portability is a bigger issue for numerical computation than for perhaps any other CommonPoint application system. There are two reasons. First, unlike higher-level frameworks, numerical computation is closely tied to features of the development environment and the underlying platform. Second, because advances in floating-point designs--especially the IEEE arithmetic standards--are strongly backward-compatible and because the numerical programming paradigm has barely changed in forty years of support for the types called float and double in C++, it is common for developers to bring to CommonPoint platforms software developed elsewhere a decade earlier.
The issue of portability arises in many forms. You might seek application compatibility between various CommonPoint platforms, despite their differences regarding floating-point computation. Or you might wish to develop numerical code to run on any system conforming to the proposed Floating-Point C Extensions. Or you might have thousands of lines of C code written using the Standard C numerical header math.h. This chapter gives portability strategies.
- A catalog of system dependencies
-
- Strategies for portability across
CommonPoint platforms
-
- Strategies for portability between
CommonPoint and other platforms
-
Click the icon to mail questions or corrections about this material to Taligent personnel.