Processor I CPSC 321 Andreas Klappenecker

Midterm 1 Thursday, October 7, during the regular class time Covers all material up to that point History MIPS assembly language programming Computer arithmetic MIPS architecture Review on Tuesday, October 5

Administrative Issues Allen Parish will not have a help session this Friday, but again in subsequent Fridays. I will have extra office-hours this Friday, 10:30am-11:30am. There will be no office hours on Tuesday morning (PhD defense).

Questions? Problems?

Today’s Menu The Processor: Datapath and Control

The Processor: Datapath and Control We want to implement portions of MIPS memory-reference instructions: lw, sw arithmetic-logical instructions: add, sub, and, or, slt control flow instructions: beq, j We ignore multiply, divide, and other integer and floating point instructions.

Instruction Fetch Instruction Decode and Register Fetch Execution, Memory Address Computation, or Branch Completion Memory Access or R-type instruction completion Write-back step Five Execution Steps

Implementations All right, give us the details if you can’t help it!

1) Fetch Instructions A memory unit is used to hold and supply the instructions given an address A state element, called the program counter, determines the address We need an adder to increment the program counter to the address of the next instruction.

2) Decode Instructions & Fetch Registers Premise: We do not know at this point which instruction we ought to execute. BUT, we can do some speculative executions: We can fetch the registers using the bits and of the instruction Which instructions do not need this information? The branch command uses bits 0-15 of the instruction (=> sign-extend to 32 bits).

Instruction Word Formats Register format Immediate format Jump format 0 rs rt rd shamt funct op-code rs rt immediate value op-code 26 bit current segment address

3) Execution, Address Calculation, or Branch Completion Execution (depending on instruction) compute a memory address (lw and sw) compute an arithmetic result (int ops) compare (for a branch) and set PC to appropriate address The content of the registers are used as an input to the arithmetic-logic unit. The instruction decode sets the appropriate control signals.

Role of the ALU The output of the arithmetic logic unit is written to a register in the case of arithmetic-logical instructions used as an address in the case of load and store instructions used to determine the next instruction address for branch instructions The ALU has a great variety of roles!

4) Memory Access or R-Type Instruction Completion For load and store instructions, the address ALU_Out calculated by the arithmetic-logic unit is used for memory reference: read MDR = Mem[ALUOut] or store Mem[ALUOut] = B For arithmetic-logical instructions, write the result of the operation into the destination register (Actually, this is the description for the multi-cycle processor)

5) Write-Back Load instructions complete by writing back the value from the memory.

Design Choices We can implement the MIPS processor as a single-cycle processor or as a multi-cycle processor. We will use a synchronous design, so the clocking strategy will be important. The control unit will be simpler for the single- cycle processor, so we start with that.

Abstract, simplified view Two types of functional units: elements that operate on data values (combinational) elements that contain state (sequential) Simplified Datapath

Combinatorial and Sequential Elements The ALU is a combinatorial element Other elements of the design are not combinatorial, but contain a state An element with some internal storage is called a state element State elements have at least two inputs: data clock (determines when data is written)

Clocking Methodology We need to decide when signals can be read and written We need to specify timing behavior For simplicity, we assume an edge-triggered clocking strategy (synchronous design) All storage elements are updated on either raising edge or falling edge: cycle time rising edge falling edge

Synchronous Design Typical execution: – read contents of some state elements, – send values through some combinational logic – write results to one or more state elements

Two inputs: the data value to be stored (D) the clock signal (C) indicating when to read & store D Two outputs: the value of the internal state (Q) and it's complement D-latch

D flip-flop Output changes only on falling clock edge

Summary We will use D-flip flops to build the register file We gradually build up the datapath Simple components will allow us to do this We add the control logic a little later You will need a firm understanding of logic design Study Chapter 4, read Appendix B

Instruction Fetch Use PC to find new instruction PC = PC + 4, preparing for next instruction

Instruction Fetch + PC update

R-Format Instructions Register format op: basic operation of instruction funct: variant of instruction rs: first register source operand rt: second register source operand rd: register destination operand shamt: shift amount op-code rs rt rd shamt funct

Implementing R-Format Instructions  R-Format ALU operations need the register file  and an implementation of the ALU

Built using D flip-flops How can we realize it? Register File

Reading the Register File Multiplexers select the outputs

D flip-flop Output changes only on falling clock edge

Writing into the Register File log n+1 input bits address n+1 registers Falling edge of clock signal determines when to write

Implementing Loads and Stores Register file, ALU, data memory unit, sign extension unit

Datapath for a load and store 1.register access 2.memory access calculation 3.read or write from memory 4.in the case of a load, write into register file

Datapath for a Branch Use ALU to evaluate the branch condition, another adder for branch target = PC (sign extended 16 bits)>>2

Summary Found implementations for R-format instructions Found implementations for loads and stores Found implementations for branches We need to compose these datapaths Simplicity of the implementations is striking!