24 Jan
2015
24 Jan
'15
6:27 p.m.
Beman Dawes wrote:
In practice, the enhanced efficiency of aligned types is usually more important than worry about someday encountering an odd-ball architecture or use in an unaligned location. But it is an engineering tradeoff, so you can make the choice whichever way seems appropriate to your application.
The more practical problem is not oddball architectures, but the fact that when a field changes in size, everything after that field needs to be rechecked for alignedness. (You correctly identified this as a maintenance problem with the fixed offset approach I outlined, and it also applies here.)