span> <span style="color:#8be9fd">testtimenp</span><span style="color:#999999">(</span><span style="color:#999999">)</span><span style="color:#999999">:</span>
    <span style="color:#ff79c6">for</span> i <span style="color:#ff79c6">in</span> <span style="color:#f1fa8c">range</span><span style="color:#999999">(</span><span style="color:#f1fa8c">1000</span><span style="color:#999999">)</span><span style="color:#999999">:</span>
        construct_Phi2_np<span style="color:#999999">(</span>x_np<span style="color:#999999">,</span> D_arr_np<span style="color:#999999">)</span>
        
<span style="color:#ff79c6">print</span><span style="color:#999999">(</span><span style="color:#f1fa8c">"time cost for cupy:"</span><span style="color:#999999">)</span>        
<span style="color:#6272a4">%</span>time testtimecp<span style="color:#999999">(</span><span style="color:#999999">)</span>

<span style="color:#ff79c6">print</span><span style="color:#999999">(</span><span style="color:#f1fa8c">"time cost for numpy"</span><span style="color:#999999">)</span>
<span style="color:#6272a4">%</span>time testtimenp<span style="color:#999999">(</span><span style="color:#999999">)</span>
</code></span>
  • 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

2.3、测试结果

矩阵大小M=2, N=2M=20, N=20M=200, N=200M=200, N=400M=400, N=200
Eigenf(6ms)、d(6ms)f(22ms)、d(30ms)f(685ms)、d(2.283s)f(2.222s)、d(4.845s))f(1.567s)、d(4.843s)
NumCppf(10ms)、d(11ms)f(38ms)、d(32ms)f(1.51s)、d(1.854)f(3.035s)、d(3.619s)f(3.255s)、d(7.278s)
NumPy195ms234ms3.25s6.32s6.32s
CuPy1.83s1.75s1.75s1.71s1.81s

三、动态矩阵计算测试

3.1 测试程序

  • float cpp
<span style="color:#000000"><code class="language-cpp"><span style="color:#f1fa8c">#<span style="color:#ff79c6">include</span> <span style="color:#f1fa8c"><iostream></span></span>
<span style="color:#f1fa8c">#<span style="color:#ff79c6">include</span> <span style="color:#f1fa8c"><vector></span></span>
<span style="color:#f1fa8c">#<span style="color:#ff79c6">include</span> <span style="color:#f1fa8c"><fstream></span></span>
<span style="color:#f1fa8c">#<span style="color:#ff79c6">include</span> <span style="color:#f1fa8c"><cmath></span></span>
<span style="color:#f1fa8c">#<span style="color:#ff79c6">include</span> <span style="color:#f1fa8c"><sys/timeb.h></span></span>
<span style="color:#f1fa8c">#<span style="color:#ff79c6">include</span><span style="color:#f1fa8c">"iomanip"</span></span>
<span style="color:#f1fa8c">#<span style="color:#ff79c6">include</span> <span style="color:#f1fa8c"><Eigen/Dense></span></span>

<span style="color:#f1fa8c">#<span style="color:#ff79c6">include</span><span style="color:#f1fa8c">"NumCpp.hpp"</span></span>
<span style="color:#f1fa8c">#<span style="color:#ff79c6">include</span><span style="color:#f1fa8c">"boost/filesystem.hpp"</span></span>

<span style="color:#ff79c6">using</span> <span style="color:#ff79c6">namespace</span> Eigen<span style="color:#999999">;</span>
<span style="color:#ff79c6">using</span> <span style="color:#ff79c6">namespace</span> std<span style="color:#999999">;</span>

<span style="color:#ff79c6">int</span> M <span style="color:#6272a4">=</span> <span style="color:#f1fa8c">200</span><span style="color:#999999">;</span> <span style="color:#6272a4">// 2 20 200</span>
<span style="color:#ff79c6">int</span> N <span style="color:#6272a4">=</span> <span style="color:#f1fa8c">200</span><span style="color:#999999">;</span> <span style="color:#6272a4">// 2 20 200</span>

<span style="color:#ff79c6">int</span> <span style="color:#8be9fd">main</span><span style="color:#999999">(</span><span style="color:#999999">)</span>
<span style="color:#999999">{</span>
	Eigen<span style="color:#6272a4">::</span>Matrix2Xf X_mat <span style="color:#6272a4">=</span> Eigen<span style="color:#6272a4">::</span>Matrix2Xf<span style="color:#6272a4">::</span><span style="color:#8be9fd">Random</span><span style="color:#999999">(</span>M<span style="color:#999999">,</span> N<span style="color:#999999">)</span><span style="color:#999999">;</span>
	Eigen<span style="color:#6272a4">::</span>Vector2f x <span style="color:#6272a4">=</span> Eigen<span style="color:#6272a4">::</span>Vector2f<span style="color:#6272a4">::</span><span style="color:#8be9fd">Random</span><span style="color:#999999">(</span>M<span style="color:#999999">,</span> <span style="color:#f1fa8c">1</span><span style="color:#999999">)</span><span style="color:#999999">;</span>


