REPOSITORYtechnical papers

covering the full range

of computer system design

The Repository, a collection of over2000 technical papers and documents re-lating to computer science and engineering,is nmintained by the Computer Society as aservice to the information processing com-munity. Some of the papers have beenrefereed; others have not.

If you have a paper of interest to thecomputer field, you are invited to submittwo copies to Dr. Warren L. Semon, IEEEComputer Society Editor-in-Chief, Divisionof Systems and Information Sciences, 313Link Hall, Syracuse University, Syracuse,NY 13210. Be sure to include a cover lettergiving permission to enter the paper in theRepository. Entry in the Repository doesnot constitute publication.

Photocopy prices are $.10 per page. A$1.00 shipping and handling charge will beadded to all orders under 50 pages. Micro-fiche copies are available at $2.50 for each50 pages of copy. Minimum incrementalcharge: $2.50.

Be sure to state the R number, listedbefore the author's last name, of each paperyou order. All Repository items must beprepaid except for companies or institutionswith established accounts A $2.00 invoicecharge will be added to all non-prepaidorders. Please make you check or moneyorder payable to the IEEE ComputerSociety.

R75-213-Mavaddat, F. and B. Parhami,"Two-Level Associative Memory Organiza-tion for Table Look-Up Applications" (21pp., Technical Report No. CSL-74-001,Computer, Systems Laboratory, Arya-MehrUniversity of Technology, Tehran, Iran)

A two-level associative memory organiza-tion is a combination of a small associativememory and a conventional memory, whichbehaves like a large associative memorymost of the time when queries from aspecial class of associative type are posed. Inthis paper, such an organization is proposedand evaluated through sinulation, for table

September 1975

look-up operations of the type encounteredin assemblers and compilers. Motivation forthis work is the prohibitive cost of largeassociative memories. Simulation results areconsistent with those obtained for cachememories and show that a large percentageof the queries can be answered by theassociative memory without a need toreference the conventional memory. Theassociative memory is specially designed forautomatic implementation of the LRU(least recently used) replacement algorithm.The conventional part of the memory maybe part of the main store of the system intowhich the two-level associative memory isincorporated.

R75-214-Nahapetian, A. and C. Lucas,"Storage Allocation for Square and N-CubeArrays of Unpredictable Size" (10 pp.,Technical Report No. CSL-74-002, Compu-ter Systems Laboratory, Arya-Mehr Univer-sity of Technology, Tehran, Iran)

Storage allocation methods for arrays,either at compile-time (static) or run-time(dynamic), by the nature of the addressingpolynomial, require the size of- eachdimension (except possibly one) to be fixedat the time the addressing is beingperformed. An allocation scheme with acorresponding addressing polynomial inde-pendent of dimension size for square andn-cube arrays (arrays having equal size ineach dimension) has been developed, whichallows the array size to vary even as theaddressing is being performed. The resultingstructure can also be considered as amultidimensional generalization of a stack.Using the notation of (aja2 ... an)p torepresent a1pn-l + a2pn-2 + ... + anp0and assuming am to be the leftmostmaximum among the indices (a,, a2,... an), the polynomial for the n-cube arrayis derived as f(al, a2, .. an) = f(O0, . 0)+ (a,a2 ... an)am+l + a - m

(a, a2 ... an)am which reduces to f(al, a2)

f(0, 0) + a2 + a, + (2 - m) * a2 for thetwo dimensional case.

R75-215-Mavaddat, F. and A. Nahapetian,"Decomposition Methods Applied to theDesign of Character Output Devices" (20pp., Technical Report No. CSL-75-001,Computer Systems Laboratory, Arya-MehrUniversity of Technology, Tehran, Iran)

Character decomposition methods arediscussed by which the performance ofcharacter output devices can be improved.This increased performance can be mani-fested both by increase of speed andextension of character space-or a traded-offcombination of both. Correspondingly, thefunctional characteristics of character out-put devices are discussed which can takeadvantage of this increased performance.These ideas are applied to the compromiseddesign of Farsi alphabet which with minormodifications are also applicable to Arabicscript. This can be useful in design ofprinting devices used in countries with morethan three hundred million people. It isbelieved that these methods could also beapplied to other scripts with the coopera-tion of the people conversant in thoselanguages.

