intel labs china · microsoft powerpoint - ppt0000000.ppt [read-only] author: hshen1 created date:...
TRANSCRIPT
Intel Labs ChinaIntel Labs China20092009
Gansha WuGansha WuSenior Staff Researcher, Manager,Senior Staff Researcher, Manager,
Parallel Programming Technology Group,Parallel Programming Technology Group,Intel Labs ChinaIntel Labs China
Intel LabsIntel Labs
MissionMission
Intel LabsIntel Labskp
Strategy Strategy ‘‘On a PageOn a Page’’
Intel LabsIntel Labskp
Intel Labs Intel Labs –– WorldwideWorldwidePlus Snapshot of University CollaborationsPlus Snapshot of University Collaborations
SeattleSeattle
HillsboroHillsboro
Santa ClaraSanta Clara
BerkeleyBerkeleyPittsburghPittsburgh
BeijingBeijing
BangaloreBangaloreGuadalajaraGuadalajara
Intel Labs Europe Intel Labs Europe (16 Sites)(16 Sites)
Intel Innovation Centers Intel Innovation Centers (Ireland, Turkey, Dubai, Russia)(Ireland, Turkey, Dubai, Russia)
Intel LabsUniv. Partners
AmericasAmericas• University of Illinois at
Urbana-Champaign• University of California at Berkeley• University of Michigan• McGill University• Massachusetts Institute of Technology • Universidade Federal do Rio de Janeiro• CINVESTAV Unidad Guadalajara Instituto
Europe, Middle EastEurope, Middle East• INSE, INRIA• Technische Universität Braunschweig• Trinity College Dublin• Universita' di Bologna• University of Jordan• Gdansk Technical University• St Petersburg University of Aerospace Instrumentation • University of Oxford• Bogazici University
AsiaAsia• Tsinghua University • Fudan University • Indian Institute of Technology • Kobe University • Universiti Teknologi Malaysia• University of the Philippines• Korea Advanced IST• National Taiwan University
kp
BarcelonaBarcelona
BraunschweigBraunschweig
St. PetersburgSt. Petersburg
Tim
e H
oriz
onTi
me
Hor
izon 33--7 yr7 yr
11--3 yr3 yr
......
Research Areas: Applications, Communications, Wireless,Research Areas: Applications, Communications, Wireless,Compilers, uArch, Components, Compilers, uArch, Components, Circuit, LowCircuit, Low--power, Virtualization power, Virtualization ……
ResearchResearch
DevelopmentDevelopment
ResearchResearch
DevelopmentDevelopment
ResearchResearch
DevelopmentDevelopment
Intel R&D Model
ResearchResearchCouncilCouncil
Academic ResearchAcademic Research >7 yr>7 yr
Intel China LabIntel China Lab
Wireless Technology PlatformWireless Technology Platform
TeraTera--scale I/O Platformscale I/O Platform
Scalable Statistical ComputingScalable Statistical Computing
Compiler & Runtime TechnologyCompiler & Runtime Technology
Parallel Programming TechnologyParallel Programming Technology
Cloud Computing ResearchCloud Computing ResearchConverged IOConverged IO Video MiningVideo Mining Ct, Computer VisionCt, Computer VisionMinistry of Science and Ministry of Science and
TechnologyTechnology
kp
Industry ResearchBased on fully understanding of
Company’s business modelCompany’s core competenceMarket landscape analysis
Questions always to ourselvesWhat are the long-term trends that will shape the computer industry?How will these trends impact industry participants?What will be the computer industry landscape in long-term?
ChallengesNeed to be flexibleOrganize new projectsWork closely with market and business teams
“We should go to where the ball is going, but not where the ball is”““We should go to where the ball is going, but not where the ball We should go to where the ball is going, but not where the ball isis””
Parallel Programming: As an Example
1980’s 1990’s 2000’s
Parallel Machines “Free Lunch” Era Multi-/Many-Core
Parallel Programming is a privileged minority
Ruled by Amdhal’s Law
Gustafson's law was recognized
From MT to MC
Right-hand turn advocates small/many cores and heterogeneity
Need for Productivity
Though Instruction-Level Parallelism dominated, people start to know threads, race conditions, synchronizations
Architects and programmers struggled with consistency models
Threads: Pthreads
Loop parallelism: OpenMP
Task parallelism: Threading Building Blocks
Data parallelism: Ct (C++ for Throughput)
General parallelisms: Concurrent Collections
Intel Parallel Programming Solutions
+ + + + + + + +
Explicit Explicit ParallelismParallelism
Implicit Implicit ParallelismParallelism
“Research was started in late 1990s”““Research was started in late 1990sResearch was started in late 1990s””
SummaryIntel Labs China is an integrated and important part of Intel’s global
research for the ongoing digital revolution
As an industry research lab, we conduct research considering thecompany’s business strategy and the industry’s long-term trends
Go to where the ball is going!
Q&AQ&A
Intel LabsIntel Labs