	std<span style="color:#6272a4">::</span>cout<span style="color:#6272a4"><<</span><span style="color:#f1fa8c">"test the time cost for Eigen with matrix"</span><span style="color:#6272a4"><<</span>std<span style="color:#6272a4">::</span>endl<span style="color:#999999">;</span>
	<span style="color:#ff79c6">struct</span> timeb startTime<span style="color:#999999">,</span> endTime<span style="color:#999999">;</span>
	<span style="color:#8be9fd">ftime</span><span style="color:#999999">(</span><span style="color:#6272a4">&</span>startTime<span style="color:#999999">)</span><span style="color:#999999">;</span>  
	<span style="color:#ff79c6">for</span> <span style="color:#999999">(</span>size_t i <span style="color:#6272a4">=</span> <span style="color:#f1fa8c">0</span><span style="color:#999999">;</span> i <span style="color:#6272a4"><</span> <span style="color:#f1fa8c">10000</span><span style="color:#999999">;</span> i<span style="color:#6272a4">++</span><span style="color:#999999">)</span>
	<span style="color:#999999">{</span>
		X_mat <span style="color:#6272a4">=</span> <span style="color:#999999">(</span>Eigen<span style="color:#6272a4">::</span><span style="color:#8be9fd">MatrixXf</span><span style="color:#999999">(</span>X_mat<span style="color:#999999">.</span><span style="color:#8be9fd">rows</span><span style="color:#999999">(</span><span style="color:#999999">)</span><span style="color:#999999">,</span> X_mat<span style="color:#999999">.</span><span style="color:#8be9fd">cols</span><span style="color:#999999">(</span><span style="color:#999999">)</span><span style="color:#6272a4">+</span>x<span style="color:#999999">.</span><span style="color:#8be9fd">cols</span><span style="color:#999999">(</span><span style="color:#999999">)</span><span style="color:#999999">)</span><span style="color:#6272a4"><<</span>X_mat<span style="color:#999999">,</span> x<span style="color:#999999">)</span><span style="color:#999999">.</span><span style="color:#8be9fd">finished</span><span style="color:#999999">(</span><span style="color:#999999">)</span><span style="color:#999999">;</span>
	<span style="color:#999999">}</span>	
	<span style="color:#8be9fd">ftime</span><span style="color:#999999">(</span><span style="color:#6272a4">&</span>endTime<span style="color:#999999">)</span><span style="color:#999999">;</span>
	std<span style="color:#6272a4">::</span>cout <span style="color:#6272a4"><<</span> <span style="color:#f1fa8c">"time cost is："</span> <span style="color:#6272a4"><<</span> <span style="color:#999999">(</span>endTime<span style="color:#999999">.</span>time<span style="color:#6272a4">-</span>startTime<span style="color:#999999">.</span>time<span style="color:#999999">)</span><span style="color:#6272a4">*</span><span style="color:#f1fa8c">1000</span> <span style="color:#6272a4">+</span> <span style="color:#999999">(</span>endTime<span style="color:#999999">.</span>millitm <span style="color:#6272a4">-</span> startTime<span style="color:#999999">.</span>millitm<span style="color:#999999">)</span> <span style="color:#6272a4"><<</span> <span style="color:#f1fa8c">"ms"</span> <span style="color:#6272a4"><<</span> std<span style="color:#6272a4">::</span>endl<span style="color:#999999">;</span>


	nc<span style="color:#6272a4">::</span>NdArray<span style="color:#6272a4"><</span><span style="color:#ff79c6">float</span><span style="color:#6272a4">></span> x1 <span style="color:#6272a4">=</span> nc<span style="color:#6272a4">::</span>random<span style="color:#6272a4">::</span>rand<span style="color:#6272a4"><</span><span style="color:#ff79c6">float</span><span style="color:#6272a4">></span><span style="color:#999999">(</span><span style="color:#999999">{</span>M<span style="color:#999999">,</span> <span style="color:#f1fa8c">1</span><span style="color:#999999">}</span><span style="color:#999999">)</span><span style="color:#999999">;</span>
	nc<span style="color:#6272a4">::</span>NdArray<span style="color:#6272a4"><</span><span style="color:#ff79c6">float</span><span style="color:#6272a4">></span> D_arr <span style="color:#6272a4">=</span> nc<span style="color:#6272a4">::</span>random<span style="color:#6272a4">::</span>rand<span style="color:#6272a4"><</span><span style="color:#ff79c6">float</span><span style="color:#6272a4">></span><span style="color:#999999">(</span><span style="color:#999999">{</span>M<span style="color:#999999">,</span> N<span style="color:#999999">}</span><span style="color:#999999">)</span><span style="color:#999999">;</span>


