+ All Categories
Home > Documents > Hardware Accelerationsoc.yonsei.ac.kr/class/material/CAD/Hardwware... · 2017-03-06 · Hardware...

Hardware Accelerationsoc.yonsei.ac.kr/class/material/CAD/Hardwware... · 2017-03-06 · Hardware...

Date post: 04-Jul-2020
Category:
Upload: others
View: 0 times
Download: 0 times
Share this document with a friend
36
Hardware Acceleration Sungho Kang Yonsei University
Transcript
Page 1: Hardware Accelerationsoc.yonsei.ac.kr/class/material/CAD/Hardwware... · 2017-03-06 · Hardware Acceleration Sungho Kang Yonsei University. CS&RSOC YONSEI UNIVERSITY 2 Outline zIntroduction

Hardware Acceleration

Sungho Kang

Yonsei University

Page 2: Hardware Accelerationsoc.yonsei.ac.kr/class/material/CAD/Hardwware... · 2017-03-06 · Hardware Acceleration Sungho Kang Yonsei University. CS&RSOC YONSEI UNIVERSITY 2 Outline zIntroduction

2CS&RSOC YONSEI UNIVERSITY

Outline

IntroductionBoeingTEGASYorktown Simulation EngineLogic Simulation MachineHALZYCADAAP-1Reconfigurable

Page 3: Hardware Accelerationsoc.yonsei.ac.kr/class/material/CAD/Hardwware... · 2017-03-06 · Hardware Acceleration Sungho Kang Yonsei University. CS&RSOC YONSEI UNIVERSITY 2 Outline zIntroduction

3CS&RSOC YONSEI UNIVERSITY

IntroductionWhy Simulation Engine

Speed up difficulty in software simulationParallel Processing

Multi-processingPipeliningArray Processing

Hardware Implementation

Simulation Engine / Hardware AcceleratorCompiledEvent DrivenMulti-ProcessorArray

Page 4: Hardware Accelerationsoc.yonsei.ac.kr/class/material/CAD/Hardwware... · 2017-03-06 · Hardware Acceleration Sungho Kang Yonsei University. CS&RSOC YONSEI UNIVERSITY 2 Outline zIntroduction

4CS&RSOC YONSEI UNIVERSITY

IntroductionSimulation Engine Performance

ArchitectureMaximumEvalution

UnitsSimulationAlgorithm

MaximumGates

AnnouncedSpeed

YSE(IBM)

Multi-ProcessorPipelining

256 Compile1M

( 4 input /1 output gate )

2000M( gates / sec )

HAL(NEC)

Multi-ProcessorPipelining

31Level

ControlledEvent Driven

1.5M 300M( gates / sec )

LE(ZYCAD)

Multi-ProcessorPipelining

16 Event Driven1.6M

( 2 input /1 output gate )

16M( gates / sec )

Page 5: Hardware Accelerationsoc.yonsei.ac.kr/class/material/CAD/Hardwware... · 2017-03-06 · Hardware Acceleration Sungho Kang Yonsei University. CS&RSOC YONSEI UNIVERSITY 2 Outline zIntroduction

5CS&RSOC YONSEI UNIVERSITY

IntroductionSimulation Engine Classification

Simulatoron AAPI

(NTT)

YSE(IBM)

HAL(NEC)LE(ZYCAD)

Array Processor Architecture

Multi-Processor Architecture

Hardware EventDriven Architecture

LSM(Bell Lab.)

Software simulator(Compile)

Software simulator(Event driven)

Actual Circuit

Number of Parallel Processing

655366processor

1 processor5 stage pipe

16 processor5 stage pipe

31 processors

256 processors

Hardware Mock-up

Page 6: Hardware Accelerationsoc.yonsei.ac.kr/class/material/CAD/Hardwware... · 2017-03-06 · Hardware Acceleration Sungho Kang Yonsei University. CS&RSOC YONSEI UNIVERSITY 2 Outline zIntroduction

6CS&RSOC YONSEI UNIVERSITY

BoeingBoeing Computer Simulator

COMMUNICATIONLOOP STATION

COMMUNICATIONLOOP STATION

COMMUNICATIONLOOP STATION

COMMUNICATIONLOOP STATION

LOGICPROCESSOR

COMMUNICATIONLOOP STATION

CONTROL ANDDISPLAY

CONSOLE

INTERFACEUNIT

GENERALPURPOSE

COMPUTERPERIPHERALS

LOGICPROCESSOR

LOGICPROCESSOR

LOGICPROCESSOR

CENTRALELECTRONICCROSSBARSWITCH

Page 7: Hardware Accelerationsoc.yonsei.ac.kr/class/material/CAD/Hardwware... · 2017-03-06 · Hardware Acceleration Sungho Kang Yonsei University. CS&RSOC YONSEI UNIVERSITY 2 Outline zIntroduction

7CS&RSOC YONSEI UNIVERSITY

TEGASTEGAS Accelerator

Control & Statistics Proc.

