Hardware differences

Any software components with real-time performance constraints must be thoroughly tested with the exact hardware on which it will eventually run. The most obvious consideration is the fact that the physical components used for serial i/o are not identical on PCs and WID systems. Less obvious things that differ on the two platforms include the granularity of the clock ticks, which controls thread scheduling, the speed of data storage devices, and the throughput of characters to the console. It cannot be emphasised too strongly that without this testing, there’s no way of finding out whether these differences make a real difference.

While there’s no doubt that WINS/WINSCW and Visual C++ are an indispensable part of the total development process, and that debugging program logic and control paths is more efficiently done on a PC, hardware differences make these environments less useful for testing real-world performance. Experience tends to show that it is actually more difficult to develop test modules that work reliably under stress for the WINS/WINSCW development environments than it is for a target machine. This means that a module which fails to perform to specification under WINS/WINSCW may nevertheless be satisfactory when run on the target system.