	std<span style="color:#6272a4">::</span>cout<span style="color:#6272a4"><<</span><span style="color:#f1fa8c">"test the time cost for numcpp"</span><span style="color:#6272a4"><<</span>std<span style="color:#6272a4">::</span>endl<span style="color:#999999">;</span>
	<span style="color:#8be9fd">ftime</span><span style="color:#999999">(</span><span style="color:#6272a4">&</span>startTime<span style="color:#999999">)</span><span style="color:#999999">;</span>  
	<span style="color:#ff79c6">for</span> <span style="color:#999999">(</span>size_t i <span style="color:#6272a4">=</span> <span style="color:#f1fa8c">0</span><span style="color:#999999">;</span> i <span style="color:#6272a4"><</span> <span style="color:#f1fa8c">10000</span><span style="color:#999999">;</span> i<span style="color:#6272a4">++</span><span style="color:#999999">)</span>
	<span style="color:#999999">{</span>
		D_arr <span style="color:#6272a4">=</span> nc<span style="color:#6272a4">::</span><span style="color:#8be9fd">column_stack</span><span style="color:#999999">(</span><span style="color:#999999">{</span>D_arr<span style="color:#999999">,</span> x1<span style="color:#999999">}</span><span style="color:#999999">)</span><span style="color:#999999">;</span>
		<span style="color:#6272a4">// std::cout<<phi3<<std::endl;</span>
	<span style="color:#999999">}</span>
	<span style="color:#8be9fd">ftime</span><span style="color:#999999">(</span><span style="color:#6272a4">&</span>endTime<span style="color:#999999">)</span><span style="color:#999999">;</span>
	std<span style="color:#6272a4">::</span>cout <span style="color:#6272a4"><<</span> <span style="color:#f1fa8c">"time cost is："</span> <span style="color:#6272a4"><<</span> <span style="color:#999999">(</span>endTime<span style="color:#999999">.</span>time<span style="color:#6272a4">-</span>startTime<span style="color:#999999">.</span>time<span style="color:#999999">)</span><span style="color:#6272a4">*</span><span style="color:#f1fa8c">1000</span> <span style="color:#6272a4">+</span> <span style="color:#999999">(</span>endTime<span style="color:#999999">.</span>millitm <span style="color:#6272a4">-</span> startTime<span style="color:#999999">.</span>millitm<span style="color:#999999">)</span> <span style="color:#6272a4"><<</span> <span style="color:#f1fa8c">"ms"</span> <span style="color:#6272a4"><<</span> std<span style="color:#6272a4">::</span>endl<span style="color:#999999">;</span>
<span style="color:#999999">}</span>
</code></span>
  • 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
  • double cpp
<span style="color:#000000"><code class="language-cpp"><span style="color:#f1fa8c">#<span style="color:#ff79c6">include</span> <span style="color:#f1fa8c"><iostream></span></span>
<span style="color:#f1fa8c">#<span style="color:#ff79c6">include</span> <span style="color:#f1fa8c"><vector></span></span>
<span style="color:#f1fa8c">#<span style="color:#ff79c6">include</span> <span style="color:#f1fa8c"><fstream></span></span>
<span style="color:#f1fa8c">#<span style="color:#ff79c6">include</span> <span style="color:#f1fa8c"><cmath></span></span>
<span style="color:#f1fa8c">#<span style="color:#ff79c6">include</span> <span style="color:#f1fa8c"><sys/timeb.h></span></span>
<span style="color:#f1fa8c">#<span style="color:#ff79c6">include</span><span style="color:#f1fa8c">"iomanip"</span></span>
<span style="color:#f1fa8c">#<span style="color:#ff79c6">include</span> <span style="color:#f1fa8c"><Eigen/Dense></span></span>

<span style="color:#f1fa8c">#<span style="color:#ff79c6">include</span><span style="color:#f1fa8c">"NumCpp.hpp"</span></span>
<span style="color:#f1fa8c">#<span style="color:#ff79c6">include</span><span style="color:#f1fa8c">"boost/filesystem.hpp"</span></span>

<span style="color:#ff79c6">using</span> <span style="color:#ff79c6">namespace</span> Eigen<span style="color:#999999">;</span>
<span style="color:#ff79c6">using</span> <span style="color:#ff79c6">namespace</span> std<span style="color:#999999">;</span>

<span style="color:#ff79c6">int</span> M <span style="color:#6272a4">=</span> <span style="color:#f1fa8c">200</span><span style="color:#999999">;</span> <span style="color:#6272a4">// 2 20 200</span>
<span style="color:#ff79c6">int</span> N <span style="color:#6272a4">=</span> <span style="color:#f1fa8c">200</span><span style="color:#999999">;</span> <span style="color:#6272a4">// 2 20 200</span>

