Cedar project: Original goals and progress to date

Technical Report · Mon Jan 01 04:00:00 EST 1990

DOI:https://doi.org/10.2172/5005497· OSTI ID:5005497

The goal of this project, was to demonstrate that supercomputers of the future can exhibit general-purpose behavior and be easy to use. The Cedar Project was based on five key developments which taken together offered a comprehensive solution to achieving high performance computation: the development of VLSI components made large memories and small, fast processors available at low cost; based on many years of work at Illinois and elsewhere, we had designed a shared memory and switch which provide high bandwidth over a wide range of computations and applications areas; the Parafrase project at Illinois had for more than 10 years been aimed at developing software for restructuring ordinary programs to effectively exploit supercomputer architectures; by using a hierarchy of control, we believed that dataflow principles could be used at a high level (macro-dataflow), thus avoiding some of the problems with traditional dataflow methods; and work in numerical algorithms indicated great promise in exploiting multiprocessors without the penalty of high synchronization overhead, which had proved fatal in some earlier studies. In this section we will outline our progress since the beginning of the Cedar Project and briefly describe specific Cedar work that remains to be done.

• Research Organization: Illinois Univ., Chicago, IL (USA)
• Sponsoring Organization: DOE/ER
• DOE Contract Number: FG02-85ER25001
• OSTI ID: 5005497
• Report Number(s): DOE/ER/25001-T4; ON: DE90006761
• Country of Publication: United States
• Language: English

Related Subjects

99 GENERAL AND MISCELLANEOUS
990200* -- Mathematics & Computers
CEDAR COMPUTERS
COMPUTER ARCHITECTURE
COMPUTERS
DATA PROCESSING
DESIGN
DIGITAL COMPUTERS
DOCUMENT TYPES
FORTRAN
MEMORY MANAGEMENT
PERFORMANCE
PROCESSING
PROGRAMMING LANGUAGES
PROGRESS REPORT
SUPERCOMPUTERS
TASK SCHEDULING