Browse Prior Art Database

INDIVIDUAL AND MULTI-PROCESSING PERFORMANCE CHARACTERISTS OF PROGRAMS ON LARGE PARALLEL COMPUTERS

IP.com Disclosure Number: IPCOM000149498D
Original Publication Date: 1970-Apr-30
Included in the Prior Art Database: 2007-Apr-01
Document File: 80 page(s) / 3M

Publishing Venue

Software Patent Institute

Related People

Draughon, E.: AUTHOR [+4]

Abstract

IMM 380 April 1970

This text was extracted from a PDF file.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 6% of the total text.

Page 1 of 80

IMM 380

April 1970

         New York University
Courant Institute of Mathematical Sciences

   INDIVIDUAL AND MULTI-PROCESSING
PERFORMANCE CHARACTER1

S'I'IC

      S OF PROGRAMS
ON

LARGE PARALLEL COMPUTERS

E. Draughon, J. Schwartz, and A. Stein

Results obtained at the Courant Institute of Mathematical
Sciences, New York University, with the National Science
Foundation, Grant No. NSF-GJ -76.

[This page contains 1 picture or other non-text object]

Page 2 of 80

[This page contains 1 picture or other non-text object]

Page 3 of 80

Table of Contents

. Introduction ... ...........,....................... 1
....................

I1 . The Simulated Operating System 3

A . System features 3

B Features desirable but not provided 6


C . Timing measurements available in the

simulated system 8

D Errors detected by the simulator 11

E Common difficulties in using the system 12

.......

I11 . Comments on Parallel Programming Techniques 16

A Programming pitfalls 16

...........................

     B . Desirable techniques 19
IV . Summary of Programs Simulated in Our Experiments,
and Associated Experience ......................... 23

A Parallel shooting 23
B . Computation of total potential energy of a
system of N particles interacting by pair-wise
Lennart-Jones forces ........................... 24
C . Monte Carlo computation of atomic energy levels 25

.......

D . Computation of shockwave characteristics 27

     E . Calculation of eigenvalues of a complex matrix . 27
V . Quantities Computed for Programs Run Outside
the Simulated Operating System .................... 28
A . Normal overhead ................................ 28
B . Efficiency . .................................... 28
C . Relative efficiency ............................ 29

D . Cost effectiveness ............................. 30

VI . Measured Results for Programs Run Singly and

Under the Operating System ........................ 31


VII . Measurements of the Overall Efficiency of the

     Simulated Operating System ........................ 59
VIII . Conclusions ....................................... 67
Bibliography ............................................ 69
Appendices ............................................. 70

.

I

................................


.

............

...............................

.

...............

.

........

.

...........................


.

..............................

[This page contains 1 picture or other non-text object]

Page 4 of 80

[This page contains 1 picture or other non-text object]

Page 5 of 80

I. Introduction

A s previously reported, we have a t the Courant Institute

a Parallel Processor Simulator of an "Athene Class" computer which i s able to simulate up to sixty CDC 6600 Central Processing Units each operating independently and all of which share a common memory; i n addition, several instructions are provided which allow for intercommunication between the Central Processor Units. A Pri...