<span style="color:#ff79c6">int</span> <span style="color:#8be9fd">main</span><span style="color:#999999">(</span><span style="color:#999999">)</span>
<span style="color:#999999">{</span>
	Eigen<span style="color:#6272a4">::</span>MatrixXd X_mat <span style="color:#6272a4">=</span> Eigen<span style="color:#6272a4">::</span>MatrixXd<span style="color:#6272a4">::</span><span style="color:#8be9fd">Random</span><span style="color:#999999">(</span>M<span style="color:#999999">,</span> N<span style="color:#999999">)</span><span style="color:#999999">;</span>
	Eigen<span style="color:#6272a4">::</span>MatrixXd x <span style="color:#6272a4">=</span> Eigen<span style="color:#6272a4">::</span>VectorXd<span style="color:#6272a4">::</span><span style="color:#8be9fd">Random</span><span style="color:#999999">(</span>M<span style="color:#999999">,</span> <span style="color:#f1fa8c">1</span><span style="color:#999999">)</span><span style="color:#999999">;</span>


	std<span style="color:#6272a4">::</span>cout<span style="color:#6272a4"><<</span><span style="color:#f1fa8c">"test the time cost for Eigen with matrix"</span><span style="color:#6272a4"><<</span>std<span style="color:#6272a4">::</span>endl<span style="color:#999999">;</span>
	<span style="color:#ff79c6">struct</span> timeb startTime<span style="color:#999999">,</span> endTime<span style="color:#999999">;</span>
	<span style="color:#8be9fd">ftime</span><span style="color:#999999">(</span><span style="color:#6272a4">&</span>startTime<span style="color:#999999">)</span><span style="color:#999999">;</span>  
	<span style="color:#ff79c6">for</span> <span style="color:#999999">(</span>size_t i <span style="color:#6272a4">=</span> <span style="color:#f1fa8c">0</span><span style="color:#999999">;</span> i <span style="color:#6272a4"><</span> <span style="color:#f1fa8c">10000</span><span style="color:#999999">;</span> i<span style="color:#6272a4">++</span><span style="color:#999999">)</span>
	<span style="color:#999999">{</span>
		X_mat <span style="color:#6272a4">=</span> <span style="color:#999999">(</span>Eigen<span style="color:#6272a4">::</span><span style="color:#8be9fd">MatrixXd</span><span style="color:#999999">(</span>X_mat<span style="color:#999999">.</span><span style="color:#8be9fd">rows</span><span style="color:#999999">(</span><span style="color:#999999">)</span><span style="color:#999999">,</span> X_mat<span style="color:#999999">.</span><span style="color:#8be9fd">cols</span><span style="color:#999999">(</span><span style="color:#999999">)</span><span style="color:#6272a4">+</span>x<span style="color:#999999">.</span><span style="color:#8be9fd">cols</span><span style="color:#999999">(</span><span style="color:#999999">)</span><span style="color:#999999">)</span><span style="color:#6272a4"><<</span>X_mat<span style="color:#999999">,</span> x<span style="color:#999999">)</span><span style="color:#999999">.</span><span style="color:#8be9fd">finished</span><span style="color:#999999">(</span><span style="color:#999999">)</span><span style="color:#999999">;</span>
	<span style="color:#999999">}</span>	
	<span style="color:#8be9fd">ftime</span><span style="color:#999999">(</span><span style="color:#6272a4">&</span>endTime<span style="color:#999999">)</span><span style="color:#999999">;</span>
	std<span style="color:#6272a4">::</span>cout <span style="color:#6272a4"><<</span> <span style="color:#f1fa8c">"time cost is："</span> <span style="color:#6272a4"><<</span> <span style="color:#999999">(</span>endTime<span style="color:#999999">.</span>time<span style="color:#6272a4">-</span>startTime<span style="color:#999999">.</span>time<span style="color:#999999">)</span><span style="color:#6272a4">*</span><span style="color:#f1fa8c">1000</span> <span style="color:#6272a4">+</span> <span style="color:#999999">(</span>endTime<span style="color:#999999">.</span>millitm <span style="color:#6272a4">-</span> startTime<span style="color:#999999">.</span>millitm<span style="color:#999999">)</span> <span style="color:#6272a4"><<</span> <span style="color:#f1fa8c">"ms"</span> <span style="color:#6272a4"><<</span> std<span style="color:#6272a4">::</span>endl<span style="color:#999999">;</span>


	nc<span style="color:#6272a4">::</span>NdArray<span style="color:#6272a4"><</span><span style="color:#ff79c6">double</span><span style="color:#6272a4">></span> x1 <span style="color:#6272a4">=</span> nc<span style="color:#6272a4">::</span>random<span style="color:#6272a4">::</span>rand<span style="color:#6272a4"><</span><span style="color:#ff79c6">double</span><span style="color:#6272a4">></span><span style="color:#999999">(</span><span style="color:#999999">{</span>M<span style="color:#999999">,</span> <span style="color:#f1fa8c">1</span><span style="color:#999999">}</span><span style="color:#999999">)</span><span style="color:#999999">;</span>
	nc<span style="color:#6272a4">::</span>NdArray<span style="color:#6272a4"><</span><span style="color:#ff79c6">double</span><span style="color:#6272a4">></span> D_arr <span style="color:#6272a4">=</span> nc<span style="color:#6272a4">::</span>random<span style="color:#6272a4">::</span>rand<span style="color:#6272a4"><</span><span style="color:#ff79c6">double</span><span style="color:#6272a4">></span><span style="color:#999999">(</span><span style="color:#999999">{</span>M<span style="color:#999999">,</span> N<span style="color:#999999">}</span><span style="color:#999999">)</span><span style="color:#999999">;</span>


