Sidney B. Gasser · 1972
Sandia Laboratories has established a consulting service designed to improve the overall efficiency of Fortran production programs in use by the scientific community. Since this service was inaugurated in May, 1971, approximately 40 jobs have been reviewed for optimization with varying degrees of success. In several instances it was not possible to improve the operating efficiency of a given program. However, in most cases it has been possible to achieve worthwhile reductions in the use of critical machine resources (primary attention is directed at use of central processor time and allocation of central memory). Average reduction of CPU usage is approximately 30% with one recorded instance of a 97% reduction.This paper addresses the problems encountered in launching and promoting a program optimization service in an open shop environment among users with widely varying levels of programming to indicate the method used to isolate and correct the sections of code that preempt the vital computer resources.
Cited by 0Open access
Sidney B. Gasser · ACM SIGCSIM Installation Management Review · 1972
Sandia Laboratories has established a consulting service designed to improve the overall efficiency of Fortran production programs in use by the scientific community. Since this service was inaugurated in May, 1971, approximately 40 jobs have been reviewed for optimization with varying degrees of success. In several instances it was not possible to improve the operating efficiency of a given program. However, in most cases it has been possible to achieve worthwhile reductions in the use of critical machine resources (primary attention is directed at use of central processor time and allocation of central memory). Average reduction of CPU usage is approximately 30% with one recorded instance of a 97% reduction.This paper addresses the problems encountered in launching and promoting a program optimization service in an open shop environment among users with widely varying levels of programming to indicate the method used to isolate and correct the sections of code that preempt the vital computer resources.
Cited by 0
Infotech Limited (Maidenhead) · 1976
Cited by 0
Infotech International. Maidenhead · CERN Document Server (European Organization for Nuclear Research) · 1976
Cited by 0
Elsevier eBooks · 2006
Cited by 0
Michael Wetter · OSTI OAI (U.S. Department of Energy Office of Scientific and Technical Information) · 1998
GenOpt is a generic optimization program for nonlinear, constrained optimization. For evaluating the objective function, any simulation program that communicates over text files can be coupled to GenOpt without code modification. No analytic properties of the objective function are used by GenOpt. Optimization algorithms and numerical methods can be implemented in a library and shared among users. GenOpt offers an interlace between the optimization algorithm and its kernel to make the implementation of new algorithms fast and easy. Different algorithms of constrained and unconstrained minimization can be added to a library. Algorithms for approximation derivatives and performing line-search will be implemented. The objective function is evaluated as a black-box function by an external simulation program. The kernel of GenOpt deals with the data I/O, result storage and report, interlace to the external simulation program, and error handling. An abstract optimization class offers methods to interface the GenOpt kernel and the optimization algorithm library.
Cited by 49Open access
Qiming Sun · Theoretical and computational chemistry · 2025
Cited by 1
Tauson, Peter O. · Defense Technical Information Center (DTIC) · 1967
This final report describes the work performed on an experimental additive heat calorimeter, designed and tested under Air Force Contract No. F33615-67-C-1473. A number of parameters were explored to establish the importance of design configurations of the calorimeter and its response to the control system while maintaining the desired readout accuracy. Fundamentally the two goals were: (1) Establish the nature of heat flow paths, and their relative importance in design and operation of the calorimeter. (2) Determine the response of the calorimeter control system to the changes in quality, and the time required to stabilize at a readable level. (Author)
Cited by 0
Кукуруза, А. О., Павлюк, Д. П., Сеник, В. В. et al. · ELARTU (Ternopil National Technical University) · 2019
Cited by 0Open access
Hiroshi Hoshino, Masahiro Esashi, Kiyoshi Agusa et al. · Lecture notes in control and information sciences · 2006
Cited by 1
Brauckmann, Alexander · ERA · 2025
The rise of specialized hardware and high-level domain-specific languages (DSLs) in the tensor domain has created new optimization challenges. First, a legacy code gap emerges, as vast amounts of code written in low-level languages like C cannot leverage modern accelerators. Second, an optimization gap exists because compilers for modern tensor DSLs rely on incomplete optimization rules, failing to achieve the full available performance. Manually addressing either of these gaps is an error-prone and unsustainable engineering effort. This thesis demonstrates that program synthesis techniques offer a powerful, automated way to address both challenges. It presents a series of such methods in order to lift numeric programs from legacy codebases to high-level tensor DSLs and subsequently, to transform high-level tensor DSL programs into computationally more efficient equivalents. The first method, MlirSynth shows the feasibility of program synthesis in the tensor DSL domain, using enumerative search to automatically lift programs from C to high-level dialects in the MLIR compiler infrastructure, resulting in significant geometric mean speedups of 3.4x on CPUs and 21.6x on TPUs over traditional compilation approaches. Second, to overcome the scalability and correctness limitations of MlirSynth's enumerative search, TENSORIZE introduces a novel top-down, symbolic program synthesis algorithm. By using a symbolic trace of the source program as a specification and recursively simplifying it with automatically generated tensor DSL program sketches, TENSORIZE achieves synthesis times that are orders of magnitude faster than prior work. This enables lifting of more complex programs, while guaranteeing correctness by construction. The third method, STENSO extends TENSORIZE's scalable program synthesis method from program lifting to superoptimization. Given a tensor DSL program, STENSO systematically explores the space of semantically equivalent programs, using a cost model and branch-and-bound algorithm to efficiently synthesize the highly optimized implementations. STENSO discovers powerful non-obvious optimizations that are missed by leading tensor compilers, resulting in geometric mean speedups of 3.8x over NumPy and 1.6x over JAX and PyTorch Inductor. In summary, this thesis presents program synthesis as a scalable and practical method for tensor program optimization. The developed techniques bridge the gap between legacy code and modern hardware, and the gap that rule-based systems leave when compiling high-level programs, enabling software to fully exploit the available performance on modern hardware.
