Showing papers on "Benchmark (computing) published in 1977"
TL;DR: This paper gives the results of benchmark tests of several different microprocessors in terms of speed of execution, size of source program (number of statements) and of memory requirements for the compiled programs.
Abstract: This paper gives the results of benchmark tests of several different microprocessors. The results are given in terms of speed of execution, size of source program (number of statements) and of memory requirements for the compiled programs. An analysis of these results is given in terms of the microprocessor architectures.
9 citations
13 Jun 1977
TL;DR: A benchmark program was run on the baseline system and the model was used to predict CPU and I/O channel utilizations for the upgraded or reconfigured system and there was good agreement between predicted and measured utilizations.
Abstract: One possible use of an analytical model of a computer system is to predict the increase in performance that results from a proposed modification of the existing system. With this aspect in mind, experiments were conducted that hypothesized situations in which a manager was considering upgrading his system and was interested in estimating the performance improvements to weigh against the anticipated costs. The basic concept was to run a benchmark program on the baseline system and to validate the model parameters. After validation, the model was used to predict CPU and I/O channel utilizations for the upgraded or reconfigured system. Lastly, the benchmark was rerun on the reconfigured system and measured utilizations were compared with predicted values from the model. There was good agreement between predicted and measured utilizations when adding a channel to an IBM 370/155, when reallocating files and adding two channels to an IBM 370/155--2, and when adding a channel to an IBM 370/168--1.
3 citations
01 Jan 1977
TL;DR: The construction and use of a general purpose FORTRAN synthetic program is outlined, and the results of a benchmark in which it was used are given.
Abstract: A general purpose synthetic program may be used to good advantage in configuring a workload for use in benchmarking several machines. A good general purpose synthetic program should be able to do a variety of types of CPU and I/O processing in a variety of source languages, should be able to simulate interactive processing, and should be able to exercise the core management routines of a demand paged (virtual) system. The construction and use of a general purpose FORTRAN synthetic program is outlined, and the results of a benchmark in which it was used are given. The development of the workload used in the benchmark is also described.
3 citations
01 Nov 1977
TL;DR: A survey of methods for measuring the quality of object programs produced by the compiler, and their performance at execution time, and experiments with the use of those methods are presented.
Abstract: : One of the concerns of the compiler writer is the quality of object programs produced by the compiler, and in particular their performance at execution time. A survey of methods for measuring this performance, and experiments with the use of those methods, is presented. We examine two general categories of evaluation: comparative evaluation, in which benchmark programs are run on groups of language systems; and analytic evaluation, in which a single system is measured in terms determined by its own structure. Besides surveying the results of various evaluation experiments, we present in detail the results of a series of experiments on a particular language system (PDP11 ALGOL 68-S).
2 citations
TL;DR: The authors propose that benchmark work in conventional neutronics be reorganized somewhat, so that key steps in the reactor design and analysis process be systematically benchmarked.
Abstract: The authors propose that benchmark work in conventional neutronics be reorganized somewhat, so that key steps in the reactor design and analysis process be systematically benchmarked. Furthermore, they suggest that benchmark work in different areas be unified, and that new benchmark problems be developed for reactor safety and fusion power neutronics calculations.
1 citations
01 Jul 1977
TL;DR: An investment model based on mathematical programming is developed which considers the aspects of investment for hardware selection and combines these methods with an optimization model.
Abstract: The selection of an EDP configuration often fixes a firm to a single manufacturer for a long time and the capabilities of the chosen computer will continually influence the firm's organization. Only few approaches exist to give assistance to the investors by developing useful decision models based on the investment theory /11, 12/. The hardware selection methods /4, 13/ used up to now, like benchmark tests, don't meet these demands. In this paper an investment model based on mathematical programming is developed which considers the aspects of investment for hardware selection.Nevertheless, the present methods stay valid because their output can be used as dalta input for the optimization model. Therefore, a concept is proposed which combines these methods with an optimization model.
TL;DR: An investment model based on mathematical programming is developed which considers the aspects of investment for hardware selection and meets the demands of the hardware selection methods.
Abstract: The selection of an EDP configuration often fixes a firm to a single manufacturer for a long time and the capabilities of the chosen computer will continually influence the firm's organization. Only few approaches exist to give assistance to the investors by developing useful decision models based on the investment theory /11, 12/. The hardware selection methods /4, 13/ used up to now, like benchmark tests, don't meet these demands. In this paper an investment model based on mathematical programming is developed which considers the aspects of investment for hardware selection.
TL;DR: There are seven methods generally used to measure the performance of the proposed architecture: Timing, Mixes, Kernel Programs, Benchmark Programs, Synthetic Programs, Mathematical Models, and Simulation.
Abstract: Having selected a Microprogrammed Instruction Set (MIS), the need to have a method of evaluation immediately arises There are seven methods generally used to measure the performance of the proposed architecture These methods are: Timing, Mixes, Kernel Programs, Benchmark Programs, Synthetic Programs, Mathematical Models, and Simulation