Status Reg.

Instr.Mem

Cmd Reg

Result Buff Proc.

Activity Srch. Proc.

Evaluation Processors

Result

Flag Mem

PIN MemInstr. Me

m

Update Processor

Host Interface Processor

Input Event

Result Buff.

Update LIFO

672 bit wideSimulation Proc. Mem

To Host

Simulation Data Bus

Maint Proc.

InstrMem

Clock/Clear

I/O Ports

Page 8: Hardware Accelerationsoc.yonsei.ac.kr/class/material/CAD/Hardwware... · 2017-03-06 · Hardware Acceleration Sungho Kang Yonsei University. CS&RSOC YONSEI UNIVERSITY 2 Outline zIntroduction

8CS&RSOC YONSEI UNIVERSITY

TEGASTEGAS Accelerator

Functional Level Block DiagramControl and

StatisticProcessor

UpdatePass

Processor

FaultList

Proessor

HostInterfaceBuffers

MaintenanceProcessorand Clock

Logic

EvaluationPass

Processor

HostInterface

Processor

SimulationProcessingMemories

Page 9: Hardware Accelerationsoc.yonsei.ac.kr/class/material/CAD/Hardwware... · 2017-03-06 · Hardware Acceleration Sungho Kang Yonsei University. CS&RSOC YONSEI UNIVERSITY 2 Outline zIntroduction

9CS&RSOC YONSEI UNIVERSITY

TEGASTEGAS Accelerator

Accelerator Update Processor

LIFOAddress

andAccessControl

MasterController

SimulationProcessing

MemoryAccess

Controller

DescriptorAddress

PipeLIFO

Addressand

AccessControl

To / FromSimulationProcessing

Memory

To/Fromresult

Buffer MemoryTo/From

Control and StatisticsProcessor

To/FromFault ListProcessor

Support Bus

To/FromHost Interface Processor

From Host InterfaceProcessor

From UpdateProcessor

LIFOMemory

From HostInterface

Processor

TO/FromUpdate

ProcessorLIFO Memory

From EvaluationProcessor

Page 10: Hardware Accelerationsoc.yonsei.ac.kr/class/material/CAD/Hardwware... · 2017-03-06 · Hardware Acceleration Sungho Kang Yonsei University. CS&RSOC YONSEI UNIVERSITY 2 Outline zIntroduction

10CS&RSOC YONSEI UNIVERSITY

TEGASTEGAS AcceleratorAccelerator Evaluation Processor

BehaviorProcessorInstruction

Memory

BehaviorialLanguageProcessor

BehavioralBuffer

StructuralEvaluationProcessor

StructuralProcessorInstruction

Memory

StructuralBuffer

StructuralProcessorScheduler

SchedulerMemory

PinAttributeMemory

Level 0

Support Bus

To/From Fault List Processor

To/Fromcontrol and Statistics Processor

To/FromTime Queue Processor,

Time Queue Memory

To/Fromcontrol and Statistics Processor

Behavioral Processor

Structural Processor

To/From Fault List Processor

To/FromActivity Search Processor

To/From Simulation Processing Memory

To/Fromcontrol and Statistics Processor

Page 11: Hardware Accelerationsoc.yonsei.ac.kr/class/material/CAD/Hardwware... · 2017-03-06 · Hardware Acceleration Sungho Kang Yonsei University. CS&RSOC YONSEI UNIVERSITY 2 Outline zIntroduction

11CS&RSOC YONSEI UNIVERSITY

TEGASTEGAS AcceleratorAccelerator Time Queue Processor

TimeQueueSearchPage

ActivityLogic

TimeQueueSearchPage

ActivityLogic

TimeQueue

ProcessorController

TimeQueue

MemoryAvailability Memory

TimeQueue

MemoryAvailabilty

Logic

TimeQueueSearchPage

ActivityLogic

To/From Evalution Processor

To/FromTime Queue

Memory

From Control and Statistics

Processor

From Activity Search ProcessorSimulation

ProcessingMemory

Address Bus

Support Bus

From Control and Statistic Processor

Page 12: Hardware Accelerationsoc.yonsei.ac.kr/class/material/CAD/Hardwware... · 2017-03-06 · Hardware Acceleration Sungho Kang Yonsei University. CS&RSOC YONSEI UNIVERSITY 2 Outline zIntroduction

12CS&RSOC YONSEI UNIVERSITY

YSEYorktown Simulation Engine

Compiled

256 x 256 Switch

Logic Proc 0

Logic Proc 2

Logic Proc 1

Logic Proc 256

ArraySimulator........

