Multiprocessors: Architecture and Applications

Patton,

doi:10.1109/mc.1985.1662921

Cited by 36 publications

(2 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This does not reduce central processing unit, charges because time oil ali processors is billed to tile job [43]. However, a substantial savings iii memory charges is achieved because the program memory residency time is reduced by the muitiprocessor speedup [44] [45]. For typical problems O and K are approximately 256.…”

Section: Aekek = A[(e-e')modo]k'0kmentioning

confidence: 99%

Generalized Approach to Inverse Problems in Tomography: Image Reconstruction for Spatially Variant Systems Using Natural Pixels *

Baker¹,

Budinger²,

Huesman³

2020

High-Performance Computing in Biomedical Research

View full text Add to dashboard Cite

show abstract

Section: Aekek = A[(e-e')modo]k'0kmentioning

confidence: 99%

Generalized Approach to Inverse Problems in Tomography: Image Reconstruction for Spatially Variant Systems Using Natural Pixels *

Baker¹,

Budinger²,

Huesman³

2020

High-Performance Computing in Biomedical Research

View full text Add to dashboard Cite

show abstract

“…where the constants A and/3 are determined by the constraints T=q + ln N__ =~ln N (13) cz+N N for sufficiently large N, and the speedup is ~ .…”

Section: ) K=l K K=lmentioning

confidence: 99%

Cost-effectiveness of concurrent supercomputers

Resnikoff

1987

J Supercomput

View full text Add to dashboard Cite

This article introduces a mathematical function that describes the latent concurrency inherent in an arbitrary program with given initial conditions. The concurrency function is used to derive asymptotic estimates for speedup, including Amdahl's Law, It provides a new method for analyzing costeffectiveness of the processor-memory-communications constituents of a computing system for applications where system cost andexecution time are mutually elastic variables. The costs of programming and input/output are not taken into account in the present study. The methods are applied to study the relative advantages of serial versus concurrent processing; the relationship of the memory/processor ratio to cost-effectiveness; the conditions that determine the relative advantages of SIMD and MIMD control structures; the effects of various interprocessor communication strategies; and cost-effectiveness implications of the choice of data path width.

show abstract

The university of Minnesota programme in computational science

Patton

1988

Commun. appl. numer. methods

View full text Add to dashboard Cite

SUMMARYThe University of Minnesota installed a Cray-1A supercomputer in December 1981. In the past five years programmes for research, instruction and public service in computational science and engineering have been developed that utilize the Cray-1A and successive machines. The University currently operates a Cray-2 and a Cyber 205.In 1981 the University of Minnesota Computer Center operated a Control Data Cyber 74 for batch FORTRAN research and instructional computing, a Cyber 172 as an instructional timesharing system and a Cyber 730 for database applications. Two departments had installed their own VAX 11/780 systems and time was also being purchased by the central facility for computationally intensive research on the United Computing Cray-1A and a Cybernet Cyber 203. In autumn 1981 a procurement was initiated to encourage bids for twenty VAX 11/780s to be either centrally situated in the computer center or distributed around the campus and networked, or for one or more major CDC or IBM mainframes, or for a Class VI supercomputer (Cray 1 or Cyber 205). Only one bid was tendered, for a lease-returned Cray-1B. This bid was accepted and the system leased through a third-party leasing company. The Cray-1B was installed in December 1981, and in mid-1982 its memory was upgraded to loh words to make the system a Cray-1A (see ReferenceThe first users of the supercomputer were research faculty and graduate students in the Chemical Engineering and Materials Science department, and the Economics department. The Chemistry, Physics and Astronomy departments had their own VAX 11/780s. Usage was slow to shift from the heavily loaded Cyber 74 background FORTRAN machine to its Cray-lA replacement, even though the Cray was at least ten times faster (without vectorizing) and three times more costeffective than the Cyber 74. All the remote user had to do to run a FORTRAN job on the Cray rather than the Cyber 74 was to change the destination code in a job header card, but this action seemed to represent a psychological threshold many users were not willing to cross. User reluctance was eventually overcome by extensive training, through general short courses and more specialized courses on code vectorization. The research use on the machine did not pick up significantly until a year after installation, when the Cyber 74 was phased out and the VAX 11/780s on campus had become heavily loaded. While many graduate students took advantage of the short courses presented by the computer centre staff, only rarely was the Cray used in formal credit courses, such as computational physics courses, primarily for graduate students.Part of the funding for the supercomputer came from sales of computer time to commercial firms for bona fide research computation. Commercial time sales were helpful in covering the cost of the Cray-lA, but never amounted to-more than 10 to 15 per cent of the computer centre's budget.2 By late 1983 the Cray load was increasing rapidly, and plans turned to formalizing computational science research and instruction a...

show abstract

Multiprocessors: Architecture and Applications

Cited by 36 publications

References 9 publications

Generalized Approach to Inverse Problems in Tomography: Image Reconstruction for Spatially Variant Systems Using Natural Pixels *

Generalized Approach to Inverse Problems in Tomography: Image Reconstruction for Spatially Variant Systems Using Natural Pixels *

Cost-effectiveness of concurrent supercomputers

The university of Minnesota programme in computational science

Contact Info

Product

Resources

About