A Global Optimization Method for Numerical Programs

Abstract

The invention provides a global optimization method of a numerical program. The global optimization method of the numerical program includes the steps that a symbol execution technology is adopted toextract algebraic representations for expressing path constraint conditions and numerical calculation processes from a source program; all the algebraic representations are analyzed, and by using different algebraic transformation rules, the algebraic representations are converted to be in algebraic forms with lower numerical calculation errors; finally, all the algebraic representations are converted into corresponding code fragments, and the code fragments are combined to generate a target program. The global optimization method of the numerical program has the advantages that all the processes are automation processes, and no human intervention is needed; a program writer can better concentrate on service logic and does not need to care about detail operation related to numerical analysis. In this way, not only is the development efficiency improved, but also code logic is visual and easy to understand and maintain.

Description

technical field [0001] The invention belongs to the application field of computer numerical calculation and program analysis, and relates to a global optimization technology for numerical programs. Generate more accurate and efficient optimization programs than the original numerical programs. Background technique [0002] According to the IEEE745 standard, single-precision floating-point numbers occupy 32 bits, and double-precision floating-point numbers occupy 64 bits. Although many software developers regard them as real values ​​when developing programs, they can actually represent limited values. and discontinuous. Therefore, for most programs, floating-point numbers inevitably introduce numerical errors because they can only represent approximations of real numbers. In a source program, a large number of floating-point operations may be involved, and the errors of these floating-point operations will accumulate as the program progresses, eventually resulting in progr...

AI Technical Summary

Problems solved by technology

Because many such detailed operations are related to the hardware platform, they are not only very cumbersome, but also extremely error-prone, resulting in the main logic of the numerical program being covered up by these detailed operations, which reduces the readability of the source code, which is not conducive to the correct development of the code. It is also not conducive to subsequent code maintenance

Images