Bus ControlControlProc.

Host

Page 13: Hardware Accelerationsoc.yonsei.ac.kr/class/material/CAD/Hardwware... · 2017-03-06 · Hardware Acceleration Sungho Kang Yonsei University. CS&RSOC YONSEI UNIVERSITY 2 Outline zIntroduction

13CS&RSOC YONSEI UNIVERSITY

YSEYorktown Simulation Engine

Partitioning of 256 PUsEach PU simulates a subcircuit consisting of up to 4k gatesSpecialized PU for RAMs and ROMs

All PUs are synchronized by a common clocksPU can evaluate a gate during every clock cyclePartitioning

Minimize the waiting timeControl processor : host to YSE

Page 14: Hardware Accelerationsoc.yonsei.ac.kr/class/material/CAD/Hardwware... · 2017-03-06 · Hardware Acceleration Sungho Kang Yonsei University. CS&RSOC YONSEI UNIVERSITY 2 Outline zIntroduction

14CS&RSOC YONSEI UNIVERSITY

YSEYorktown Simulation Engine

Logic Processor

Program Counter

Fetch

Address

InstructionMemory

(1024 x 80)

Function Memory

Fetch Store

Addr Addr

DATA2048 x 2 x 5

Signal Data

Fetch Addr

Delay Value

Memory1024 x16

Logic Unit

Operands 1- 5

Function

Page 15: Hardware Accelerationsoc.yonsei.ac.kr/class/material/CAD/Hardwware... · 2017-03-06 · Hardware Acceleration Sungho Kang Yonsei University. CS&RSOC YONSEI UNIVERSITY 2 Outline zIntroduction

15CS&RSOC YONSEI UNIVERSITY

YSEYorktown Simulation Engine

Logic ProcessorPC provide the index to the next gate to be evaluated (compiled)Signal values(0,1,X,Z) are stored in data memory Up to 4 inputs for each gate

Generalized DeMorgan Code(GDM)16 functions of 4 valued variablesEvaluation is done in zoom table

Page 16: Hardware Accelerationsoc.yonsei.ac.kr/class/material/CAD/Hardwware... · 2017-03-06 · Hardware Acceleration Sungho Kang Yonsei University. CS&RSOC YONSEI UNIVERSITY 2 Outline zIntroduction

16CS&RSOC YONSEI UNIVERSITY

YSEYorktown Simulation Engine

Function UnitConcatenating gate type with input values by GDM code

Functionmemory GDM

GDM

GDM

GDM1 code

GDM4 code type

GDM5 code

...

input 1value

input 4value

outputvalue

Page 17: Hardware Accelerationsoc.yonsei.ac.kr/class/material/CAD/Hardwware... · 2017-03-06 · Hardware Acceleration Sungho Kang Yonsei University. CS&RSOC YONSEI UNIVERSITY 2 Outline zIntroduction

17CS&RSOC YONSEI UNIVERSITY

LSMLogic Simulation Machine

Event Driven Logic Simulation Tasks

Processor TaskCURRENTEVENTPROC

1) retrieve and distribute event2) record event3) oscillation check

MODELACCESSINGUNIT

1) determine first fanout2) determine next fanout

SIMPLECONFPROC

1) update source configuration2) update and transmit

fanout configuration

EVAL1) evaluate2) check for repeated evaluations

SCHED 1) schedule2) timing analysis

EVENT LISTMANAGER insert in event list

Page 18: Hardware Accelerationsoc.yonsei.ac.kr/class/material/CAD/Hardwware... · 2017-03-06 · Hardware Acceleration Sungho Kang Yonsei University. CS&RSOC YONSEI UNIVERSITY 2 Outline zIntroduction

18CS&RSOC YONSEI UNIVERSITY

LSMLogic Simulation Engine

EVENTLISTMEM

CURRENTEVENTPROC

EVENTLIST

MANAGER

MODELACCESSING

UNIT

UPDATESAND

EVALUATIONS

SCHED

FROM STIMULUS

FILE TO RESULTS FILE

Page 19: Hardware Accelerationsoc.yonsei.ac.kr/class/material/CAD/Hardwware... · 2017-03-06 · Hardware Acceleration Sungho Kang Yonsei University. CS&RSOC YONSEI UNIVERSITY 2 Outline zIntroduction

