COMP303 - Computer Architecture Lecture 10. Multi-Cycle Design & Exceptions

Transcription

1 COP33 - Computer Architecture Lecture ulti-cycle Design & Exceptions

2 Single Cycle Datapath We designed a processor that requires one cycle per instruction RegDst busw 32 Clk RegWr Rd ux imm6 Rt 5 5 Rs 5 Rw Ra Rb bit Registers 6 Rt busb 32 Sign Extender npc_sel Clk busa ux ALUSrc Fetch Unit ALUctr Data In ALU 32 Clk Equal 32 <3:> Rt <2:25> WrEn Rs <6:2> emwr Adr Data emory Rd emrd <:5> 32 <:5> Imm6 ux emtoreg

3 What s wrong with our CPI= processor? Arithmetic & Logical PC Inst emory Reg File mux ALU mux setup Load PC Inst emory Reg File mux ALU Data em mux setup Critical Path Store PC Inst emory Reg File mux ALU Data em Branch PC Inst emory Reg File cmp mux Long cycle time All instructions take as much time as the slowest Real memory is not so nice as our idealized memory cannot always get the job done in one (short) cycle

4 Reducing Cycle Time Cut combinational dependency graph and insert register / latch Do same work in two fast cycles, rather than one slow one storage element storage element Acyclic Combinational Logic Acyclic Combinational Logic (A) => storage element storage element Acyclic Combinational Logic (B) storage element

5 Basic Limits on Cycle Time Next address logic PC <= branch? PC offset : PC + 4 Fetch Reg <= em[pc] Operand Fetch A <= R[rs], B <= R[rt] ALU execution ALUOut <= A op B Data emory Access <= em[aluout] Result Store R[rd] <= ALUOut Control npc_sel ALUSrc ALUctr emrd emwr RegDst RegWr Next PC PC Fetch Operand Fetch Exec emory Access Reg. File Result Store

6 Partitioning the CPI= Datapath Add registers between smallest steps Exec Next PC PC Fetch Operand Fetch emory Access Reg. File npc_sel ALUSrc ALUctr emrd emwr RegDst RegWr Result Store Allow the instruction to take multiple cycles.

7 Example ulticycle Datapath Fetch Operand Fetch Result Store em Access Next PC PC IR Ext ALU Reg. File npc_sel ALUSrc ALUctr emrd emwr emtoreg RegDst RegWr Reg File A B S Additional registers are added to store values between stages.

8 R-type instructions (add, sub,...) inst Logical Register Transfers ADD R[rd] R[rs] + R[rt]; PC PC + 4 inst ADD Physical Register Transfers IR E[PC] A R[rs]; B R[rt] S A + B R[rd] S; PC PC + 4 Next PC PC Inst. em IR Reg File A B Exec S em Access Reg. File

9 Load instruction inst LW inst LW Logical Register Transfers R[rt] E(R[rs] + sx(im6); PC PC + 4 Physical Register Transfers IR E[PC] A R[rs]; B R[rt] S A + SignEx(Im6) E[S] R[rd] ; PC PC + 4 Next PC PC Inst. em IR Reg File A B Exec S em Access Reg. File

10 Store instruction inst SW inst SW Logical Register Transfers E(R[rs] + sx(im6) R[rt]; PC PC + 4 Physical Register Transfers IR E[PC] A R[rs]; B R[rt] S A + SignEx(Im6); E[S] B PC PC + 4 Next PC PC Inst. em IR Reg File A B Exec S em Access Reg. File

11 Branch instruction inst BEQ Logical Register Transfers if R[rs] == R[rt] then PC PC + sx(im6) else PC PC + 4 inst Physical Register Transfers IR E[pc] A R[rs]; B R[rt] Eq =(A - B = = ) BEQ&Eq PC PC + sx(im6) Next PC PC Inst. em IR Reg File A B Exec S em Access Reg. File

12 ulticycle Implementation Each step in a multicycle implementation will take clock cycle. ulticycle implementation allows a functional unit to be used more than once per instruction, as long as it is used on different clock cycles. Using a functional unit more than once can help to reduce the amount of hardware required. The ability to allow instructions to take different number of clock cycles is another advantage of multicycle implementation.

13 ulticycle Datapath (Figure 5.26, p.32) PC u x Address emory emdata Write data [25-2] [2-6] [5-] Instr. [5-] register [5-] emory data register u x u x Read reg Read reg 2 Write register Write data Read data Registers Read data 2 Sign extend 6 32 A Shift left 2 B 4 u x 2 3 ALU zero ALU result ALUout Differences between single-cycle and multicycle datapath A single memory unit is used for both instruction and data. There is a single ALU, rather than an ALU and two adders. One or more registers are added after every major functional unit to hold the output of that unit until the value is used in a subsequent clock cycle.

