bluegene01

Information about bluegene01

Published on September 18, 2007

Author: Malbern

Source: authorstream.com

Content

Emulating Massively Parallel (PetaFLOPS) Machines:  Emulating Massively Parallel (PetaFLOPS) Machines Neelam Saboo, Arun Kumar Singla Joshua Mostkoff Unger, Gengbin Zheng, Laxmikant V. Kalé Department of Computer Science Parallel Programming Laboratory http://charm.cs.uiuc.edu Roadmap:  Roadmap BlueGene Architecture Need for an Emulator Charm++ BlueGene Converse BlueGene Future Work Blue Gene: Processor-in-memory Case Study:  Blue Gene: Processor-in-memory Case Study Five steps to a PetaFLOPS, taken from: http://www.research.ibm.com/bluegene/ FUNCTIONAL MODEL: 34X34X36 cube of shared memory nodes each having 25 processors. SMP Node:  SMP Node 25 processors 200 processing elements Input/Output Buffer 32 x 128 bytes Network Connected to six neighbors via duplex link 16 bit @ 500 MHz = 1 Gigabyte/s Latencies: 5 cycles per hop 75 cycles per turn Processor:  Processor STATS: 500 MHz Memory-side cache eliminates coherency problems 10 cycles local cache 20 cycles remote cache 10 cycles cache miss 8 integer units sharing 2 floating point units 8 x 25 x ~40,000 = ~8 x 106 processing elements! Need for Emulator:  Need for Emulator Emulator – enables programmer to develop, compile, and run software using programming interface that will be used in actual machine Emulator Objectives:  Emulator Objectives Emulate Blue Gene and other petaFLOPS machines. Memory limitations and time limitations on single processor requires that simulation MUST be performed on parallel architecture. Issues: Assume that program written for processor-in-memory machine will handle out-of-order execution and messaging. Therefore don’t need complex event queue/rollback. Emulator Implementation:  Emulator Implementation What are basic data structures/interface? Machine configuration (topology), handler registration Nodes with node-level shared data Threads (associated with each node) representing processing elements Communication between nodes How to handle all these objects on parallel architecture? How to handle object-to-object communication? Difficulties of implementation eased by using Charm++, object-oriented parallel programming paradigm. Experiments on Emulator:  Experiments on Emulator Sample applications implemented: Primes Jacobi relaxation MD prototype ApoA-I: 92k Atoms 40,000 atoms, no bonds calculated, nearest neighbor cutoff Ran full Blue Gene (with 8 x 106 threads) on ~100 ASCI-Red processors Collective Operations:  Collective Operations Explore different algorithms for broadcasts and reductions RING LINE OCTREE x y z Use 'primitive' 30 x 30 x 20 (10 threads) Blue Gene emulation on 50 processor Linux cluster Converse BlueGene Emulator Objective:  Converse BlueGene Emulator Objective Performance estimation (with proper time stamping) Provide API for building Charm++ on top of emulator. Bluegene Emulator :  Bluegene Emulator Node Structure Communication threads Non-affinity message queue Affinity message queue Worker thread inBuffer Performance:  Performance Pingpong Close to Converse pingpong; 81-103 us v.s. 92 us RTT Charm++ pingpong 116 us RTT Charm++ Bluegene pingpong 134-175 us RTT Charm++ on top of Emulator:  Charm++ on top of Emulator BlueGene thread represents Charm++ node; Name conflict: Cpv, Ctv MsgSend, etc CkMyPe(), CkNumPes(), etc Future Work: Simulator:  Future Work: Simulator LeanMD : Fully functional MD with only cutoff How can we examine performance of algorithms on variants of processor-in-memory design in massive system? Several layers of detail to measure Basic: Correctly model performance, timestamp messages with correction for out-of-order execution More detailed: network performance, memory access, modeling sharing of floating-point unit, estimation techniques

