NSA Supercomputing Research Budget Code:  
NSA has a perennial requirement for the highest performance computing technology in order to perform its mission, the national security "Grand Challenge." The Supercomputing Research Program is directed to the discovery and application of methods seeking order of magnitude improvements in the computer support required for deriving intelligence from mathematical and signal processing problems. Activities range from invention and prototyping of new concepts, to improvement in the ability to use leading edge commercial products. A current focus of the research is on parallelism, the ability to apply many processing elements simultaneously to a single problem. The following projects are included within the Supercomputing Research Program:
  • SPLASH - A field programmable gate array-based architecture;
  • Parallel language work - AC (for T3D); implementation of `futures` model of message-passing programming;
  • MARQUISE/SOLITAIRE - Repackage a High Performance Computer (HPC) using diamond-based Multi-Chip Modules (MCMs). This computer will demonstrate the performance and reliability characteristics of diamond substrate and MCMs.
  • TEARDROP - Area array I/O (AAIO) microelectronic technologies have the potential to improve device interconnect bandwidths by an order of magnitude. This project will evaluate the impact of AAIO integrated circuits and MCMs on the performance and scalability of uniform-memory-access (UMA) supercomputer architectures.
  • 3-D DIAMOND MCM CUBE COMPUTER - DARPA program to build test vehicle to demonstrate a 3-D computer architecture with a nanosecond system clock
  • MICRO SPRAY COOLING - Heat extraction values as high as 250 Watts per square centimeter have been demonstrated using micromachined spray head cooling. Values as high as 700 Watts per square centimeter are claimed to be possible by using dynamic feedback temperature sensors.
  • FIBER-OPTIC LOGIC - Optical techniques can theoretically support data rates of several hundred Gb/s for future communications and computing systems.
  • OPTOELECTRONIC CIRCUITS - supports research into a high performance spectrometer on a chip, and a semiconductor optical amplifier.
  • PROGRAMMING METHODS AND LANGUAGES - Provides computational methods and languages for massively parallel, distributed heterogeneous computing platforms and special-purpose processors.
  • ADVANCED COMPUTATIONAL TECHNIQUES - Develops mathematical theory and computational methods to analyze many complex problems akin to factoring discrete logarithms, digital signatures, etc.
  • GLOBAL NETWORK ISSUES - Understanding and using methods and technology concerning computing systems and the "Global Network" to support the NSA's communications missions.
  • TECHNOLOGY OF COMPUTING - Assist in the research of extremely fast communications and computations based on exotic concepts such as nanosecond devices, quantum mechanics, DNA, and non-linear optics.
Budget ($ M)
FY 95 Act 25.90
FY 96 Pres 27.40
FY 96 Est 27.48
FY 97 Rqst 23.43
Program Component Areas
  FY 96 FY 97
HECC 27.48 23.43
LSN    
HCS    
HuCS    
ETHR    
Agency Ties
DARPA Partner
NSF  
DOE  
NASA Partner
NIH User
NSA  
NIST  
NOAA  
EPA  
ED  
AHCPR  
VA  
Milestone Changes  
FY 1995 Actual Milestones FY 1996 Estimated Milestones FY 1997 Agency Requested Milestones
SPLASH commercialization completed.

PARALLEL LANGUAGE WORK - Continued development of C-based compilers - improved mechanisms for distributed memory access model of computation with high efficiency and ease-of-use.

MARQUISE - Design and demonstration of a repackaged commercial supercomputer board using processor modules implemented with diamond-based MCMs.

TEARDROP - Initiation of evaluation study.

3-D DIAMOND MCM CUBE COMPUTER -- Accomplished testing and MCM development work.

CARNOT/CYC - Build 2 cooperating intelligent agents.

LIGHTNING - A number of papers were published by peer-review journals on the concepts behind project LIGHTNING. The architecture was completed. It was verified in software at Sarnoff Laboratories under contract to NSA.

OPTOELECTRONIC CIRCUITS - Ridge wave guides that have very low optical loss factors were designed, manufactured, and tested.

HIGH POWER LIGHT SOURCE - Successfully tested the spray cooling assembly to cool the computer. Five diamond MCM substrates were assembled at Cray Research and Texas Instruments using a copper polyamide interconnect process.

HIGH SPEED OPTICAL SWITCHING - A high speed bit error rate tester was procured for NSA to verify test results of the research in high speed optical switching experiments. 		PARALLEL LANGUAGE WORK - Continued development of C-based compilers - expansion of AC-supported parallel processing platforms.

MARQUISE/SOLITAIRE - Demonstration of repackaged commercial supercomputer with all logic and network components implemented with diamond-based MCMs for reliability and size reduction. Resultant machine will use micro-machined spray cooling. Completion of coolant study and spray cooler environment testing.

TEARDROP - Completion of evaluation study.

3-D DIAMOND MCM CUBE COMPUTER - Completion and demonstration of all components and test vehicle.

MICRO SPRAY - Delivery of several prototype systems and establishment of a reliability data base for system operation.

FIBER-OPTIC LOGIC - Demonstration of cascaded multiple optical gates to support high speed logic operation.

OPTOELECTRONIC CIRCUITS - Develop an infrared-sensitive photodetector which, when assembled with a liquid crystal, will form an optically-addressed spatial light modulator. In addition a circular grating infrared laser source will be developed for open beam communication of 1 Gb/s data.

CYC - Evaluate Cyc for use in message understanding, machine translation, and text retrieval applications. 		BRIOLETTE -- A follow-on program to MARQUISE. The MARQUISE program implemented all of the technologies demonstrated in the DARPA 3-D Diamond MCM Cube except for the Z axis. Phase 1 of the BRIOLETTE program would be the design and evaluation phase (including limited hardware test vehicles) of a UMA supercomputer without a backplane, thereby improving processor-to-memory latency and bandwidth by an order of magnitude.

New initiative in the area of materials, devices, optoelectronics for very high communication rates and computing.

LIGHTNING - Support for a wavelength division multiplexing testbed. Develop an optically-addressed spatial light modulator.

FIBER-OPTIC LOGIC - Techniques that can restore data signals so that high bit error rates can be maintained without compromise of signal quality.

DATA HANDLING TOOLS - Applications-driven knowledge-based tool research.