R75-216-Parhami, B., "Application of APLfor Rapid Verification of a Digital SystemArchitecture" (9 pp., Technical Report No.CSL-75-002, Computer Systems Labora-tory, Arya-Mehr University of Technology,Tehran, Iran)

An instruction set for a special-purposeassociative processor, designed for informa-tion storage and retrieval applications, isdefined in APL/360. These definitions arethen used to test the validity of the system'sarchitecture for the proposed applications.This is accomplished by writing andsimulating the execution of sample pro-grams on a sample data set and selectivelyobserving the corresponding transformationsperformed on the data set.

Order by R-number.Use the Repository order form

on page 72.

67

R75-217-Parhami, B., "Errors in DigitalComputers: Causes and Cures" (23 -pp.,Te-chnical Report No. CSL-75-003, Compu-ter Systems Laboratory, Arya-Mehr Univer-sity of Technology, Tehran, Iran)

Errors in digital computers are caused bytwo classes of faults: design faults whichmay exist in hardwvare or software of thesystem, and operational faults due tocomponent failures or external interferenceduring system operation. This paper pro-vides an introduction to various causes oferrors in digital computers and measuresthat can be taken to avoid such errors.

R75-218-Parhami, B., "Modeling of Trade-offs in Fault-Tolerant Homogeneous ArrayProcessors" (17 pp., Technical Report No.CSL-75-004, Computer Systems Labora-tory, Arya-Mehr University of Technology,Tehran, Iran)

Array processing has been used as atechnique to improve the performance ofdigital computer systems. The problem ofreliability for array processors is of majorconcern since the extreme complexity ofthese systems makes them highly susceptibleto hardware failures. In this paper, areliability modeling procedure is presentedfor a class of fault-tolerant, homogeneousarray processors with a hard-core recon-figuration switch. The cost/reliability trade-offs implied by the model are discussed andan example is presented for illustration. It isshown that reliability gain is possible byincreasing the total system cost, but only toa certain point. However, before reachingthis point, many different reliability levelscan be attained at varying costs.

R75-219-Mavaddat, F., "An Experimentwith Teaching of Programming Languages"(34 pp., Technical Report No. CSL-75-005,Computer Systems Laboratory, Arya-MehrUniversity of Technology, Tehran, Iran)

After some reflections upon difficultiesinvolved in teaching programming in anintroductory course, a simple programmablemachine of mechanical nature with a smallset of very tangible instructions is proposed.All the necessary concepts such as sequenc-ing, looping, subroutines, and declarativestatements are introduced in the frameworkof this machine, and a number of problemsare introduced for gaining experience in useof such concepts. At closing, it is arguedthat a simplification of most programminglanguages is easily teachable by relatingthem to the concepts developed for thismachine. A simplification of Fortran istreated for this purpose.

R75-220-Parhami, B., "Representation ofDigits by Optically Weighted Dot-MatrixCharacters" (17 pp., Technical Report No.CSL-75-006, Computer Systems Labora-tory, Arya-Mehr University of Technology,Tehran, Iran)

Representation of digits by opticallyweighted characters is a way of representinginformation in combined digital and analogform. Such characters can be used torepresent three-dimensional patterns onconventional printers, with the third dimen-sion visualized by the darkness of variousprinted page areas. In this paper, a generalframework for the design of opticallyweighted characters to represent digits is

68

established, and several examples are givento illustrate the general concepts. It isshown that Farsi digits lend themselves verywell to optically weighted representation,and examples of such character sets aregiven for both decimal and hexadecimalFarsi digits.

