An assembly language subroutine DYNAL has been programmed to provide dynamic allocation of array storage used by subroutines written in PDP-11 FORTRAN IV. DYNAL can dynamically allocate up to 16 arrays in the free cor...
详细信息
An assembly language subroutine DYNAL has been programmed to provide dynamic allocation of array storage used by subroutines written in PDP-11 FORTRAN IV. DYNAL can dynamically allocate up to 16 arrays in the free core between DOS MONITOR and the program low address (or low address of the largest overlay). Two example program situations are discussed to demonstrate the usefulness and applicability of DYNAL.
The purpose of this paper is to consider the problem of validation as it applies to the development of discrete digital simulation models. The described validation techniques, discussed in the context of model develop...
详细信息
The purpose of this paper is to consider the problem of validation as it applies to the development of discrete digital simulation models. The described validation techniques, discussed in the context of model development, are utilized during the developmentf of a simulation model of the Univac 1108 Exec 8 computer system at the Marshall Space Flight Center in Huntsville, Alabama. These techniques are of a general nature and are demonstrated with examples from the Univac 1108 Exec 8 simulation model development.
The performance measurement and analysis of software operating systems which extend basic computing machinery is discussed. The description of an external monitoring technique which facilitates the correlation of hard...
详细信息
The performance measurement and analysis of software operating systems which extend basic computing machinery is discussed. The description of an external monitoring technique which facilitates the correlation of hardware events with software functions without the need for software monitors is presented. A time-related Event is defined to provide the basis for the technique used to implement the monitor system. In addition, Event analysis methods are introduced which allow a software system execution profile to be constructed.
In several existing virtual-memory systems, the memory hierarchy is managed by a policy of the global LRU type. Two algorithms for restructuring programs to be executed in such an environment are described. The princi...
详细信息
In several existing virtual-memory systems, the memory hierarchy is managed by a policy of the global LRU type. Two algorithms for restructuring programs to be executed in such an environment are described. The principles forming the bases of these algorithms, which try to tailor the referencing behavior of programs to the one expected by the global LRU policy so as to improve their paging performance, are stated in general terms. The results obtained in a series of experiments on the reference strings generated by a real program under various inputs confirm the expectations about the practical applicability and the performance improvements of the new algorithms.
An objective function is provided and an approach is outlined to heuristically scheduling production oriented batch computersystems based upon the existence of a fast (approximately 100 times faster than real time) m...
详细信息
An objective function is provided and an approach is outlined to heuristically scheduling production oriented batch computersystems based upon the existence of a fast (approximately 100 times faster than real time) means for predicting the performance of a given system for a given schedule.
A technique is presented for instrumenting Concurrent Pascal systems at the component level. The key technical problem is that of implanting in each component a unique index which can be matched to the name of the com...
详细信息
A technique is presented for instrumenting Concurrent Pascal systems at the component level. The key technical problem is that of implanting in each component a unique index which can be matched to the name of the component by the instrumentation package. The implantation technique requires two monitors for parent-parent synchronization and parent-offspring synchronization. An additional monitor provides percent utilization statistics by component and a report process periodically produces summary figures. The entire procedure can be automated. The source Concurrent Pascal System can be read by an instrumentation program and the Instrumented Source Program produced as output. No modifications are required to the Concurrent Pascal Compiler, hence the same compiler can be used to compile and run either the instrumented or uninstrumented version of the system.
This article compares the performance of two page replacement algorithms (RAs) about which little system performance measurement data is available. They are: the Global Least Recently Used (LRU) and the Local LRU with...
详细信息
This article compares the performance of two page replacement algorithms (RAs) about which little system performance measurement data is available. They are: the Global Least Recently Used (LRU) and the Local LRU with fixed and equal size main memory buffer allotted to each task. The number of page faults caused during execution of programs under each RA is used as an inverse criterion for its effectiveness.
This article develops the theory of the statistical dynamics of a fixed degree of multiprogramming Markov process model. The model is a generalization of Buzen's central server model and permits the inclusion of s...
详细信息
This article develops the theory of the statistical dynamics of a fixed degree of multiprogramming Markov process model. The model is a generalization of Buzen's central server model and permits the inclusion of such features as multiple central processing units and direct access storage devices with rotational position sensing and sector queueing.
Presentation of an approximate analysis for a cyclic queueing model that has one station with S parallel servers. The approximation uses a single server with a rate equal to the combined service rate of the S servers....
详细信息
Presentation of an approximate analysis for a cyclic queueing model that has one station with S parallel servers. The approximation uses a single server with a rate equal to the combined service rate of the S servers. A comparison with the exact analysis of the cyclic queueing network shows that the approximation works well provided the number of jobs in the system is greater than twice the number of parallel servers, S. The approximate analysis is computationally simpler.
In a typical multiprogramming operating system, there are many functions that the central processor unit must perform that are classified as overhead. This paper studies the effects of attaching microprocessors to a m...
详细信息
In a typical multiprogramming operating system, there are many functions that the central processor unit must perform that are classified as overhead. This paper studies the effects of attaching microprocessors to a medium to large scale computer system to perform these overhead functions. The aspects investigated are those of process management, memory management, and some of the input/output (I/O) management functions. These functions are investigated using a configuration with the microprocessor(s) attached to one (or more) of the I/O channels. The central processor then sends the information to the microprocessor unit at the same time that this information would have been posted to the operating system's internal data structures. The microprocessor maintains this information, and, on demand, provides the central processor with the responses as though the central processor had computed them. This configuration requires that the microprocessors be totally self-contained. That is, they must have sufficient memory space.
暂无评论