Computer Architecture Wrap-Up, Lecture Slide - Computer Science, Slides of Computer Architecture and Organization

Download Computer Architecture Wrap-Up, Lecture Slide - Computer Science and more Slides Computer Architecture and Organization in PDF only on Docsity!

Randal E. Bryant

Carnegie Mellon University

CS:APP2e

CS:APP Chapter 4

Computer Architecture

Wrap-Up

http://csapp.cs.cmu.edu

Overview

Wrap-Up of PIPE Design

 Exceptional conditions

 Performance analysis

 Fetch stage design

Modern High-Performance Processors

 Out-of-order execution

Exception Examples

Detect in Fetch Stage

irmovl $100,%eax

rmmovl %eax,0x10000(%eax) # invalid address

jmp $-1 # Invalid jump target

.byte 0xFF # Invalid instruction code

halt # Halt instruction

Detect in Memory Stage

Exceptions in Pipeline Processor #

Desired Behavior

 rmmovl should cause exception

 Following instructions should have no effect on processor state

# demo-exc1.ys

irmovl $100,%eax

rmmovl %eax,0x10000(%eax) # Invalid address

nop

.byte 0xFF # Invalid instruction code

0x000: irmovl $100,%eax

1 2 3 4

F D E M 0x006: rmmovl %eax,0x1000(%eax) F D E 0x00c: nop 0x00d: .byte 0xFF

F D

F

W

5

M

E

D

Exception detected

Exception detected

Maintaining Exception Ordering

 Add status field to pipeline registers

 Fetch stage sets to either “AOK,” “ADR” (when bad fetch address), “HLT” (halt instruction) or “INS” (illegal instruction)

 Decode & execute pass values through

 Memory either passes through or sets to “ADR”

 Exception triggered only when instruction hits write back

F predPC

W stat icode^ valE^ valM^ dstE^ dstM

M stat icode^ Cnd valE^ valA^ dstE^ dstM

E stat icode^ ifun^ valC^ valA^ valB^ dstE^ dstM^ srcA^ srcB

D stat icode^ ifun^ rA rB^ valC^ valP

Exception Handling Logic

Fetch Stage

Memory Stage

Writeback Stage

dmem_error

Determine status code for fetched instruction

int f_stat = [ imem_error: SADR; !instr_valid : SINS; f_icode == IHALT : SHLT; 1 : SAOK; ];

Update the status

int m_stat = [ dmem_error : SADR; 1 : M_stat; ];

int Stat = [

SBUB in earlier stages indicates bubble

W_stat == SBUB : SAOK; 1 : W_stat; ];

Avoiding Side Effects

Presence of Exception Should Disable State Update

 Invalid instructions are converted to pipeline bubbles

 Except have stat indicating exception status

 Data memory will not write to invalid address

 Prevent invalid update of condition codes

 Detect exception in memory stage

 Disable condition code setting in execute

 Must happen in same clock cycle

 Handling exception in final stages

 When detect exception in memory stage

» Start injecting bubbles into memory stage on next cycle

 When detect exception in write-back stage

» Stall excepting instruction

 Included in HCL code

Control Logic for State Changes

Setting Condition Codes

Stage Control

 Also controls updating of memory

Should the condition codes be updated?

bool set_cc = E_icode == IOPL &&

State changes only during normal operation

!m_stat in { SADR, SINS, SHLT } && !W_stat in { SADR, SINS, SHLT };

Start injecting bubbles as soon as exception passes

through memory stage bool M_bubble = m_stat in { SADR, SINS, SHLT } || W_stat in { SADR, SINS, SHLT };

Stall pipeline register W when exception encountered

bool W_stall = W_stat in { SADR, SINS, SHLT };

Performance Metrics

Clock rate

 Measured in Gigahertz

 Function of stage partitioning and circuit design

 Keep amount of work per stage small

Rate at which instructions executed

 CPI: cycles per instruction

 On average, how many clock cycles does each instruction require?

 Function of pipeline design and benchmark programs

 E.g., how frequently are branches mispredicted?

CPI for PIPE

CPI ≈ 1.

 Fetch instruction each clock cycle

 Effectively process new instruction almost every cycle

 Although each individual instruction has latency of 5 cycles

CPI > 1.

 Sometimes must stall or cancel branches

Computing CPI

 C clock cycles

 I instructions executed to completion

 B bubbles injected (C = I + B)

CPI = C/I = (I+B)/I = 1.0 + B/I

 Factor B/I represents average penalty due to bubbles

Fetch Logic Revisited

During Fetch Cycle

**1. Select PC

1. Read bytes from** **instruction memory
2. Examine icode to** determine **instruction length
3. Increment PC**

Timing

 Steps 2 & 4 require significant amount of time

Standard Fetch Timing

 Must Perform Everything in Sequence

 Can’t compute incremented PC until know how much to increment it by

Select PC

Mem. Read Increment

need_regids, need_valC

1 clock cycle

Modified Fetch Timing

29-Bit Incrementer

 Acts as soon as PC selected

 Output not needed until final MUX

 Works in parallel with memory read

Select PC

Mem. Read

Incrementer

need_regids, need_valC 3-bit add

MUX

1 clock cycle

Standard cycle

More Realistic Fetch Logic

Fetch Box

 Integrated into instruction cache

 Fetches entire cache block (16 or 32 bytes)

 Selects current instruction from current block

 Works ahead to fetch next block

 As reaches end of current block

 At branch target

Instruction Cache

Instruction Cache

Byte 0 Bytes 1-

Current Block

Next Block

Current Instruction

Current Instruction

Instr. Length

Instr. Length

Fetch Control

Fetch Control

Other PC Controls