The stack. This lives in the general random-access memory (RAM) area, but has direct support from the processor via its stack pointer. The stack pointer is moved down to create new memory and moved up to release that memory. This is an extremely fast and efficient way to allocate storage, second only to registers. The Java compiler must know, while it is creating the program, the exact size and lifetime of all the data that is stored on the stack, because it must generate the code to move the stack pointer up and down. This constraint places limits on the flexibility of your programs, so while some Java storage exists on the stack—in particular, object references—Java objects themselves are not placed on the stack
The heap. This is a general-purpose pool of memory (also in the RAM area) where all Java objects live. The nice thing about the heap is that, unlike the stack, the compiler doesn’t need to know how much storage it needs to allocate from the heap or how long that storage must stay on the heap. Thus, there’s a great deal of flexibility in using storage on the heap. Whenever you need to create an object, you simply write the code to create it by using new,and the storage is allocated on the heap when that code is executed. Of course there’s a price you pay for this flexibility. It takes more time to allocate heap storage than it does to allocate stack storage (if you even could create objects on the stack in Java, as you can in C++).
栈有一个很重要的特殊性，就是存在栈中的数据可以共享。假设我们同时定义：
int a = 3;
int b = 3；
编译器先处理int a = 3；首先它会在栈中创建一个变量为a的引用，然后查找栈中是否有3这个值，如果没找到，就将3存放进来，然后将a指向3。接着处理int b = 3；在创建完b的引用变量后，因为在栈中已经有3这个值，便将b直接指向3。这样，就出现了a与b同时均指向3的情况。 这就是张老师讲的享元模式，节约内存。