Embedded Software and Floating Point Data
Integer arithmetic is used to create many embedded applications but sometimes their range they afford is not enough. At this time, embedded engineers have to consider floating point. Most embedded systems these days are built on 32 bit CPUs. In most cases, these CPUs provide enough computing power to perform arithmetic processing. But sophisticated operations present new challenges and here you have to turn to floating point data crunching.
In theory, floating point is quite simple to understand. Most embedded engineers are aware of it. The value consists of two integers- the mantissa and the exponent. And the best thing about floating point is that you can represent a large range of values. But when it comes to practice, there are some tradeoffs, especially in coding.
For example, even though most embedded developers are not 100 percent comfortable with binary representation, they understand it. Floating point numbers can also be shown in binary but it is a bit confusing. Moreover, floating point operations take time. And then the values are not exact. Many embedded engineers find this frustrating. The conclusion? Use floating point only if it is essential and then only after you have investigated and discarded all other ways to crunch integers.