R75-221-Ramapriyan, H. K., "Data Hand-ling for the Geometric Correction of LargeImages" (31 pp., Computer Sciences Cor-poration, Huntsville, Alabama)

Several geometric distortions are presentin remotely sensed images depending on thetype of sensors and the object beingobserved. It is often desirable to compensatefor these distortions and store the images inreference to a standard coordinate system.Digital techniques for correction are versa-tile and introduce a minimum of radio-metric errors. The main problems to beconsidered in this area are the determinationof the corrective transformation, resam-pling, and the management of the largequantities of data. This paper considers thelast problem. It is shown here that, by ajudicious rearrangement of the input data,considerable reductions in the requiredmemory capacity can be achieved. Therearrangement can be accomplished inseveral states. The method presented here isamenable to pipeline implementation forprocessing a continuous stream of images. Itcan also be used on a relatively smallcomputer with the staging being in timesequence rather than hardware. Examplesare shown to illustrate the reductions inoverall cost depending on the relative costof memory and the -peripheral devices.

Order by R-number.Use the Repository order form

on page 72.

R75-222-Huberts, G. J., "NORMAN(NORMalizing ANalyzer)" (64 pp., Techni-cal Report UIUCDCS-R-75-7 15, Dept. ofComputer Science, University of Illinois,Urbana, Illinois)NORMAN is a pattern recognition

machine which recognizes two-dimensionalfigures under conditions of translation,rotation, and magnification. It uses "intel-ligent scanning" to build up a profile of anyarbitrary figure and stochastic (i.e., probab-ilistic) processing to normalize the profilewith respect to the three permutations.

R75-223-Bailey, D. W. and M. J. Ferguson,"Variable Length Zero-Error Coding forCorrelated Sources" (23 pp., TechnicalReport A75-1, University of Hawaii, Hono-lulu, Hawaii)

The theory of variable length coding fora discrete memoryless information source isextended to the problem of two correlatedsources. It is well known that the outputsequence from a single source X can beencoded and subsequently reconstructed bya decoder with zero probability of error if,and only if, the average codeword length nxsatisfies nx > H(X). This familiar result isgeneralized to cover correlated source

coding under the assumption of indepen-dent encoders and coupled decoders. Amethod is developed to determine theminimum average codeword lengths Ex andny needed to encode the output sequencesfrom two correlated sources X and Y so thatthey can be reproduced by the decoder withzero distortion. The results are presented asan admissible rate region R in the nx - nyplane.

R75-224-Binder, R., "Variable-LengthPackets in an Unslotted ALOHA Channel"(19 pp., Technical Report B75-2, Universityof Hawaii, Honolulu, Hawaii)

Some simulation results are presentedfor an unslotted ALOHA channel when thetraffic consists of variable-length packets.The latter consist of 80 bits of overhead,with the number of data bytes approxi-mately geometrically distributed betweenone and eighty. The packet delay, numberof users, and channel throughput are foundas a function of the mean retransmissiontime, and performance compared to that ofa channel with constant-length packets. Alarge range of retransmission mean valuesare found necessary to optimize the channel(minimize average delay) over the full rangeof throughput values, for both constant andvariable-length packets. Another result isthat the maximum number of users can beincreased from four to ten times for a meanof ten characters per packet relative to80-character packets, depending on thedesired average delay.

R75-225-Binder, R., "A Simple Mixed-Traffic Technique for Ground-BasedALOHA Channels" (25 pp., TechnicalReport B75-6, University of Hawaii, Hono-lulu, Hawaii)

This report concerns the problem of howto regulate fie traffic in a slotted orunslotted ALOHA channel when interactiveuser traffic is also present. Some simulationresults are given for a simple technique inwhich a constant interval of time is insertedby all file nodes between successive newpacket transmissions. The interactive trafficnodes operate in a normal (uncontrolled)manner; the constant time interval used byfie nodes is adjusted to achieve the desireddelay characteristics for the interactiveusers. The technique is intended for aground-based environment in which inter-active user traffic predominates, with onlyoccasional requirements for file transfers. Itdoes not appear efficient for satellitechannels due to their large propagationdelays.

