CS501 Final Term Highlighted Handouts Set 3 by Masters

Pipelined SRC

Vincent P. Heuring & Harry F. Jordan Chapter 5 Computer Systems Design and Architecture 5.1.3

Summary

• Pipelined Version of the SRC
• Control Signals for Pipelined SRC

Pipelined Version of the SRC

In this lecture, a pipelined version of the SRC is presented. The SRC uses a five-stage pipeline. Those five stages are given below:

1. Instruction Fetch
2. Instruction decode/operand fetch
3. ALU operation
4. Memory access
5. Register write

As shown in the next diagram, there are several registers between each stage.

After the instruction has been fetched, it is stored in IR2 and the incremented value of the program counter is held in PC2. When the register values have been read, the first register value is stored in X3, and the second register value is stored in Y3. IR3 holds the opcode and ra. If it is a store to memory instruction, MD3 holds the register value to be stored.

After the instruction has been executed in the ALU, the register Z4 holds the result. The op-code and ra are passed on to IR4. During the write back stage, the register Z5 holds the value to be stored back into the register, while the op-code and ra are passed into IR5. There are also two separate memories and several multiplexers involved in the pipeline operation. These will be shown at appropriate places in later figures.

The number after a particular register name indicates the stage where the value of this register is used. multiplexers 5 stages of pipeline As mentioned earlier, the SRC instructions fall into the following three categories:

1. ALU Instructions
2. Load/Store instructions
3. Branch Instructions

We will now discuss how to design a common pipeline for all three categories of instructions.

1. ALU instructions

ALU instructions are usually of the form:

op-code ra, rb, rc or op-code ra, rb, constant.

In the diagram shown, X3 and Y3 are temporary registers to hold the values between pipeline stages. X3 is loaded with operand value from the register file. Y3 is loaded with either a register value from the register file or a constant from the instruction. The operands are then available to the ALU. The ALU function is determined by decoding the op-code bits. The result of the ALU operation is stored in register Z4, and then stored in the destination register in the register write back stage. There is no activity in the memory access stage for ALU instructions. Note that Z5, IR3, IR4, and IR5 are not shown explicitly in the figure. The purpose of not including these registers is to keep the drawing simple. However, these registers will transfer values as instructions progress through the pipeline. This comment also applies to some other figures in this discussion.

2. Load/Store instructions

Load/Store instructions are usually of the form:

op-code ra, constant(rb)

The instruction is loaded into IR2 and the incremented value of the PC is loaded in PC2. In the next stage, X3 is loaded with the value in PC2 if the relative addressing mode is used, or the value in rb if the displacement addressing mode is used. Similarly, C1 is transferred to Y3 for the relative addressing mode, and c2 is transferred to Y3 for the displacement addressing mode. The store instruction is completed once memory access has been made and the memory location has been written to. The load instruction is completed once the loaded value is transferred back to the register file. The following figure shows the schematic for a load instruction. A similar schematic can be drawn for the store instruction.