T echnologies and
....to the PetaFLOPS Applications and Software home page!
The needs of important applications are the motivation for designing and building PetaFLOPS machines.
The 1995 PetaFLOPS Computing Summer School/Workshop was held August 14 - 23, 1995, Bodega Bay, California. This Summer Study focused on Applications and Algorithms Challenges for PetaFLOPS Computing.
The characteristics of potential PetaFLOPS scale applications can be used to guide other research areas such as: devices, architectures, and software for PetaFLOPS computers.
Applications Requiring PetaFLOPS Scale Computing
Applications Area I
- Image Processing and Recognition
- Neural Networks
- Symbolic Processing
- Databases and Query Processing
- Vision and Geometric Computing
- Digital Signal Processing
Applications Area II
- Computational Quantum Chemistry
- Molecular Mechanics and Dynamics
- Genome Informatics and Phylogeny
- Reactive Flows and Computational Combustion
- Computational Fluid Dynamics
- Structural Mechanics/Statics
- Geophysical Modeling (GCM, OGCM)
- Biosphere Modeling (bio/geo/chemical)
Applications Area III
- Reservoir Modeling/Seismic
- Astrophysics/General Relativity
- QCD and Computational Particle Physics
- Flow in Porus Media
- Electronic Structure and Materials Properties
- Scattering and Plasma Physics
- Orbital Mechanics
- Rendering and Graphics
Applications Area IV
- Dynamical Systems Simulation
- Electronic Device Simulation
- Discrete Event Simulations
- Circuit Simulation
- Control Systems
- Economic Models
- Stocks and Options Pricing Models
- Digital Libraries and Multimedia
Applications Area V
- Automated Reasoning
- Computer Algebra
- Number Theory
- Graph Algorithms
- Coding Theory
- Game Playing and Searching
- Optimization and Scheduling Problems
- Multidisciplinary Applications
Architecture (Hardware) and Software of PetaFLOPS Machines
Software technology is the system component that bridges the gap between applications and computing hardware. It includes tools for application development as well as software components that regulate system operation and the allocation of system resources. On top of the actual hardware system, these components combine to build a "virtual system" that is effective and pleasant for programmers and end-users to use, and also makes efficient use of the underlying hardware resources.
Software Technology Issues:
- The Challenge of Software Technology
- Technology Trends and Strategic Opportunities
- BLISS versus the Metasystem: Two Visions of PetaFLOPS Computing
- Software Technology Areas
- Programming Languages and Models
- Programming Technology
- Resource Management/Scheduling
Scaling Behavior of Applications
- Typical users have access to GF class machines today
- PF machines are a million times faster
- 86,400 seconds per day and 31,536,000 seconds per year
- PF machines will do in five minutes what it takes GF machines to do in 10 years
- Will PF machines need PB size memories?
- Will PF machines need high I/O bandwidths?
- Can a PF machine be usable with TF class storage and networking?
- What kind of performance will fixed sized problems gain?
Questions for PF Architects
Discussion Issues I
- What is the right paradigm for examining PF applications?
- Extensions of today's architectures or something radical?
- Will the ratios (flops/ram/disk/IO) be maintained?
- Will PF machines have new interfaces for users?
- Voice and video?
- Virtual Environments and Virtual Reality?
- 3D scanner and computer based sensors?
- Will PF machines be in one place?
- What programming models will be supported?
- Global shared memory?
- Message passing?
- Tuple space?
- Neural space?
- Graphical or geometrical metaphors?
Discussion Issues II
- What will be the latency and bandwidth visible to users?
- What kind of fault tolerance will be available?
- How deep will the memory hierarchies become?
- What will be the ratio of integer to floating point performance?
- What new instruction types will be available?
- Will the machine be "open"?
- Will higher level functions be available in hardware? (e.g. FFT,
SolveLinearSystem, Transpose, etc.)
Goals for the Applications Group
- Develop some general outlines for possible PF machines
- Partition known applications into those that are likely to benefit from PF machines and those not likely too
- Develop parallel scaling behavior for number of applications areas and focus on those that are problematic
- New "Yet to be Invented" Applications: develop list of speculative
applications that may be enabled by PF machines
- Develop list of attributes any PF machine must have to enable a broad spectrum of applications
- Develop list of software and systems features/attributes that PF machines should have to enable applications
- Develop economic impact from PF applications scenarios
- VR MUDS and MOOS
- Gesture recognition and gesture based indexing
- Individual based ecological and economic models
- Image based searching and classification of video streams
- Consumer driven interactive consumer product design
- Deep Logic: @ 1000 instructions/inference PF -> TeraInferences
- Circuit verification, circuit design, reasoning assistant
- Verification of proofs (QED) active digital libraries
- Merging CAD and video based simulations
- Synthetic movies (news)
- Mass speech and handwriting recognition
- Complete Knowledge bases
- Inversions of Input/Output models of the global economy and energy models
- Instant free market pricing of global products and services
Return to the P.E.T.A. Directory
Return to the P.E.T.A. Directory
Authorizing NASA Official: Bill Feiereisen, Program Manager, NASA HPCC Program
Senior Editor: Thomas Sterling
Curators: Michele O'Connell (Michele.OConnell@hq.nasa.gov),
Lawrence Picha (Larry.Picha@hq.nasa.gov),
Revised: 21 AUGUST 96 (lpicha)