19CS&RSOC YONSEI UNIVERSITY

LSMLogic Simulation Engine

Pipeline - uniformly distribute the work load How to solve bottleneck of special processors or FIFO buffersElement evaluation is indicated as soon as one of its inputs changeMost time consuming

Fanout determinationGate evaluation

Function evaluations are executed concurrently by parallel processors

Page 20: Hardware Accelerationsoc.yonsei.ac.kr/class/material/CAD/Hardwware... · 2017-03-06 · Hardware Acceleration Sungho Kang Yonsei University. CS&RSOC YONSEI UNIVERSITY 2 Outline zIntroduction

20CS&RSOC YONSEI UNIVERSITY

HALHAL

Block Level HardwareLogic Simulator

Router Cell Network

IC Node Proc.1

Event Set

Out Pin Status

Comparison

IC Pin Conn Tab

Input Pin Status

Tab

Event Fetch

Event Pack

Xmission

IC Node Proc.2

9

IC Node Proc.3

0

IC Node Proc.3

1

Packet Xcvr

Exec. Contro

l

Dynamic Gate Array

PIN Array Data Mem

Gate Array Mem

Addr. Mem

MNS

MNS MNS

Main Memroty

State Mem

Master Mem Acces

s

IC Addr Map

Master Mem Acces

s

IC Addr Map

State Mem

Main Mem Data

DMA Contrlr

CPU

RAM Data

Master Cntr. Proc.

Memory IC Procs.

Page 21: Hardware Accelerationsoc.yonsei.ac.kr/class/material/CAD/Hardwware... · 2017-03-06 · Hardware Acceleration Sungho Kang Yonsei University. CS&RSOC YONSEI UNIVERSITY 2 Outline zIntroduction

21CS&RSOC YONSEI UNIVERSITY

HALHAL

Event searchSignal propagationSystem clock changeExecutes a level controlled event driven mechanismSimulation model consists of logic blocks24 bit message packets solve conflicts when more than one access

Page 22: Hardware Accelerationsoc.yonsei.ac.kr/class/material/CAD/Hardwware... · 2017-03-06 · Hardware Acceleration Sungho Kang Yonsei University. CS&RSOC YONSEI UNIVERSITY 2 Outline zIntroduction

22CS&RSOC YONSEI UNIVERSITY

ZYCADZYCAD Logic Evaluator

Event driven Up to 16 processorsEvent processor - 5 stage pipeline

Get fanout list and determine delayRead fanout and update output statesUpdate input states and determine gate typeEvaluate gate modelSchedule new event

3 input 1 output gate onlyEvent stack - reduce overheadFuture event scheduler - timing wheel

Page 23: Hardware Accelerationsoc.yonsei.ac.kr/class/material/CAD/Hardwware... · 2017-03-06 · Hardware Acceleration Sungho Kang Yonsei University. CS&RSOC YONSEI UNIVERSITY 2 Outline zIntroduction

23CS&RSOC YONSEI UNIVERSITY

AAP-1AAP-1

ARRAYCONTROL

UNIT

INSTRUCTIONMEMORY

INTERFACEUNIT

DATABUFFERMEMORY

PE PE PE PE

PEPEPEPE

PE PE PE PE

256 X 256 PE ARRAY

256

HOST

COMPUTER

16

16

Page 24: Hardware Accelerationsoc.yonsei.ac.kr/class/material/CAD/Hardwware... · 2017-03-06 · Hardware Acceleration Sungho Kang Yonsei University. CS&RSOC YONSEI UNIVERSITY 2 Outline zIntroduction

24CS&RSOC YONSEI UNIVERSITY

AAP-1AAP-1

Processing Element

MUX-ID1

MUX-RUT

REG-RS

MUX-OD

MUX-ID2 RAM-B(32WX1b)

RAM-B(64X1b)

LAT-A LAT-B LAT-S

MUX-ALDS

MUX-CRY

REG-10

REG-C

A BALU

CORALU

'1'

CiU

CiRCiCo

F

DO2

DTU2DTU1

DO1

RF-A RF-B

PELEVEL

BYPASS

8 NEIGHBOR PEs UPPER AND LOWER PEs

Page 25: Hardware Accelerationsoc.yonsei.ac.kr/class/material/CAD/Hardwware... · 2017-03-06 · Hardware Acceleration Sungho Kang Yonsei University. CS&RSOC YONSEI UNIVERSITY 2 Outline zIntroduction