	std<span style="color:#6272a4">::</span>cout<span style="color:#6272a4"><<</span><span style="color:#f1fa8c">"test the time cost for numcpp"</span><span style="color:#6272a4"><<</span>std<span style="color:#6272a4">::</span>endl<span style="color:#999999">;</span>
	<span style="color:#8be9fd">ftime</span><span style="color:#999999">(</span><span style="color:#6272a4">&</span>startTime<span style="color:#999999">)</span><span style="color:#999999">;</span>  
	<span style="color:#ff79c6">for</span> <span style="color:#999999">(</span>size_t i <span style="color:#6272a4">=</span> <span style="color:#f1fa8c">0</span><span style="color:#999999">;</span> i <span style="color:#6272a4"><</span> <span style="color:#f1fa8c">10000</span><span style="color:#999999">;</span> i<span style="color:#6272a4">++</span><span style="color:#999999">)</span>
	<span style="color:#999999">{</span>
		D_arr <span style="color:#6272a4">=</span> nc<span style="color:#6272a4">::</span><span style="color:#8be9fd">column_stack</span><span style="color:#999999">(</span><span style="color:#999999">{</span>D_arr<span style="color:#999999">,</span> x1<span style="color:#999999">}</span><span style="color:#999999">)</span><span style="color:#999999">;</span>
		<span style="color:#6272a4">// std::cout<<phi3<<std::endl;</span>
	<span style="color:#999999">}</span>
	<span style="color:#8be9fd">ftime</span><span style="color:#999999">(</span><span style="color:#6272a4">&</span>endTime<span style="color:#999999">)</span><span style="color:#999999">;</span>
	std<span style="color:#6272a4">::</span>cout <span style="color:#6272a4"><<</span> <span style="color:#f1fa8c">"time cost is："</span> <span style="color:#6272a4"><<</span> <span style="color:#999999">(</span>endTime<span style="color:#999999">.</span>time<span style="color:#6272a4">-</span>startTime<span style="color:#999999">.</span>time<span style="color:#999999">)</span><span style="color:#6272a4">*</span><span style="color:#f1fa8c">1000</span> <span style="color:#6272a4">+</span> <span style="color:#999999">(</span>endTime<span style="color:#999999">.</span>millitm <span style="color:#6272a4">-</span> startTime<span style="color:#999999">.</span>millitm<span style="color:#999999">)</span> <span style="color:#6272a4"><<</span> <span style="color:#f1fa8c">"ms"</span> <span style="color:#6272a4"><<</span> std<span style="color:#6272a4">::</span>endl<span style="color:#999999">;</span>
<span style="color:#999999">}</span>
</code></span>
  • 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
  • python for numpy and cupy
<span style="color:#000000"><code class="language-python"><span style="color:#ff79c6">import</span> cupy <span style="color:#ff79c6">as</span> cp
<span style="color:#ff79c6">import</span> numpy <span style="color:#ff79c6">as</span> np
<span style="color:#ff79c6">import</span> time

M <span style="color:#6272a4">=</span> <span style="color:#f1fa8c">2</span> <span style="color:#6272a4"># 2 20 200</span>
N <span style="color:#6272a4">=</span> <span style="color:#f1fa8c">2</span> <span style="color:#6272a4"># 2 20 200</span>

x1_np <span style="color:#6272a4">=</span> np<span style="color:#999999">.</span>random<span style="color:#999999">.</span>rand<span style="color:#999999">(</span>M<span style="color:#999999">,</span> <span style="color:#f1fa8c">1</span><span style="color:#999999">)</span>
D_arr_np <span style="color:#6272a4">=</span> np<span style="color:#999999">.</span>random<span style="color:#999999">.</span>rand<span style="color:#999999">(</span>M<span style="color:#999999">,</span> N<span style="color:#999999">)</span>

x1_cp <span style="color:#6272a4">=</span> cp<span style="color:#999999">.</span>random<span style="color:#999999">.</span>rand<span style="color:#999999">(</span>M<span style="color:#999999">,</span> <span style="color:#f1fa8c">1</span><span style="color:#999999">)</span>
D_arr_cp <span style="color:#6272a4">=</span> cp<span style="color:#999999">.</span>random<span style="color:#999999">.</span>rand<span style="color:#999999">(</span>M<span style="color:#999999">,</span> N<span style="color:#999999">)</span>

