It's good software engineering practice to minimize gratuitous portability problems in the code. Techniques to minimize potential portability problems are:
- The integral and floating type sizes should be considered as minimums. Algorithms should be designed to continue to work properly if the type size increases.
- Floating point computations can be carried out at a higher precision than the size of the floating point variable can hold. Floating point algorithms should continue to work properly if precision is arbitrarily increased.
- Avoid depending on the order of side effects in a computation
that may get reordered by the compiler. For example:
a + b + c
can be evaluated as (a + b) + c, a + (b + c), (a + c) + b, (c + b) + a, etc. Parentheses control operator precedence, parentheses do not control order of evaluation.
If the operands of an associative operator + or * are floating point values, the expression is not reordered.
- Avoid dependence on byte order; i.e. whether the CPU is big-endian or little-endian.
- Avoid dependence on the size of a pointer or reference being the same size as a particular integral type.
- If size dependencies are inevitable, put a static assert in
the code to verify it:
static assert(int.sizeof == (int*).sizeof);
32 bit processors and operating systems are still out there. With that in mind:
- Integral types will remain the same sizes between 32 and 64 bit code.
- Pointers and object references will increase in size from 4 bytes to 8 bytes going from 32 to 64 bit code.
- Use size_t as an alias for an unsigned integral type that can span the address space. Array indices should be of type size_t.
- Use ptrdiff_t as an alias for a signed integral type that can span the address space. A type representing the difference between two pointers should be of type ptrdiff_t.
- The .length, .size, .sizeof, .offsetof and .alignof properties will be of type size_t.
Endianness refers to the order in which multibyte types are stored. The two main orders are big endian and little endian. The compiler predefines the version identifier BigEndian or LittleEndian depending on the order of the target system. The x86 systems are all little endian.
The times when endianness matters are:
- When reading data from an external source (like a file) written in a different endian format.
- When reading or writing individual bytes of a multibyte type like longs or doubles.
System specific code is handled by isolating the differences into separate modules. At compile time, the correct system specific module is imported.
Minor differences can be handled by constant defined in a system specific import, and then using that constant in an IfStatement or StaticIfStatement.