R75-226-Wax, D., "Design Considerationsfor the ALOHA Radio CommunicationSub-System" (25 pp., Technical ReportB75-11, University of Hawaii, Honolulu,Hawaii)

In the development of any communica-tion system, the design engineer has at hisdisposal a given set of resources andlimitations with which to work, and his jobis to optimize the system capability withinthis given set. The purpose of this report isto describe the important constraints andmajor resources available in the develop-ment of The ALOHA System radio com-munication sub-system, and to show thegeneral engineering solutions generated tomeet system requirements.

COMPUTER

R75-227-Wax, D., "A Description of theALOHA Radio Communication Sub-System" (24 pp., Technical Report B75-12,University of Hawaii, Honolulu, Hawaii)

The purpose of this report is to describeThe ALOHA System's radio communicationsub-system at a functional and blockdiagram level. The material is separated intodescriptions of the types of stations used inthe system, the data interface between theradio sub-system and the data terminalequipment, the modulation scheme em-ployed, and radio transceiver and modemequipment. In addition, results of radiorange measurements are provided withcomparisons to predicted ranges. Also, someresults of impulse noise measurements aregiven, with a discussion about the possibleeffects on burst data transmission by urbanimpulse noise.

R75-228-Ferguson, M. J., "A Study ofUnslotted ALOHA with Arbitrary MessageLengths" (24 pp., Technical Report B75-13,University of Hawaii, Honolulu, Hawaii)

The major result of this paper is thatfixed length packets maximize throughputand minimize delay for an unslottedALOHA random access channel. The modelstudied assumes that a terminal sends apacket with arbitrary length and then waitsan exponential length time (mean X- 1)before attempting to transmit again. Eachtime it transmits, it obtains a newindependent message length according toLo(t). This model discriminates against longpackets by not requiring repeated successiveretransmissions in the event of failure. Evenwith this optimistic view, fixed lengthpackets maximize throughput. The through-put for arbitrary Lo(t) is computed andspecialized for a few examples. This paperrepresents an extension of a previous paperwhere Lo(t) was assumed exponential.

R75-229-Abramson, N. and E. R. Caccia-mani, Jr., "Satellite Data Communications"(20 pp., Technical Report B75-14, Univer-sity of Hawaii, Honolulu, Hawaii)

In this paper we provide a survey of thehistory, technology, and economics ofcommercial satellite data communications.Technological and economic trends insatellite data communications are examined,and the differences between the satellitechannel and terrestrial channels are empha-sized. After a discussion of existing standardtariffed satellite data services and specialtariffed data services, we list some possiblefuture data services which might be offeredto take advantage of the special propertiesof satellite data channels.

R75-230-Thayse, A., "Cellular Hazard-FreeDesign of Combinational Networks" (38pp., MBLE Research Laboratory, Brussels,Belgium)

The theoretical concepts of meet andjoin derivatives, generalized extended vec-tors, and unate envelopes of Booleanfunctions are defined. The derivatives andthe extended vectors give rise to newmethods, using the Kronecker matrixproduct, for prime implicant and primeimplicate extraction. The unate envelopesallow us to derive new canonical forms forBoolean functions. All the above theoreticalconcepts are used to derive new hazard-freedesign techniques for combinational net-works. For example, one deals with cellularmultilevel hazard-free networks, hazard-freedesign with hazard-containing components,and three-level hazard-free networks con-taining a smaller number of dynamichazards than some correspondent two-levelnetworks.

