为什么 ALU 结果将地址写入 memory 用于 R-Type 指令？

Question

I'm trying to understand single-cycle datapath for MIPS instructions. Currently I can trace R-type, I-type and J-type instructions and I'm aware of control signals on different instructions but there are still unclear points in my mind. In this example I assumed there is a R-type instruction let's say add instruction.

I think add instruction has nothing to do with memory. Certainly it doesn't write result value to memory but it points to "Address" input in data memory. This is where I'm confused.

Answer 1

It does not "write" addresses to memory, it gives addresses to the Data Memory.

A memory operation requires 3 things:

• an address
• a value
• a size

Some memory operations are "loads" — these transfer a value from memory to the CPU, so the load specifies a target register for the value; some memory operations are "stores" — these transfer a value to memory from the CPU, so stores specify a source register for the value.

Regardless of whether load or store, an address must be specified to perform a Data Memory operation. On MIPS we have just one addressing mode , which is register + sign extended immediate. In simple MIPS processors, the ALU performs this addition, since it is free/available in that clock cycle for that use.

The control signals turn MemRead on for loads and MemWrite on for stores. Otherwise both of those are off, so the Data Memory is inactive for R-Type instructions, branch instructions, any instruction that doesn't load or store.

For R-Type instructions, the ALU result is routed (around the Data Memory, and eventually) back to the registers as the Write Data. For load instructions, the Data Memory value becomes the Write Data.