<span style="color:#ff79c6">def</span> <span style="color:#8be9fd">testtimecp</span><span style="color:#999999">(</span><span style="color:#999999">)</span><span style="color:#999999">:</span>
    <span style="color:#ff79c6">global</span> D_arr_cp
    <span style="color:#ff79c6">for</span> i <span style="color:#ff79c6">in</span> <span style="color:#f1fa8c">range</span><span style="color:#999999">(</span><span style="color:#f1fa8c">10000</span><span style="color:#999999">)</span><span style="color:#999999">:</span>
        D_arr_cp <span style="color:#6272a4">=</span> cp<span style="color:#999999">.</span>column_stack<span style="color:#999999">(</span><span style="color:#999999">(</span>D_arr_cp<span style="color:#999999">,</span> x1_cp<span style="color:#999999">)</span><span style="color:#999999">)</span>
        
        
<span style="color:#ff79c6">def</span> <span style="color:#8be9fd">testtimenp</span><span style="color:#999999">(</span><span style="color:#999999">)</span><span style="color:#999999">:</span>
    <span style="color:#ff79c6">global</span> D_arr_np
    <span style="color:#ff79c6">for</span> i <span style="color:#ff79c6">in</span> <span style="color:#f1fa8c">range</span><span style="color:#999999">(</span><span style="color:#f1fa8c">10000</span><span style="color:#999999">)</span><span style="color:#999999">:</span>
        D_arr_np <span style="color:#6272a4">=</span> np<span style="color:#999999">.</span>column_stack<span style="color:#999999">(</span><span style="color:#999999">(</span>D_arr_np<span style="color:#999999">,</span> x1_np<span style="color:#999999">)</span><span style="color:#999999">)</span>
        
<span style="color:#ff79c6">print</span><span style="color:#999999">(</span><span style="color:#f1fa8c">"time cost for cupy:"</span><span style="color:#999999">)</span>        
<span style="color:#6272a4">%</span>time testtimecp<span style="color:#999999">(</span><span style="color:#999999">)</span>

<span style="color:#ff79c6">print</span><span style="color:#999999">(</span><span style="color:#f1fa8c">"time cost for numpy"</span><span style="color:#999999">)</span>
<span style="color:#6272a4">%</span>time testtimenp<span style="color:#999999">(</span><span style="color:#999999">)</span>   
</code></span>
  • 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

3.2 测试结果

矩阵大小M=2, N=2M=20, N=20M=200, N=200
Eigenf(203ms)、d(377ms)f(1.427s)、d(2.875s)f(20.29s)、d(42.458s)
NumCppf(130ms)、d(144ms)f(1.416s)、d(1.319s)f(15.673s)、d(18.797s)
NumPy54ms324ms8.36s
CuPy308ms312ms3.03s

四、结论与建议

4.1 结论

  • 矩阵计算效率由高到低：Eigen>NumCpp>NumPy ；
  • 动态矩阵运算（此处不断增加一列）的效率由高到低：NumPy>NumCpp>Eigen；
  • CuPy在数据维数较低时，没有体现出加速，但是应对高维数据时，有明显的加速；
  • 一个有意思的现象，此处试验结果中数据为float型时，所需的计算成本越低。（最开始，我一直认为32位的float要比64位的double要省时，但是网上有人说：新版C++会将float转为double再处理，等于用float会增加额外的转换步骤，计算效率更低，但是本博客的试验结果又验证回最初的直觉。不知所以然，希望大神路过能给予解答。）

4.2 建议

  • 当不存在大量动态矩阵扩展的算法，C++使用Eigen；
  • 当数据维数很高时，强烈建议GPU加速；
  • 当算法存在动态矩阵扩展时，需要通过试验确定合适的库，可能组合各种库为较合适的方案。

4.3 源起

写了一个动态RBF网络，原来一直在python语言下用，采用NumPy实现。最近需要在C++环境中使用，于是用Eigen实现了。发现，Eigen实现版本的计算成本比NumPy还高。于是又实现了NumCpp版本，发现还是NumPy效率最高。经过以上试验，终于窥见一些缘由：算法里同时需要大量矩阵乘、加、减运算，同时又有大量动态矩阵运算（增加一行或一列）。矩阵乘、加、减的运算相差较各库的动态矩阵计算效率相差小。这也是为什么会产生与直觉不一样的结果。其中，double型计算更加耗时。。。注意和下面结果相矛盾！！！

最终，我没用Eigen,而是采用C++调用Pyhon(也即NumPy实现)的方案。

下面列出动态RBF网络算法利用各库实现的计算成本：

方法EigenNumCppNumPy
time costf(2.575s)、d(1.355s)f(12.048s)、d(5.707s)770ms

其中，float型计算更加耗时。。。怎么和上面结果相矛盾！！！

注意事项

C++编译存在DeBug与Release两种方式，DeBug一般用于调试，编译的版本会存储大量调试信息，实际运行时并不能反应算法的真实计算成本。一定要利用Release方式编译测试代码。