R75-231-Lavington, S. H., G. Thomas, andD. B. G. Edwards, "The MU5 Multi-computer Communication System" (20 pp.,Dept. of Computer Science, The University,Manchester, United Kingdom)

The MU5 multicomputer system consistsof processors, stores, etc. interconnected viacentral hardware known as the exchange. Itis shown that the exchange should providefor a total data throughput in excess of 40Mbytes/sec among 12 units, with separateprovision for local rates of up to 200Mbytes/sec. The configuration adopted isbased on a localized bus network with avery short common highway to permit highswitching speeds. The basic operation of theexchange is described, and the time allowedfor data transmission and priority decisionsis analyzed. By suitable design a datathroughput of 73 Mbytes/sec is achieved.One of the units in the MU5 complex is afour-channel autonomous page transfersystem, intended primarily for moving pagesbetween various levels of random-accessmemory during management of the seg-mented paged virtual store. The design andperformance of the page transfer unit isdescribed. The technology and MTBF of theexchange and page transfer unit arediscussed, and a comparison is made of thecost and fail-soft capability of other relatednetworks.

R75-232-Borucki, W. J., D. H. Card, andG. C. Lyle, "A Method of Using ClusterAnalysis to Study Statistical Dependence inMultivariant Data" (31 pp., Ames ResearchCenter, NASA, Moffett Field, California)

IEEE Computer Society RepositoryCumulative Index 1966-1973

Published by request of Repository users, the cumulative index lists just under1600 entries for 1966 through 1973. It consists of three parts: an author index,a subject index based on keywords, and a chronological index in the order inwhich the items have been listed through the years.Price-$16.00 Member Price-$12.00

To order: Use the multipurpose order form on page 71.

September 1975

A technique is presented that uses bothcluster analysis and a Monte Carlo signifi-cance test of clusters to discover associa-tions between variables in multidimensionaldata. The method is applied to an exampleof a noisy function in three-dimensionalspace and also to a sample from a mixtureof three bivariate normal distributions.

R75-233-Vohdin, B., "ANDIS 74-SolvingDifferential Equations on a Minicomputer"(11 pp., Duesseldorf, West Germany)

Differential equations can easily besolved on an analog computer. This paperoutlines how the same method can be usedon a digital computer. For this purpose thenew programming language ANDIS 74 isintroduced. ANDIS 74 allows for very fastand easy to read programming of differen-tial equations. A compiler for this languageis also discussed. It is shown how evencomplex differential equations can be solvedon a minicomputer.

R75-234-Ram, G., "M. I.S.-TheMETEOSAT Image Simulator" (25 pp.,Weiterstadt, West Germany)

A simulator has been implementedwhich produces simulated telemetry datafor the high speed image (radiometer)channel and the attitude part of thehousekeeping channel. In this paper wedescribe parts of this simulator which areinteresting from the point of view of imageprocessing and computer graphics. Thesimulator was implemented on an IBM370/155 using an IBM 2250 display, amicrofilm recorder SD4060, and a photo-write unit-OPTRONICS International.

R75-235-Casasent, D. and T. McWilliams,"Use of an Interactive Simulator in theDesign of a High-Speed Minicomputer"(29 pp., Carnegie-Mellon University, Dept.of Electrical Engineering, Pittsburgh,Pennsylvania)

A high-speed minicomputer with acache memory, complete I/O, break andinterrupt facilities, implementing theentire PDP-8 instruction set has beendesigned. Complete instruction executiontimes of 125 nsec are possible. Thedesign of this machine was greatly facili-tated by the use of a fast interactivesimulation program with macro and arrayfeatures. Several output timing formatsprovided minimum output data of maxi-mum usefulness because of the dependenceof the density of the data on its dynamicactivity.

R75-236-Chu, Y., "Architecture of a

Nonnumerical Accumulator" (28 pp.,Dept. of Computer Science, University ofMaryland, College Park, Maryland)

This paper presents the architecture ofa nonnumerical accumulator which makesuse of a specially organized associativememory. In contrast to a numericalaccumulator, this nonnumerical accumula-tor is capable of matching, moving,inserting, deleting, altering, counting, andmarking bits, characters, words, and sen-tences stored in the nonnumerical accum-ulator. In this paper, the configurationand instructions of the nonnumericalaccumulator together with programmingtechniques for some applications are pre-sented.

