Presentation on theme: "1 System 2020: Research Grand Challenges in Computer Architecture Mary Jane Irwin Penn State University John Shen Intel."— Presentation transcript:
1 System 2020: Research Grand Challenges in Computer Architecture Mary Jane Irwin Penn State University John Shen Intel
2 Mainframes Mini’s Workstations PC’s ??? Eniac What is the next big thing ?
3 What are the mega trends ? 1.Wired Wireless y Telecommunication y Internet/Computing 2.Patch-work Wireless Blanket Wireless 3.Personal Computer Mobile Computer Persistent/Transparent Computer 4.Embedded vs. High-end Convergence? 5.Client vs. Server Convergence?
4 And anticipated usage models ? z Human-centric: y Intelligent spaces x At home, work, school, leisure, … x Active displays, sensory devices, immersive experiences, … y Personal agents x Feature rich gadgets, real-time information x Highly mobile - roam seamlessly from space to space z Infrastructure-centric: y Traditional server farms and data centers x Very large scale information fusion, storage, analysis x Supporting an enormous number of roaming “servers” y Fabric for supporting human-centric uses x Proactively pushing information to roaming agents x Communication and synchronization between spaces
5 The computing paradigm ala Google
6 The computing paradigm ala Nokia
7 CRA System 2020 Workshop z Logistics y December 4 (pm), 5, 6, and 7 (am), 2005 at the Seascape Resort, Monterey Bay y 55 participants (15 industry, 33 academia, 7 CRA and NSF) z Structure y two “bookend” keynotes: Shekhar Borkar (Intel) on “Microarchitecture Challenges for 2015” and Jim Larus (MSR) on “Software Challenges in the Nanoscale Technologies.” y one “industry” panel: Bob Colwell (Consultant), Chuck Moore (AMD), Ravi Nair (IBM), Justin Rattner (Intel), and Steve Scott (Cray) y rest was brainstorming with afternoon report out plenaries z Managed by CRA, funded by NSF
8 What are the components of a GC? z A “grand” scale problem that will require at least a decade of concentrated research to make substantive progress 1.that has a measurable outcomes/milestones, 2.that will excite and engage the computer architecture research community, 3.and that is deserving of considerable investment by funders because it will materially advance the capabilities and conduct of society.
9 1W Featherweight Supercomputer 1.For the goal of 1TOP/W will need up to 10,000X improvement in EPO (energy/op) y 1TOP/W =.001 nJ/op vs. today’s ~10nJ/op 2.Architects are already engaged 3.Societal impacts … and funding y Societal impacts are clear and compelling: pervasive intelligent sensors, embedded supercomputing appliances,... y Funding investments ?
10 Featherweight Challenges z power/energy reduction issues y dynamic and leakage, HW/SW mode controls,... z performance improvement issues y CMPs & SMT, heterogeneous cores, programmable accelerators, eDRAMs, NoCs,... z technology concerns (65nm 45nm 32nm) y↑ process variation, ↑ transient/permanent faults, advanced packaging (SoC MCP 3D),... z design and fabrication concerns y design time & tools, verification & test, fab costs... z programmability...
11 Popular Parallel Programming (P 3 ) 1.Software and architecture support that makes parallel programming easy y If 2X per two year perf. gains continue, will soon have 1000-way chip-level parallelism 2.Architects are becoming engaged but can’t do the job alone y Need compiler, prog. languages, OS & application developers 3.Societal impacts … and funding y A necessary enabling technology for future chips (e.g., the 1W Featherweight Supercomputer) y Funding investments ?
12 P 3 Challenges z new languages/models/compiler issues y that are correct, efficient, scalable, portable,... y that require minimal exposure of the programmer to low-level details y and that support multi-modal parallelism x data-parallel, embarrassingly parallel, irregularly parallel z microarchitecture support issues y lightweight thread/process launch, communication and synchronization, monitoring for reliability and thermal hot spots with dynamic adaptation,... z development support y benchmarks, prototyping platforms, tools for debugging, performance tuning,...
13 Dependable Systems 1. Self-healing hardware and software systems that you can trust your life on y 2x improvement in mean work-to-failure per generation y while reducing the cost of ownership and vendor costs for liability/repair 2. Architects already engaged but can’t do the job alone y A system stack problem – devices, circuits, archs, languages, OS, applications, dependability analysts 3. Societal impacts … and funding y The s/w problem alone is ~ 0.6% GDP of the US y Funding investments ?
14 Dependable Systems Challenges z Host of hardware reliability problems y Transient – SEU, temperature/process variations, … y Permanent – aging, TDDB, NBTI, EM, … z Software reliability issues z Increasing security issues z Dynamically adapt to system constraints of reliability, security, performance, power z Architects can provide low cost solutions y Workload-aware, selective, fast, adaptive y Bring dependability to h/w-s/w interface y Integrated cross-layer solution from devices to applications
15 New Computing Models 1.Beyond the stored program architecture y data flow? neural network? 2.“Expanding the box” for architects y neuroscientists, biologists, chemists,... 3.Societal impacts … and funding y Neuro-prosthetics, telepathy,... y Funding investments ?
16 “Brain” Challenges z High risk – but high payoff y Neuroscientists are a long way from unraveling the mysteries of the neocortex y Take partial steps – augment certain brain functions (hearing for the deaf, vision for the blind, mobility for the quadrapeligic), z Take advantage of emerging technologies y Heterogeneous systems: silicon + nanosensors and actuators, emerging nanotechnologies (CNT, QCAs, quantum,...)
17 Watch for the final report /architecture/home.html z And check out the reports from the previous Grand Challenges conferences