This paper presents an architecture for a restructurable computer system which reconfigures its resources according to the problem environment for efficient performance. It converts the user's program into an intermediate level language called Realist which is capable of specifying arbitrary resource structures such as an array or a pipeline and the computation to be performed upon these structures. An architectural design for the system is presented with special attention to bus units. It is shown how APL, a vector processing language, can be implemented on the system. Some storage schemes are considered for organizing vectors and matrices to facilitate efficient retrieval and manipulation. The paper is concluded with a comparison of the proposed system to existing high speed architectures.

This research note suggests that before we design hardware or software for the task of problem solving, we re-evaluate the task of problem solving in terms of the linguistic constituents which will be required and the manner in which these linguistic constituents will be combined. Utilizing the principles of composition and abstraction-to-specifics, we conclude that all data and programs might be realized in a structured format called messages. We conclude with a few preliminary thoughts and questions as to how an architecture designed for such structured operands and operators might be designed.

This paper proposes that all data elements in a computer memory be made to be self-identifying by means of a tag. The paper shows that the advantages of the change from the traditional von Neumann machine to tagged architecture are seen in all software areas including programming systems, operating systems, debugging systems, and systems of software instrumentation. It discusses the advantages that accrue to the hardware designer in the implementation and gives examples for large- and small-scale systems. The economic costs of such an implementation for a minicomputer system are examined. The paper concludes that such a machine architecture may well be a suitable replacement for the traditional von Neumann architecture.

This book provides an excellent method to learn about compilers. At an introductory, intermediate, and advanced level, it discusses topics central and related to the construction of compilers, many of which have previously appeared only in journal articles or publications not generally available. The work has only two faults. It was composed on an IBM 360 line printer using FORMAT and thus suffers typographically. It is not a cookbook for advanced compiler implementation. That is, when one attempts to build an advanced compiler with optimization, and when he attempts to find the answer to his pressing practical problems in the book, the answers may well not be there. In fairness, however, the book provides excellent indices and a very good historical and general bibliography that might aid the serious implementer.

This paper investigates the effects of dependence on rank tests, in particular on a class of recently defined nonparametric tests called "mixed" statistical tests. It is shown that the mixed test statistic is asymptotically normal for Gaussian processes with mild regularity properties justifying the use of asymptotic relative efficiency (ARE) as a figure of merit. Results are presented in terms of variations on three well-known statistics--the one-sample Wilcoxon, the two-sample Mann-Whitney, and the Kendall\tau. It is found that the effects of dependence on ARE with respect to a parametric test can be offset to some extent by appropriately grouping sample values. If, however, a constant false-alarm rate is to be attained, either the form of the dependence must be known or some learning scheme must be applied.

A recent work by the authors describes a new class of nonparametric statistics formed by summing the rank statistics on data subsets of a fixed number of sample valuesm. This paper derives the locally-most-powerful tests of this class under the assumption of independent sampling.

Mixed statistical tests are described. It is shown that these tests have a much higher efficiency than conventionally used statistics such as the sign test and polarity coincidence correlation without the high operational complexity of the Wilcoxon, Mann-Whitney, Kendall\tau, or Fisher-Yates: Terry-Hoeffding tests.