在c++ 11中可以:

auto array = new double[M][N]; 

这样，内存就不会被初始化。要初始化它，可以这样做:

auto array = new double[M][N]();

示例程序(用"g++ -std=c++11"编译):

#include <iostream>
#include <utility>
#include <type_traits>
#include <typeinfo>
#include <cxxabi.h>
using namespace std;

int main()
{
    const auto M = 2;
    const auto N = 2;

    // allocate (no initializatoin)
    auto array = new double[M][N];

    // pollute the memory
    array[0][0] = 2;
    array[1][0] = 3;
    array[0][1] = 4;
    array[1][1] = 5;

    // re-allocate, probably will fetch the same memory block (not portable)
    delete[] array;
    array = new double[M][N];

    // show that memory is not initialized
    for(int r = 0; r < M; r++)
    {
        for(int c = 0; c < N; c++)
            cout << array[r][c] << " ";
        cout << endl;
    }
    cout << endl;

    delete[] array;

    // the proper way to zero-initialize the array
    array = new double[M][N]();

    // show the memory is initialized
    for(int r = 0; r < M; r++)
    {
        for(int c = 0; c < N; c++)
            cout << array[r][c] << " ";
        cout << endl;
    }

    int info;
    cout << abi::__cxa_demangle(typeid(array).name(),0,0,&info) << endl;

    return 0;
}

输出:

2 4 
3 5 

0 0 
0 0 
double (*) [2]

我在创建动态数组时使用这个。如果你有一个类或结构。这是可行的。例子:

struct Sprite {
    int x;
};

int main () {
   int num = 50;
   Sprite **spritearray;//a pointer to a pointer to an object from the Sprite class
   spritearray = new Sprite *[num];
   for (int n = 0; n < num; n++) {
       spritearray[n] = new Sprite;
       spritearray->x = n * 3;
  }

   //delete from random position
    for (int n = 0; n < num; n++) {
        if (spritearray[n]->x < 0) {
      delete spritearray[n];
      spritearray[n] = NULL;
        }
    }

   //delete the array
    for (int n = 0; n < num; n++) {
      if (spritearray[n] != NULL){
         delete spritearray[n];
         spritearray[n] = NULL;
      }
    }
    delete []spritearray;
    spritearray = NULL;

   return 0;
  } 

Typedef是你的朋友

在回顾并查看了许多其他答案之后，我发现需要进行更深层次的解释，因为许多其他答案要么存在性能问题，要么迫使您使用不寻常的或繁重的语法来声明数组，或访问数组元素(或以上所有问题)。

首先，这个答案假设您在编译时知道数组的尺寸。如果你这样做，那么这是最好的解决方案，因为它将提供最好的性能，并允许您使用标准数组语法来访问数组元素。

The reason this gives the best performance is because it allocates all of the arrays as a contiguous block of memory meaning that you are likely to have less page misses and better spacial locality. Allocating in a loop may cause the individual arrays to end up scattered on multiple non-contiguous pages through the virtual memory space as the allocation loop could be interrupted ( possibly multiple times ) by other threads or processes, or simply due to the discretion of the allocator filling in small, empty memory blocks it happens to have available.

其他好处是声明语法简单，数组访问语法标准。

在c++中使用new:

#include <stdio.h>
#include <stdlib.h>

int main(int argc, char **argv) {

typedef double (array5k_t)[5000];

array5k_t *array5k = new array5k_t[5000];

array5k[4999][4999] = 10;
printf("array5k[4999][4999] == %f\n", array5k[4999][4999]);

return 0;
}

或使用calloc的C样式:

#include <stdio.h>
#include <stdlib.h>

int main(int argc, char **argv) {

typedef double (*array5k_t)[5000];

array5k_t array5k = calloc(5000, sizeof(double)*5000);

array5k[4999][4999] = 10;
printf("array5k[4999][4999] == %f\n", array5k[4999][4999]);

return 0;
}

如果只在本地或类中使用数组，则可以使用lambda表达式创建类数组对象。

int main(int argc, char **argv) {
  srand(time(0));
  int w = rand() % 20 + 10;
  int h = rand() % 20 + 10;
  {
    // Here. Only one line and no memory leak.
    auto arr = [=, p = std::make_unique<int[]>(w * h)](int i, int j) -> int & { return *&p[i * w + j]; };
    arr(1, 0) = 6;
    std::cout << arr(1, 0) << '\n';
  }
}

该对象与int[][]不兼容，要将其作为参数传递，需要模板。但是当你访问一个元素时，它可以让你摆脱正确获取[i * w + j]的麻烦。

你也可以使用一个裸std::unique_ptr<int[]>和一个在lambda中实现的getter。但你可能会犯错误，引用不明确。

首先使用指针定义数组(第1行):

int** a = new int* [x];     //x is the number of rows
for(int i = 0; i < x; i++)
    a[i] = new int[y];     //y is the number of columns

int **arr = new int* [NoOfRows]
for(int i = 0; i<NoOfRows; i++)
   arr[i] = new int[noOfColumn];

这是在c++中使用new声明2D数组的好方法 这里，NoOfRow显示了你需要多少行 ans noOfColumn显示你需要多少列