Related presentations


Other presentations created by Malbern

Ford
04. 08. 2007
0 views

Ford

aganek ams03
18. 09. 2007
0 views

aganek ams03

carl2004
18. 09. 2007
0 views

carl2004

Innovation CEO
18. 09. 2007
0 views

Innovation CEO

Websight TAV WEB July 2004
18. 09. 2007
0 views

Websight TAV WEB July 2004

BlueRidge
18. 09. 2007
0 views

BlueRidge

sis statewide training
18. 09. 2007
0 views

sis statewide training

060307DFID
18. 09. 2007
0 views

060307DFID

Blue Jean expt 2005
18. 09. 2007
0 views

Blue Jean expt 2005

APA 2005 Strategies for Genetics
18. 09. 2007
0 views

APA 2005 Strategies for Genetics

ross pearce
18. 09. 2007
0 views

ross pearce

present 05
18. 09. 2007
0 views

present 05

mcgee
18. 09. 2007
0 views

mcgee

SIS Asmnt Score Pres 20070522
18. 09. 2007
0 views

SIS Asmnt Score Pres 20070522

mergers
26. 08. 2007
0 views

mergers

WalkTwoMoons
26. 08. 2007
0 views

WalkTwoMoons

a2 pale view narrative technique
26. 08. 2007
0 views

a2 pale view narrative technique

IDAR26
26. 08. 2007
0 views

IDAR26

SundaySept30
26. 08. 2007
0 views

SundaySept30

would you recognize
13. 08. 2007
0 views

would you recognize

Youth
13. 08. 2007
0 views

Youth

WEAD0201
13. 08. 2007
0 views

WEAD0201

wp 10 e
13. 08. 2007
0 views

wp 10 e

VALS Keahey
13. 08. 2007
0 views

VALS Keahey

Workshop Presentation june 2002
13. 08. 2007
0 views

Workshop Presentation june 2002

DisplacedEllerman
04. 08. 2007
0 views

DisplacedEllerman

Framing Effect and Age
04. 08. 2007
0 views

Framing Effect and Age

Freud2004
04. 08. 2007
0 views

Freud2004

feist ch01Intro
04. 08. 2007
0 views

feist ch01Intro

empowering
04. 08. 2007
0 views

empowering

weston
13. 08. 2007
0 views

weston

Coyote Hills Historic B 000
26. 08. 2007
0 views

Coyote Hills Historic B 000

Olshan cegawrkshp 2
18. 09. 2007
0 views

Olshan cegawrkshp 2

EmbryologyLect5
04. 08. 2007
0 views

EmbryologyLect5

FBrancaLuxembourg3mar
04. 08. 2007
0 views

FBrancaLuxembourg3mar

E Newsletter June2006
18. 06. 2007
0 views

E Newsletter June2006

Urdy
18. 06. 2007
0 views

Urdy

tim feeney
18. 06. 2007
0 views

tim feeney

Sports vision 2004
18. 06. 2007
0 views

Sports vision 2004

PAD Cost Effectiveness
18. 06. 2007
0 views

PAD Cost Effectiveness

Geriatric Exercise Handout
18. 06. 2007
0 views

Geriatric Exercise Handout

Expositions Solidaires
18. 06. 2007
0 views

Expositions Solidaires

erasmus aout05en
18. 06. 2007
0 views

erasmus aout05en

ENGLISH IS FUN
18. 06. 2007
0 views

ENGLISH IS FUN

English Culture Sports2 ready
18. 06. 2007
0 views

English Culture Sports2 ready

6746
18. 06. 2007
0 views

6746

20060410 uncch sils
18. 06. 2007
0 views

20060410 uncch sils

20060408 seaall
18. 06. 2007
0 views

20060408 seaall

20050823 Nagoya
18. 06. 2007
0 views

20050823 Nagoya

060217 CLEMENT B talk
18. 06. 2007
0 views

060217 CLEMENT B talk

DREAM ME
18. 06. 2007
0 views

DREAM ME

evaporaion
15. 06. 2007
0 views

evaporaion

Division with 0
15. 06. 2007
0 views

Division with 0

Division Trouble
15. 06. 2007
0 views

Division Trouble

decimals add subtract
15. 06. 2007
0 views

decimals add subtract

coniferousrees
15. 06. 2007
0 views

coniferousrees

conducing a successful online
15. 06. 2007
0 views

conducing a successful online

Caribou Hills Fire pt1
26. 08. 2007
0 views

Caribou Hills Fire pt1

Caribou Hills Fire pt2
26. 08. 2007
0 views

Caribou Hills Fire pt2

7560
18. 06. 2007
0 views

7560

EOCT GOLF REVIEW ALG
15. 06. 2007
0 views

EOCT GOLF REVIEW ALG

StanescuAndreas
18. 09. 2007
0 views

StanescuAndreas

07Sp Retention
26. 08. 2007
0 views

07Sp Retention

civ3 tech tree rev100
26. 08. 2007
0 views

civ3 tech tree rev100

PPoPP2006
18. 09. 2007
0 views

PPoPP2006

Digitalt innhold
04. 08. 2007
0 views

Digitalt innhold

bateman powerpoint figures
26. 08. 2007
0 views

bateman powerpoint figures

LBC BlueObserving
18. 09. 2007
0 views

LBC BlueObserving

conrad brittannia
26. 08. 2007
0 views

conrad brittannia

PMR bullets2
18. 06. 2007
0 views

PMR bullets2

Dr Levitons Slides
04. 08. 2007
0 views

Dr Levitons Slides