这里要感谢同事小张的提醒。最初测试时，发现Eigen和NumCpp的效率都要比NumPy要差。今天午饭时跟小张抱怨这个疑惑时，小张便给出问题的所在。

在CMakeLists.txt文件里添加下面两行，启动Release编译方式，并启动最高效率编译模式。

modelTSS
关注
  • 4
    点赞
  • 踩
  • 7
    收藏
    觉得还不错? 一键收藏
  • 0
    评论
专栏目录
opencv矩阵转eigen_numpy opencv matlab eigen SVD结果对比
weixin_39624360的博客
01-12 143
参考https://zhuanlan.zhihu.com/p/26306568https://byjiang.com/2017/11/18/SVD/http://www.bluebit.gr/matrix-calculator/https://stackoverflow.com/questions/3856072/single-value-decomposition-implementation-...
Eigen-3.4.0
04-07
Eigen-3.4.0,适用Windows，Visual Studio 2022项目属性管理器中需要配置包含目录Eigen是一个用于线性运算的C++ 模板库，支持 矩阵和矢量运算，数值分析及其相关的算法。
C++ eigen，Matlab CPU，Matlab CPU多核，Matlab GPU谁更快？
李宏宗的博客
04-01 1101
本文测试了C++ eigen，Matlab CPU，Matlab CPU多核，Matlab GPU计算矩阵的效率，废话不多说，直接上测试代码。 模拟了400维的problem，300个DHN网络，一共经行十次测试的的情况（本来选择的是4000维的problem，3000个DHN，计算100次的情况，因为C++ eigen算的实在是太慢了） C++的代码： #include <iostream> #include <Eigen/Dense> #include <wi
Eigen 稀疏矩阵构建效率
最新发布
qq_28129141的博客
05-23 273
来修改，随着矩阵规模增加，该操作效率迅速下降。然后尝试先采用三元组进行预先分配内存，仍然无法提高效率。Eigen 中的SparseMatrix修改元素值时，采用。
Eigen比较运算
douzi949389的博客
12-11 1885
#include <Eigen/Dense> #include <unsupported/Eigen/CXX11/Tensor> int main(int argc, char *argv[]) { float th =0.2; std::vector<float> pd = {0.1f, 0.2f, 0.3f, 0.4f}; auto eigenTensorMap = Eigen::TensorMap<Eigen::Tensor<.
numpy和eigen索引方式比较
xinxiangwangzhi_的博客
04-08 1131
Eigen(c++) 矩阵元素的访问：在矩阵的访问中，行索引总是作为第一个参数，Eigen中矩阵、数组、向量的下标都是从0开始。矩阵元素的访问可以通过”()”操作符完成。例如m(2, 3)既是获取矩阵m的第2行第3列元素。 针对向量还提供”[]”操作符，注意矩阵则不可如此使用 Numpy: 从0开始，并接受从数组末尾开始索引的负索引，使用方括号’‘[]’'进行索引 如：x[1,3] 也可以x[1][3] 二者结果等价，但第二种方式效率低于第一种方式，第二种会生成临时数组。 参考： numpy eigen
记一次win10安装cupy的失败经历(附 cupy与numpy对比)
noblaker的博客
11-08 9137
win10安装cupy中踩的n多个坑 问题的根源应该是我在装cupy之前把Nvidia cuda更新到了最新的11.5版本，原本是10.2版本。中间经历真是太狗血了，花了将近一天时间来搞，当然最后终于好了 记录一下中间大概经历的事情吧，也算是个教训了，中间瞎几把拐弯抹角搞了好多东西，也看了很多的博客跟官方文档，但是都没啥用 中间因为一直在摸索，没想着会整这么久，绕这么多弯，很多就没有截图记录 问题的开始 想着先把nvidia的cuda (Compute Unified Device Archit
Eigen vs Numpy时间对比
weixin_30776273的博客
06-28 1427
Eigen vs Numpy时间对比 有人说Eigen太慢了,用Numpy好一点,我们来看一看是不是这样的 Eigen C++ 程序 t_time.cpp #include <iostream> #include <Eigen/Dense> using namespace Eigen; void integer_time(){ int n_a_rows = ...
eigen3.4.0库
07-31
Eigen3的核心特性在于它的模板元编程技术，这使得它在编译时就能确定矩阵和向量的操作，从而提高了运行效率。库中的数据结构包括动态大小的矩阵、固定大小的矩阵、向量以及更复杂的结构如方程系统和稀疏矩阵。Eigen3...
eigen3.3.3版本安装包
03-16
4. **兼容性**: Eigen与许多其他库（如OpenCV、NumPy等）兼容，方便在不同项目中集成。 5. **多线程支持**: 虽然Eigen本身并不直接实现并行计算，但它设计的接口使得用户可以轻松地利用多核处理器进行并行计算。 6...
eigen3.3.9版本安装包
03-16
这将生成可执行的示例程序，你可以运行它们以检查Eigen是否正确安装。 5. **集成到项目**： 在你的C++项目中，通过添加Eigen头文件的路径到编译器的包含目录，即可开始使用Eigen。例如，如果你的Eigen库位于`/usr...
