An abstract model is developed for a numerical computation and its decomposition into parallel tasks. The theoretical model is then used to evaluate the performance of iterative algorithms on an MIMD processor archite...
详细信息
An abstract model is developed for a numerical computation and its decomposition into parallel tasks. The theoretical model is then used to evaluate the performance of iterative algorithms on an MIMD processor architecture. Analytical models are derived for communication through a time-shared bus or the shared memory. The focus is on the determination of the optimum number of processors to be allocated to an iterative algorithm, using each of the two communication methods.
The Interface Agreement Processor has been designed to facilitate the development, coordination, and control of large software projects. The processor is external to, but dependent upon, a network based project manage...
详细信息
The Interface Agreement Processor has been designed to facilitate the development, coordination, and control of large software projects. The processor is external to, but dependent upon, a network based project management system. It will be an integral part of a management system to support the development of switching systems software. The use of the management system will help software designers and implementors carry out their tasks and provide for a uniform application of the software development methodology adopted.
An approach to the design of simple, highly reliable, general purpose computersystems is described. An experimental distributed system has been implemented to examine the approach critically. This experimental system...
详细信息
An approach to the design of simple, highly reliable, general purpose computersystems is described. An experimental distributed system has been implemented to examine the approach critically. This experimental system uses five single-board microcomputers, which send messages to each other via a bit-serial, contention-arbitrated bus based on Ethernet principles. A representative user interface has been provided to allow application programs implemented in Pascal to be run.
An algorithm is described for defining the set of active segments that eliminates store fragmentation;this involves analyzing the dynamic matrix for the jumps between segments. High performance is confirmed by simulat...
详细信息
An algorithm is described for defining the set of active segments that eliminates store fragmentation;this involves analyzing the dynamic matrix for the jumps between segments. High performance is confirmed by simulation.
The mathematical analysis of clustering in linear probing, in comparison with that in uniform hashing, is treated. The probability distribution of clusters is first derived by formulating the manner of clustering, and...
详细信息
The mathematical analysis of clustering in linear probing, in comparison with that in uniform hashing, is treated. The probability distribution of clusters is first derived by formulating the manner of clustering, and then the average numbers of probes are expressed in terms of this, which quantitatively indicates the adverse effects of clustering. The probability distribution is also applied in the performance analysis of the new algorithm with parallel linear probing.
The decision-table (DETAB) representation of an algorithm composed of independent modules is discussed, along with the application of Boolean-algebraic methods for the translation of DETAB into procedural language. Th...
详细信息
The decision-table (DETAB) representation of an algorithm composed of independent modules is discussed, along with the application of Boolean-algebraic methods for the translation of DETAB into procedural language. The need to solve this problem arises in the processing of reports on the hauling results of commercial fishing boats.
Consider the problem of sorting a large number of large records according to some key field, so that the records end up in sorted order on disk. The conventional methods of solving this problem are merge-based. Possib...
详细信息
Consider the problem of sorting a large number of large records according to some key field, so that the records end up in sorted order on disk. The conventional methods of solving this problem are merge-based. Possibilities are explored when particular partition-based strategies are employed that make use of distributional properties of the data. The major issues arise out of the desirability of overlapping essentially all computation and I/O.
The paper presents the concept of a rectangular cellular array intended for sorting. For an algorithm proposed the sorting time depends linearly on the number of ordered words.
The paper presents the concept of a rectangular cellular array intended for sorting. For an algorithm proposed the sorting time depends linearly on the number of ordered words.
For various microcomputer subassemblies, MICOS/80, a multitasking and multiprocessing operating system, offers high-performance basic software that considerably simplifies the implementation of an application. The nuc...
详细信息
For various microcomputer subassemblies, MICOS/80, a multitasking and multiprocessing operating system, offers high-performance basic software that considerably simplifies the implementation of an application. The nucleus is an efficient dispatcher that supervises and controls the system activities and tasks. By way of a dispatcher call a task can use the service for I/O and file management. The program is completely re-entrant and works with interrupt control and several processors, thus providing a high-level system throughput and low overhead.
A Concurrent Pascal Machine implemented on the ACE (Adaptive computer Experiment) Multiprocessor System efficiently executes concurrent Pascal programs and makes use of Concurrent Pascal as a high level language for m...
详细信息
A Concurrent Pascal Machine implemented on the ACE (Adaptive computer Experiment) Multiprocessor System efficiently executes concurrent Pascal programs and makes use of Concurrent Pascal as a high level language for multiprocessors. The author presents and evaluates some of the microprogram implementation methods, a mechanism for implementation specially on multiprocessors, and results of some simulated trials.
暂无评论