GB/T 17966-2024Information technology - Microprocessor systems - Floating-point arithmetic (English PDF)

Foreword

This document is in accordance with the provisions of GB/T 1.1-2020 "Guidelines for standardization work Part 1: Structure and drafting rules for standardization documents" Drafting.

This document replaces GB/T 17966-2000 "Binary floating-point operations for microprocessor systems".

In addition to structural adjustments and editorial changes, the main technical changes are as follows.

a) Added the description of the floating-point arithmetic programming environment considerations in the scope (see 1.5);

b) Added binary and decimal interchange format encoding content (see Chapter 5);

c) Changed the specific provisions of rounding, added the property description of rounding direction and the dynamic properties of rounding (see 6.1, 6.2, 6.3);

d) Added rules for various floating point operations, including binary and decimal format conversion, exponent calculation, silent calculation, signal the contents of computational and non-computational operations (see 7.2, 7.5, 7.6, 7.7);

e) Added description of exceptions and flags in general exception handling (see 9.1);

f) Modified the provisions on some abnormal situations that trigger self-traps (see 9.2 and 9.3);

g) Added some suggested operations and expression evaluation for floating-point operations (see Chapters 11 and 12).

This document is modified to adopt ISO/IEC 60559:2020 "Information technology microprocessor systems floating-point operations".

Compared with ISO/IEC 60559:2020, this document has made the following structural adjustments.

--- Added a chapter on "Normative References", and the subsequent chapter numbers are postponed accordingly;

--- Modified 6.3, 6.12, 7.2, Chapter 10, and the suspension section of 10.2;

---Appendix A and Appendix B correspond to Appendix B and Appendix C of ISO/IEC 60559:2020 respectively.

The technical differences between this document and ISO/IEC 60559:2020 and their reasons are as follows.

--- Deleted 1.6 "Use of auxiliary verbs", which is not applicable to my country's national conditions.

The following editorial changes were made to this document.

--- Move Appendix A (informative) "References" to after the last appendix;

--- Deleted the informative Appendix D.

Introduction

This document provides rules for performing floating-point operations that produce results independent of the hardware, software, or a combination of hardware and software used to perform the operations.

For the operations specified in the normative portion of this document, numerical results and anomalies are uniquely determined by the values of the input data, operation, and target, and are determined by the user.

control.

This document defines a system that can commercialize binary and decimal floating-point operations. The requirements met by this document include the following content.

a) facilitate the migration of existing programs from other types of computers and from computers that partially conform to this document to computers that fully conform to this document;

Computer.

b) Enhance the compatibility and safety of users and programmers who are not experts in numerical methods when writing complex numerical programs.

c) Encourage experts to develop stable and efficient numerical programs that can be ported by simply editing and recompiling on any computer that complies with this document and has sufficient capacity. Combined with language control, this can be achieved on all systems that comply with the standard.

The procedure produces the same result.

d) Provide direct support for.

1) Execution time anomaly diagnosis;

2) Handle exceptions more smoothly;

3) Implement interval algorithms at reasonable cost.

e) Provide conditions for the development of.

1) Common basic functions, such as exp or cos;

2) High-precision (multi-character) operations;

3) Coupled numerical and symbolic algebraic computations.

f) Allow rather than exclude further optimization and extension.

In the programming context, this document will also serve as a basis for dialogue between the mathematical community and programming language designers. It is hoped that in the coming years A programming language that can control the evaluation of expressions and abnormal results can be introduced so that programs can be written that run on all systems that conform to the standards and However, people realize that practicality and safety are sometimes in opposition in programming languages, just as efficiency and reliability are sometimes in opposition.

The same is true for portability.

Therefore, programming language designers are expected to use the full set of operations, precision, and exception controls described in this document as a framework to provide programmers with it is also hoped that programming language designers will be able to use this document as a framework to control the ability of expressions and exceptions in a portable manner.

Provides extensions in a completely portable manner.

Appendix A presents programming environment features for debugging support.

Information technology - Microprocessor systems - Floating point operations

1 Scope