14 ulticycle Datapath with Control Lines emread IorD emwrite IRWrite RegDst RegWrite ALUSrcA PC u x Address emory emdata Write data [25-2] [2-6] [5-] Instr. [5-] register [5-] emory data register u x u x Read reg Read reg 2 Write register Write data Read data Registers Read data 2 Sign extend 6 32 A Shift left 2 B 4 u x 2 3 ALU zero ALU result ALU control ALUout [5-] emtoreg ALUSrcB ALUop

15 Figure 5.28: Complete Datapath & Control Signals for ulticycle Implementation (including jump instruction) PCWriteCond PCWrite lord emread emwrite emtoreg IRWrite Outputs Control Op [5-] PCSource ALUOp ALUSrcB ALUSrcA RegWrite RegDst [25-] Shift left 2 Jump address [3-] u 2 x PC u x Address emory emdata Write data [3-26] [25-2] [2-6] [5-] Instr. [5-] register [5-] emory data register u x Read reg Read reg 2 Write register Write data Read data Registers Read data 2 u x Sign extend 6 32 Shift left 2 A B 4 u x u 2 x 3 PC [3-28] ALU control ALU Zero ALU result ALUOut [5-]

16 Execution Steps () Fetch IR = emory[pc]; PC = PC + 4;

17 Execution Steps (2) Decode and Register Fetch A = Reg[IR[25..2]]; B = Reg[IR[2..6]]; ALUOut = PC + (signextend(ir[5..]) << 2);

18 Execution Steps (3) Execution, memory address computation or branch completion emory Reference: ALUOut = A + signextend(ir[5..]); Arithmetic/Logical Operation: ALUOut = A op B; Branch: If (A == B) PC = ALUOut; Jump: PC = PC[3..28] (IR[25..) << 2);

19 Execution Steps (4) emory access or R-type instruction completion emory Reference: DR = emory[aluout]; or emory[aluout] = B; Arithmetic/Logical s (R-type): Reg[IR[5..]] = ALUOut; Branch, Jump: Nothing

20 Execution Steps (5) emory Read completion (Load only) Reg[IR[2..6]] = DR;

21 Finite State achine Control for ulticycle Datapath Start emread ALUSrcA = lord = IRWrite ALUSrcB = ALUOp = PCWrite PCSource = fetch decode/ register fetch ALUSrcA = ALUSrcB = ALUOp = emory address computation Execution Branch completion ALUSrcA = ALUSrcA = ALUSrcA = ALUSrcB = ALUSrcB = ALUSrcB = PCWrite ALUOp = ALUOp = ALUOp = PCWriteCond PCSource = PCSource = (Op = J ) Jump completion (Op = LW ) emory access emory access R-type completion emread lord = emwrite lord = RegDst = RegWrite emtoreg = 4 emory read completion step RegDst = RegWrite emtoreg =

22 Implementation of Finite State achine Controller PCWrite Datapath control outputs ALUOp ALUSrcB ALUSrcA RegWrite RegDst Combinational control logic Outputs PCWriteCond IorD emread emwrite IRWrite emtoreg PCSource Datapath control outputs Inputs NS3 NS2 NS NS Op5 Op4 Op3 Op2 Op Op instruction register opcode field S3 S2 S S State register Next State

23 Logic Equation for Control Signal Outputs Output PCWrite PCWriteCond IorD emread emwrite IRWrite emtoreg PCSource PCSource ALUOp ALUOp ALUSrcB ALUSrcB ALUSrcA RegWrite RegDst Current States state + state9 state8 state3 + state5 state + state3 state5 state state4 state9 state8 state6 state8 state + state2 state + state state2 + state6 + state8 state4 + state7 state7 For Example: PCWrite = S3 S2 S S + S3 S2 S S

24 Logic Equation for Next State Outputs Output Current States Op NextState state4 + state5 + state7 + state8 + state9 NextState state NextState2 state (Op = lw ) + (Op = sw ) NextState3 state2 (Op = lw ) NextState4 state3 NextState5 state2 (Op = sw ) NextState6 state (Op = R-type ) NextState7 state6 NextState8 state (Op = beq ) NextState9 state (Op = jump ) For Example: NextState = State = S3 S2 S S NextState3 = State2 (Op[5-]= lw ) = S3 S2 S S Op5 Op4 Op3 Op2 Op

25 Performance Evaluation What is the average CPI? state diagram gives CPI for each instruction type workload gives frequency of each type Type CPI i for type Frequency CPI i x freq i Arith/Logic 4 4%.6 Load 5 3%.5 Store 4 %.4 branch 3 2%.6 Average CPI: 4.