69

R75-237-Kohout, L. and B. R. Gaines,"The Logic of Protection" (24 pp., Dept.of Electrical Engineering Science, Unier-sity of Essex,'United Kingdom)

This paper presents a brief expositionof the role of various mathematicaltechniques in the development and utiliza-tion of 'resource protection structures forcomputers. The first section is concernedwith the' semantics of the problem-thedistinction between protection problemsin general and those whose complexityrnecessitate's deeper theoretical treatment.The 'second section considers the roles ofalgebraic, topological, and modal/multi-valued logical, techniques in the analysis ofprotection. Finally, we give an analysis ofa current protection model to illustratethe problems and techniques.

R75-238-Tse, B. K. P., "Caecotron: ATube for the Blind" (83 pp., TechnicalReport UIUCDCS-R-75-705, Dept. ofComptuter Science, 'University of Illinois,Urbana, fllinois)

Caecotron is a project in visual pros-theses that derives visual informationfrom two television cameras and outputsthe information to the user, namely avisually deprived person, in the form of amodulated tone sequen'ce. The aural pic-ture consists of a 16 x 16 matrix 'mappedonto the field of the television cameras,an-d the audio system scans across thismatrix picture in a way similar to that inwhich a television camera scans its field.Each of the 16 lines is represented in- theaural picture by a distinct frequency, andeach' f the 16 cells of each line' isdisplayed serially. Frequency 'shifts corres-pond to brightness variations, while anamplitude modulation of the tone cor-responds to the distance of a givenelement.

R75-239-Hohulin, K. R., "Networktransduction Programs Based on Con-nectable and Disconnectable Conditionswith Fan-in and Fan-out Restrictions" (50pp., Techhical Report 'UIUCDCS-R-75-719, Dept. of Computer Science, Univer-sity of Illinois, Urbana, 'Illinois)

Two computer programs (NETTRA-G I-FIFO and NETTRA-G2-FIFO) whichimplement network transduction pro-cedures based on connectable and discon-nectable conditions with fan-in 'and fan-out' testrictions are presented in thispaper. These programs can synthesizeNOR (NAND) networks with either com-pletely or incompletely specified outputfunctions. The results of computer experi-ments on NETTRA-Gl-FIFO andNETTRA-G2-FIFO for both completelyand incompletely specified functions arealso presented.

R75-240-Arora, R. K. and P. S. Sinha,"An Optimal Page Allocation Strategy forMinimum Page Faults" (9 pp., ComputerCentet, Indian Institute of Technology,New Delhi, India)

This paper gives an algorithm to findthe optimal number of pages in theworking set of each program in a multi-programming system so as to reduce thetotal number of page faults to a mini-

70

mum. It has been shown that under theassumptions specified, all programs shouldshare the memory equally or full alloca-tion is done for the program whose sizeis less than its share.

R75-241 -Stigall, P. D., "Designing andSupporting HMO, an Integrated HardwareMicrotode Optimizer" (72 pp., Dept. ofElectrical Engineering, University of Mis-souri, Rolla, Missouri)

This paper discusses an algorithm foroptimizing the density and parallelism -ofmicrocoded routines in microprogram-mable machines. Besides presenting thealgorithm itself, this research also analyzesthe algorithm's uses, design integrationproblems, -architectural requirements, andadaptability to conventional machinecharacteristics. Even 'though the pape'rproposes a hardware implementation ofthe algorithm, the algorithm is viewed asan integral part of the entire microcodegeneration and usage process, from initialhigh-level input into a software microcodecompiler down to machine-level executionof the resultant microcode' on the hostmachine. It is believed that, by removingmuch of the traditionally time-consumingand machine-dependent microcode optimi-zation from the software portion of thisprocess, the algorithm can improve theoverall process.