Cited by 0Open access
Not Available · OSTI OAI (U.S. Department of Energy Office of Scientific and Technical Information) · 1980
The purpose of this program was to identify up to 10 good near-term opportunities for cogeneration in 5 major energy-consuming industries which produce food, textiles, paper, chemicals, and refined petroleum; select, characterize, and optimize cogeneration systems for these identified opportunities to achieve maximum energy savings for minimum investment using currently available components of cogenerating systems; and to identify technical, institutional, and regulatory obstacles hindering the use of industrial cogeneration systems. The analysis methods used and results obtained are described. Plants with fuel demands from 100,000 Btu/h to 3 x 10/sup 6/ Btu/h were considered. It was concluded that the major impediments to industrial cogeneration are financial, e.g., high capital investment and high charges by electric utilities during short-term cogeneration facility outages. In the plants considered an average energy savings from cogeneration of 15 to 18% compared to separate generation of process steam and electric power was calculated. On a national basis for the 5 industries considered, this extrapolates to saving 1.3 to 1.6 quads per yr or between 630,000 to 750,000 bbl/d of oil. Properly applied, federal activity can do much to realize a substantial fraction of this potential by lowering the barriers to cogeneration and by stimulating wider implementation of this technology. (LCL)
Cited by 4Open access
A. Ananda Rao, P. Radhika Raju · ACM SIGSOFT Software Engineering Notes · 2014
Program invariants can be detected using both static and dynamic approaches. Dynamic methods that complement static methods will be more useful and effective in analyzing software artifacts. Existing dynamic invariant detection techniques detect both relevant and irrelevant invariants due to the presence of relevant and irrelevant/unused variables in the program. Therefore, the speed and efficiency of the techniques are affected. Also, displaying properties about irrelevant variables distracts the user from concentrating on properties of relevant variables. Moreover, these irrelevant invariants do not contribute to the correctness of the program. Therefore, there is a need to overcome the above deficiencies. This is done in three phases in this paper. In the first phase, only the relevant variables are considered by ignoring irrelevant variables. In the second phase, relevant variables are classified into two different types as design variables and non-design variables by using proposed metrics. These metrics are validated analytically against Weyuker's principles. Similarly, relevant invariants are categorized into three types as design invariants, non-design invariants and hybrid invariants in the third phase. For this purpose a set of rules are proposed. This process enormously improves the speed and efficiency of dynamic invariant detection techniques. Also, it drastically reduces the effort that is required to validate program correctness.
Cited by 0
A. J. Dodd, K. E. Kadrinka, Matti J. Loikkanen et al. · Journal of Aircraft · 1990
A major programming effort to produce an interdisciplinary optimization program for the static, dynamic, and aeroelastic analysis of finite element models has been underway at Douglas Aircraft Company since 1985. The current status of the program is described together with test and production applications and plans for future developments
Cited by 24
R.L. Turner, C.F. Poor · OSTI OAI (U.S. Department of Energy Office of Scientific and Technical Information) · 1960
In the past two years a number of process optimization studies have been made with the aid of IBM machine programs. These programs have been designed to calculate optimum power levels and optimum goal exposures so that better insight could be gained for use in resolving plant operating plans. We feel that this approach has been highly successful. In January of 1960, a new program was written for the IBM 709 to incorporate the desirable features of the previous programs and to provide additional output data. This document is a description of the new programs.
Cited by 1Open access
Kraft,E. H. · Defense Technical Information Center (DTIC) · 1974
The development of a eutectic alloy composed of two ductile phases has proceeded. Cobalt additions to the gamma-alpha (Ni-37 w/o Cr-18 w/o W-0.1 w/o Al) eutectic improved phase alignment and increased strength at both room and elevated temperatures. No effect of cobalt on phase equilibria was observed. The gamma phase was substantiated as face centered cubic (FCC), while the crystallography of the alpha phase remained uncertain. Improvement of 73C eutectics in stability in thermal cyclic environments was sought by alloying and study of the mechanism of damage. Additions of aluminum, iron and nickel stabilized the FCC structure of the cobalt matrix of 73C eutectics, thus increasing stability in thermal cycling environments. Carbide size and volume fraction also had an influence on thermal cyclic stability, but their contribution was found to be less important than carbide perfection.
Cited by 0
Bao, Bin (1982 - ), Ding, Chen (1970 - ) · UR Research (University of Rochester)
Thesis (Ph. D.)--University of Rochester. Dept. of Computer Science, 2013.
Cited by 0Open access
Jens Knoop · Lecture notes in computer science · 1998
Cited by 11
Paul Odell · 1978
Cited by 0
smkavishwar-jpg · Zenodo (CERN European Organization for Nuclear Research) · 2026
Optimization Program
Cited by 0Open access