26 Exceptions and Interrupts Exceptions are exceptional events that disrupt the normal flow of a program Terminology varies between different machines Examples of Interrupts User hitting the keyboard Disk drive asking for attention Arrival of a network packet Examples of Exceptions Divide by zero Overflow Invalid instruction Page fault (non-resident page in memory)

27 Exception Flow When an exception (or interrupt) occurs, control is transferred to the OS User program Operating System Event exception System exception handler Exception return (optional)

28 IPS convention Exception means any unexpected change in control flow, without distinguishing internal or external; Use the term interrupt only when the event is externally caused. Type of event From where? IPS terminology I/O device request External Interrupt Invoke OS from user program Internal Exception Arithmetic overflow Internal Exception Using an undefined instruction Internal Exception Hardware malfunctions Either Exception or Interrupt

29 Handling Exceptions and Interrupts When do we jump to an exception? Upon detection, invoke the OS to service the event What about in the middle of executing a multi-cycle instruction Difficult to abort the middle of an instruction Processor checks for event at the end of every instruction Processor provides EPC & Cause registers to inform OS of cause EPC a 32-bit register used to hold the address of the affected instruction. Cause a register used to record the cause of the exception. To simplify the discussion, assume undefined instruction= arithmetic overflow=

30 Handling Exceptions and Interrupts Status - interrupt mask and enable bits and determines what exceptions can occur. Control signals to write EPC, Cause, and Status Be able to write exception address into PC, increase mux set PC to exception address (8 8 hex ). ay have to undo PC = PC + 4, since want EPC to point to offending instruction (not its successor); PC = PC 4

31 Figure 5.39: The multicycle datapath with the addition needed to implement exceptions PCWriteCond PCWrite lord emread emwrite emtoreg IRWrite Outputs Control Op [5-] CauseWrite IntCause EPCWrite PCSource ALUOp ALUSrcB ALUSrcA RegWrite RegDst PC u x Address emory emdata Write data [3-26] [25-2] [2-6] [5-] Instr. [5-] register [5-] emory data register [25-] u x Read reg Read reg 2 Write register Write data Read data Registers Read data 2 u x Sign extend 6 32 Shift left 2 A B 4 u x u 2 x Shift left 2 PC [3-28] ALU control ALU Zero ALU result Jump address [3-] 8 8 ALUOut u 2 x 3 EPC u x Cause [5-]

32 How Control Detects Exceptions Undefined detected when no next state is defined from state for the op value. We handle this exception by defining the next state value for all op values other than lw, sw, (R-type), j, and beq as new state. Shown symbolically using other to indicate that the op field does not match any of the opcodes that label arcs out of state. Arithmetic overflow included logic in the ALU to detect overflow, and a signal called Overflow is provided as an output from the ALU. This signal is used in the modified finite state machine to specify an additional possible next state. Note: Challenge in designing control of a real machine is to handle different interactions between instructions and other exceptioncausing events such that control logic remains small and fast. Complex interactions makes the control unit the most challenging aspect of hardware design

33 Figure 5.4: Finite state machine to handle exception detection Start emread ALUSrcA = lord = IRWrite ALUSrcB = ALUOp = PCWrite PCSource = fetch decode/ Register fetch ALUSrcA = ALUSrcB = ALUOp = emory address computation Execution Branch completion Branch completion ALUSrcA = ALUSrcA = ALUSrcA = ALUSrcB = ALUSrcB = ALUSrcB = PCWrite ALUOp = ALUOp = ALUOp = PCWriteCond PCSource = PCSource = (Op = J ) (Op = LW ) emory emory access access R-type completion IntCause = CauseWrite emread RegDst = emread ALUSrcA = lord = RegWrite Overflow lord = ALUSrcB = emtoreg = ALUOp = EPCWrite PCWrite PCSource = Write-back step Overflow 4 IntCause = CauseWrite ALUSrcA = ALUSrcB = ALUOp = EPCWrite PCWrite PCSource = RegWrite emtoreg = RegDst =

34 Summary Disadvantages of the Single Cycle Processor Long cycle time Cycle time is too long for all instructions except the Load ulticycle implementations have the advantage of using a different number of cycles for executing each instruction. ulticycle Processor: Divide the instructions into smaller steps Execute each step (instead of the entire instruction) in one cycle Control is specified by finite state diagram (icroprogramming is used for complex instruction set) The most widely used machine implementation is neither single cycle, nor multicycle it s the pipelined implementation (next improvement we will study).

35 Optional Homework In Ch.5: 2, 8,, 27, 3, 33, 36, 43