No abstract available.

Abstract This paper discusses rendezvous on multiprocessors. Three different approaches are compared, represented by three specific systems: Ada, Harmony and BNR Pascal. All three permit tasks to run on multiple processors and use blocking communications primitives, but there are significant differences. For example, control over replying to messages out of sequence and over the allocation of tasks to processors is omitted in Ada, but is available in Harmony. The approach represented by BNR Pascal follows a middle road between Harmony and Ada: a low level protocol, invisible to the programmer, is used to ensure communications reliability, but the programmer is aware of when a rendezvous is remote. If performance considerations and verbosity and robustness are ignored, all three approaches are equivalent. To illustrate this equivalence, and to demonstrate clearly the complexity of the Ada rendezvous, an Ada rendezvous administrator written using Harmony is described. A second method of adapting Harmony to Ada is also presented, in which the Harmony primitives are modified to be closer to Ada. In practice, using Harmony primitives directly will usually result in better programs. It is argued that something very much like the rendezvous adminstrator is needed for any actual implementation of the Ada rendezvous.

Abstract Subroutines for numerical computation have in the past been written mainly in Fortran and in Algol 60, whereas most programming in the future is expected to be in more advanced languages, for example in Ada. It seems to be a great waste if it will become necessary to convert all existing numerical software into the new languages. What is needed is a facility to include subroutines written in the old languages into programs being written in the newer languages. In this paper an automatic procedure to facilitate the use of mixed languages is suggested, the purpose of the paper is to indicate the desirability for providing the required facilities, and to encourage discussion and further research in the area of mixed language programming.

Abstract Message passing as a method of synchronizing and communicating among sequential processes is not as widely understood as monitors, and there is greater diversity in the semantics of what is meant by message passing. This paper surveys some of the issues on which there are differences and gives the rationale behind certain choices and some examples of experience with them. Paradigms are presented for process structuring found to be commonly useful.

In this paper it is shown how data movement, rather than arithmetic operations, can be the hmltmg factor in the performance of parallel computers on matrix computations In particular it is proved that for machines with two-dimensional rectangular grid connectivity (such as ILLIAC IV), muittphcatlon and inversion of NxN matrices inherently require O(N) steps, even if the processing elements are not constrained to execute identical instructions.

Abstract Although overlaying code segments is a very old technique, there is very little literature 68 on what the ground rules are, what the options are and how to select an overlay structure. As a result, the facilities provided by most linkage editors are much poorer than they need be. In this paper we define the problem, give an algorithm for obtaining ‘minimal’ overlay structures and describe the surrounding commands and environment that are necessary to turn this algorithm into a practical tool.

The problem of computing the deternunant of a matrix of polynomials IS considered; two algo. rithms (expansion by minors and expansion by Gaussian ehmmatmn) are compared; and each is examined under two models for polynomial computatmn (dense univanate and totally sparse). The results, while interesting in themselves, also serve to display two points: (1) Asymptotic results are sometimes misleading for noninfimte (e.g. practical) problems. (2) Models of computa-tion are by definition simplifications of reahty: algorithmic analysis should be carried out under several distract computatmnal models and should be supported by empirical data. Key Words and Phrases: symbohc algebraic computation, analysm of algorithms, determinants of polynomial matrices

We consider the problem of computing the determinant of a matrix of polynomials; we compare two algorithms (expansion by minors and Gaussian elimination), examining each under two models for polynomial computation (dense univariate and totally sparse). The results, while interesting in themselves, also serve to display two points:

Goertzel's method, also known as Watt's algorithm, is one of the three standard methods of computing Fourier coefficients, and is especially commonly used when only a small number of coefficients is desired for a given sequence. This paper gives a floating-point error analysis of the technique, and shows why it should be avoided, particularly for low frequencies.