Distinguishing
the three long
double formats
The table suggests the high cost of a relaxed standard. To accommodate a variety of computer architectures, the IEEE binary floating-point standard did not fully specify its extended formats, and as a result the long double types on the X86 and PA-RISC platforms differ slightly in range and significantly in precision. The long double type on the PowerPC does not conform to IEEE specifications for an extended type (because it offers no extra exponent range beyond double and a number's precision depends on its value), though it offers extra precision. Chapter 11 discusses the portability issues of long double.
NOTE 
On platforms such as PowerPC and PA-RISC that lack a hardware long double type, the specifications in Table 4 pertain to the most common implementation. For example, a C++ compiler on such platforms might also offer the option of using the 64-bit double type for long double. Consult your development environment documentation for details.
[Contents]
[Previous]
[Next]
Click the icon to mail questions or corrections about this material to Taligent personnel.
Copyright©1995 Taligent,Inc. All rights reserved.
Generated with WebMaker