并行算法2-Introduction to High Performance Computing Something 高性能计算导论

The serial RAM model.串行内存模型

There is a single serial processor connected to memory. This processor issues instructions that operate on data. The operands of these instructions always live in memory.
When you analyze the cost of a serial RAM algorithm, you assume that all instructions have a cost that is bounded by some constant. Now from thar starting point,you do a big O style analysis in terms of the input size. So,that is the serial ram model.

有一个单一的串行处理器连接到内存,这个处理器发出操作数据的指令,这些指令的操作数总是存储在内存中.
当你分析一个串行RAM算法的cost时,假设所有指令的成本都是有限的常数级。现在从这个起点开始，针对输入大小进行一个大O分析。这就是串行内存模型。

What are the alternatives to the serial RAM model?串行RAM模型的替代方案是什么?

The parallel RAM or PRAM model. 并行RAM或PRAM模型.

Instead of one processor,there may be many.They all see the same memory,and you still assume a bounded constant cost per operation. But you get to work with more than one processor.
有多个处理器,它们都看到相同的内存，但仍然假设每个操作的固定cost是有限的常数级,但可以使用多个处理器一起工作。

And since the processors all see the same memory, they can coordinate and communicate by modifying shared variables.
由于处理器都看到相同的内存，所以它们可以通过修改共享变量来进行协调和通信。
Now in this model,your algorithmic analysis will involve big O as before.but you will try to reduce the total cost by up to a factor of P.
在这个模型中，你的算法分析会像以前一样涉及到大O,但要试着将cos降低P倍.