在VS2019中配置Boost C++、NumCpp、Eigen 和opencv4.3.0库环境.pdf
08-22
在VS2019中配置Boost C++、NumCpp、Eigen 和opencv4.3.0库环境.pdf
Eigen ，MKL和 matlab 矩阵乘法速度比较
weixin_33698823的博客
01-06 1454
Eigen 矩阵乘法的速度 < MKL矩阵乘法的速度，MKL矩阵乘法的速度与matlab矩阵乘法的速度相差不大，但matlab GPU版本的矩阵乘法速度是CUP的两倍，在采用float数据类型时10000*10000的矩阵乘法不到1秒 转载于:https://www.cnblogs.com/hyl2018/p/10229110.html...
计算python执行时间
nemo2011的专栏
08-08 3718
方法1 import datetime starttime = datetime.datetime.now() #long running endtime = datetime.datetime.now() print (endtime - starttime).seconds 方法 2 start = time.time() run_fun() end = time.t
[Eigen中文文档] 稠密矩阵分解函数对比
淋曦的进击手记
06-09 774
该基准测试已在配备英特尔酷睿 i7 @ 2.6 GHz 的笔记本电脑上运行，并使用启用了 AVX 和 FMA 指令集但没有多线程的 clang 进行编译。: 此分解不支持对过度约束问题的直接最小二乘求解，并且报告的时间包括计算对称协方差矩阵。文件生成，可以随意修改它以生成与你的硬件、编译器和问题大小相匹配的表格。的成本，对于前四个基于 Cholesky 和 LU 的求解器，用符号。方阵是对称的，对于过约束矩阵，测试报告的时间包括计算对称协方差矩阵。分解有关， LLT分解速度最快，但也是最不通用和鲁棒的。
Eigen, NumCpp, NumPy, CuPy计算时间成本对比
昔风不起，唯有努力生存！
10-30 5795
对于矩阵运算，著名的C++库有Eigen，Python库有NumPy。由于编译环境的初筛，本来是没有比较两者时间成本的必要。但是，又出了个NumCpp,介绍里说是一个高性能的数学计算 C++ 库，提供了一个简单的 Numpy/Matlab 类似的接口。有了中间这个NumCpp，对比三者的计算时间成本就显得有点意义了。博客园|Eigen vs Numpy时间对比中对Eigen与Numpy的效率进行了比较，但是比较时忽略了C++的编译模式[DeBug与Release]对结果的影响，对比出的结果是不够准确的。本.
python计算运行时间_Python：程序运行时间 计算方法
weixin_40009063的博客
11-22 224
import numpy as npimport mathimport datetimelt = ['I','get','it']everything = ['that','is','all']# 方法一:程序执行的时间starttime = datetime.datetime.now() # 获取当前日期for chunk in lt:everything.append(chunk)endtim...
Python计算程序运行时间
weixin_30248399的博客
08-31 595
方法1 import datetime starttime = datetime.datetime.now() #long running endtime = datetime.datetime.now() print (endtime - starttime).seconds 方法 2 start = time.time() run_fun() end = time.time(...
python与机器学习（二）Numpy / Pandas /矩阵相乘速度对比
Tang_Klay的博客
11-30 2525
python与机器学习（二） NumPy(Numerical Python)是Python语言的一个扩展程序库，支持大量的维度数组与矩阵运算，此外也针对数组运算提供大量的数学函数库。 NumPy的官方文档：https://numpy.org/doc/stable/reference/index.html Pandas是一个强大的分析结构化数据的工具集，它的使用基础是NumPy(提供高性能的矩阵运算)，用于数据挖掘和数据分析,同时也提供数据清洗功能。 Pandas的官方文档：https://pandas.py
numcpp eigen
09-04
Numcpp和Eigen都是用于数值计算的C++开源库。 Numcpp是一个基于C++的科学计算库，它提供了类似于Python中NumPy库的功能。Numcpp提供了多维数组（ndarray）和矩阵运算的功能，可以高效地进行数值计算和数据处理。它...

“相关推荐”对你有帮助么？

  • 非常没帮助
  • 没帮助
  • 一般
  • 有帮助
  • 非常有帮助
提交
评论
添加红包

请填写红包祝福语或标题

个

红包个数最小为10个

元

红包金额最低5元

当前余额3.43元 前往充值 >
需支付：10.00元
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付元
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明：

1.余额是钱包充值的虚拟货币，按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载，可以购买VIP、付费专栏及课程。

余额充值