 | |
Introduction
This paper focuses on the programming-related aspects of endianness. Its primary goal is to present
mechanisms for producing endian-independent source code that can be compiled for either big-endian or
little-endian environments from one source.
Readers unfamiliar with endianness as applied to programming (as opposed to eggs) will find some needed background in the next section. The "Overview" section outlines the remainder of this white paper.
What Is Endianness?
Nearly all computers today address their memory in units of 8-bit "bytes." They address larger storage
units--typically, but not always, called halfwords (16 bits), words (32 bits), and doublewords
(64 bits)--by giving the address of the byte at one end or the other of the storage unit.
Unfortunately, some computers, the "little-endians," use the address of the numerically least significant
byte, while others, the "big-endians," use the address of the numerically most significant byte for the
address of the multibyte storage unit[1]. The
Intel x86 family and Digital Equipment Corporation architectures (PDP-11, VAX, Alpha) are
representative little-endians, while the Sun SPARC, IBM 360/370, and
Motorola 68000 and 88000 architectures are big-endians. Still other architectures (PowerPC, MIPS, and
Intel's IA-64[2]) are capable of operating in
either big-endian or little-endian mode.
Consider a little-endian and a big-endian computer, each containing in its first 16 bytes of memory the number representing the address of the byte. Table 1-1 illustrates this situation. Regarded as an array of bytes, the situation is the same on both computers. Table 1-1. Memory as an Array of Bytes
If instead of considering the memory as an array of bytes we consider the same memory contents as four
32-bit words, the view that the two processors have is different, as shown in
Table 1-2 and
Table 1-3 In these tables, the bytes are grouped into
four byte words, which are shown in the Arabic numeral form with the most significant byte ("digit") on
the left. In Table 1-2 we put the word address column
on the right ("little end") because the computer uses the address of the least significant byte, the
byte on the right, to address the word. In Table 1-3,
the address column is on the left ("big end"), showing that the computer addresses the most significant
byte in word operations. As a result, the little-endian processor loading the 32-bit word at word
address 0 would obtain a different value (03.02.01.00) than the big-endian processor (00.01.02.03).
[3]
Table 1-2. Little-Endian Memory as an Array of Words
Table 1-3. Big-Endian Memory as an Array of Words
Overview
There are two fundamental sources of endian-related problems:
Most programmers will never encounter endianness problems. Some will never even write programs that will have to run on machines of different endianness. This paper is intended to help those who may have to deal with endianness issues either in developing a new product or porting an existing product. It offers advice on how to avoid the problems where possible and how to recognize and overcome the more difficult ones. Although most of the material is not operating system specific, it highlights the solutions that the Solaris operating environment offers. It also points out areas in which hardware assistance may be brought to bear in dealing with endianness problems. The paper is organized into three main sections as follows:
Finally, the last section in this white paper, "Operating System and Bi-Endian Architectures," discusses
issues involved in choosing the endianness for an operating system on a bi-endian system, one with
hardware that can support either endianness.
Notes appearing throughout this paper highlight the
practical application of the techniques and mechanisms discussed.
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| ||||||||||||
Print-friendly Version
