"create your own type to encapsulate the underlying type (float or double, 32-bit or 64-bit), and use this type in your code. This can be done with a define is

"create your own type to encapsulate the underlying type (float or double, 32-bit or 64-bit), and use this type in your code. This can be done with a define is C/C++ or a class in Java." No. A Java class would certainly have worse performance than just float or double, since each instance would be individually heap allocated, have the usual per-Object overhead. Better use text preprocessing (or just "double") than that. (otherwise, this is all fine advice)