R75-242-Tosser, A. J., "Optimization ofSome NAND-NOR Networks with MultipleInverter" (10 pp., U.E.R. des Sciences' etTechniques, Le' Havre, France)

The bubble symbolism is used totabulate the numbers of NAND- andNOR-gates required to realize disjunctive,conjunctive, or inhibited forms of logicfunctions; availability of multiple invertergates is taken in account. Two modes ofinhibition are considered.

R75-243-Birolini, A., "Hardware Simula-tion of Semi-Markov and Related Proces-ses-Part 2: Simulation of Noise Pulsesand Some Further Applications" (62 pp.,Institute of Applied Physics, Swiss Fede-ral Institute of Technology, Zurich,Switzerland)

This second part of a two part paperdeals with applications of generators ofsemi-Markov and related processes. Atfirst, the problem of model-building andsimulation of noise pulses which occur inbursts on data channels is investigated indetail. A general classification of themodels is proposed, the known modelsare reviewed and two new models basedon semi-Markov processes with a hierarch-ical structure are introduced. The hard-ware simulation of noise pulses is con-sidered for all the models treated. Genera-tion of random test signals and someother applications in the fields of relia-bility and queueing theory, as well as inthose of biomedical and nuclear engineer-ing, are then briefly discussed.

R75-244-Goyal, D. K., "Parallel Evalua-tion of Elementary Symmetric Functions"(16 pp., Technical Report No. 184, Dept.of Electrical Engineering, ComputerScience Laboratory, Princeton University,.Princeton, New Jersey)

A method is presented for evaluatingin parallel the elementary symmetric func-tions:

Fj(xl, x2, *.., xN)

I j N,

of N elements over an arbitrary commuta-tive ring in 0((102gN)2) time using O(N2)operations. Some efficient algorithms forcomputing individually the functions F1,F2, FN_1 and FN in time 0(0og2N) using0(N) processors are also suggested.

R75-245-Kioustelidis, J. B., "On Piece-wise Linear Minimax Approximations ofConvex Functions" (10 pp., ResearchCenter of' National Defense, Galatsi,Athens, Greece)

The problem of finding a piecewiselinear -minimax approximation of a con-vex function is considered, 'and the gene-ral equations which solve this problem arederived. Application is made for thefunctions 2X, I/x, VTR75-246-Baker, F. T., "Structured Pro-gramming in a Production -ProgrammingEnvironment" (14 pp., IBM' FederalSystems Division, Gaithersburg, Maryland)

This paper discusses how structuredprogramming methodology has been intro-duced into a large production program-ming organization using an integrated butflexible approa-ch. It next analyzes theadvantages and disadvantages of eachcomponent of the methodology and -pre-sents some quantitative results on its use.It concludes with recommendations basedon this generally successful experience,which could- be useful to'other organiza-tions interested in improving reliabilityand productivity.

R75-247-Preparata, F. P. and D. E.Muller, "Efficient --Parallel Evaluation ofBoolean Expressions" (3 pp., CoordinatedScience Lab., University of Illinois,Urbana, Illinois)

A boolean expression with n literalscan be evaluated by a parallel processingsystem in at most 1.81 log2n steps, or,equivalently, by a network constructedwith two-input AND- and OR-gates andhaving at most 1.81 log2n levels.

R75-248-Belgard, R., "BLAISE-1726" (5pp., Burroughs Corporation, Goleta, Cali-fornia)

Microprogramming as a valuable toolfor both experimental and productionproducts is shown through the effectiverealization' of an alien, experimental archi-tecture by a m'icroprogrammable hostmachine. Described is the micro-imple-mentation of the BLAISE machine onthe Burroughs B 1726. Some of thefeatures of the high-level languageoriented architecture are discussed withrespect to implementation schemes.' Astructured microprogram' organization ispresented, illustrating the benefits of theapproach.

COMPUTER

xii Xi2 ... xii I

I -<- i 1 -<- i2 <... < ij <N