25CS&RSOC YONSEI UNIVERSITY

ReconfigurableReconfigurable Array Architecture

Configuration

Page 26: Hardware Accelerationsoc.yonsei.ac.kr/class/material/CAD/Hardwware... · 2017-03-06 · Hardware Acceleration Sungho Kang Yonsei University. CS&RSOC YONSEI UNIVERSITY 2 Outline zIntroduction

26CS&RSOC YONSEI UNIVERSITY

ReconfigurableMapping

Mapping onto PE Array

Page 27: Hardware Accelerationsoc.yonsei.ac.kr/class/material/CAD/Hardwware... · 2017-03-06 · Hardware Acceleration Sungho Kang Yonsei University. CS&RSOC YONSEI UNIVERSITY 2 Outline zIntroduction

27CS&RSOC YONSEI UNIVERSITY

ReconfigurableMapping

Expanded Circuit Mapping

Page 28: Hardware Accelerationsoc.yonsei.ac.kr/class/material/CAD/Hardwware... · 2017-03-06 · Hardware Acceleration Sungho Kang Yonsei University. CS&RSOC YONSEI UNIVERSITY 2 Outline zIntroduction

28CS&RSOC YONSEI UNIVERSITY

ReconfigurableArray Reconfiguration

Page 29: Hardware Accelerationsoc.yonsei.ac.kr/class/material/CAD/Hardwware... · 2017-03-06 · Hardware Acceleration Sungho Kang Yonsei University. CS&RSOC YONSEI UNIVERSITY 2 Outline zIntroduction

29CS&RSOC YONSEI UNIVERSITY

ReconfigurablePE Array Reconfiguration

Folding

Page 30: Hardware Accelerationsoc.yonsei.ac.kr/class/material/CAD/Hardwware... · 2017-03-06 · Hardware Acceleration Sungho Kang Yonsei University. CS&RSOC YONSEI UNIVERSITY 2 Outline zIntroduction

30CS&RSOC YONSEI UNIVERSITY

ReconfigurableNode Descriptor Memory

Page 31: Hardware Accelerationsoc.yonsei.ac.kr/class/material/CAD/Hardwware... · 2017-03-06 · Hardware Acceleration Sungho Kang Yonsei University. CS&RSOC YONSEI UNIVERSITY 2 Outline zIntroduction

31CS&RSOC YONSEI UNIVERSITY

ReconfigurablePE Cell Block

Page 32: Hardware Accelerationsoc.yonsei.ac.kr/class/material/CAD/Hardwware... · 2017-03-06 · Hardware Acceleration Sungho Kang Yonsei University. CS&RSOC YONSEI UNIVERSITY 2 Outline zIntroduction

32CS&RSOC YONSEI UNIVERSITY

ReconfigurableFirst Expansion

Page 33: Hardware Accelerationsoc.yonsei.ac.kr/class/material/CAD/Hardwware... · 2017-03-06 · Hardware Acceleration Sungho Kang Yonsei University. CS&RSOC YONSEI UNIVERSITY 2 Outline zIntroduction

33CS&RSOC YONSEI UNIVERSITY

ReconfigurableExpansion of XOR and XNOR

Page 34: Hardware Accelerationsoc.yonsei.ac.kr/class/material/CAD/Hardwware... · 2017-03-06 · Hardware Acceleration Sungho Kang Yonsei University. CS&RSOC YONSEI UNIVERSITY 2 Outline zIntroduction

34CS&RSOC YONSEI UNIVERSITY

ReconfigurablePossible Expansion of 4 Input AND

Page 35: Hardware Accelerationsoc.yonsei.ac.kr/class/material/CAD/Hardwware... · 2017-03-06 · Hardware Acceleration Sungho Kang Yonsei University. CS&RSOC YONSEI UNIVERSITY 2 Outline zIntroduction

35CS&RSOC YONSEI UNIVERSITY

ReconfigurableSecond Expansion

Page 36: Hardware Accelerationsoc.yonsei.ac.kr/class/material/CAD/Hardwware... · 2017-03-06 · Hardware Acceleration Sungho Kang Yonsei University. CS&RSOC YONSEI UNIVERSITY 2 Outline zIntroduction

36CS&RSOC YONSEI UNIVERSITY

ReconfigurableFault Simulation Algorithmmake a fault list attached to each elementfor all patterns

perform good simulationstore the results of good simulationfor all levels

insert the first fault in the fault listsimulate the faultcompare the valueif the value is different from that of good simulation

propagate the fault to the next levelelse

drop the faultinsert the first fault in the fault listsimulate the faultcompare the valueif the value is different from that of good simulation

propagate the fault to the next levelelse

drop the faultend

end


Recommended