任何编程语言的内存分配取决于编译器还是系统架构？

Question

Memory allocation of any programming language depends on compiler or the system architecture? If it depends on compiler, then what difference 32 bit/64 bit architecture makes? If it depends on architecture, then why the memory size of variables are constant for 16/32/64 bit architectures? What is the impact of slack bytes on the system architecture?

Answer 1

It's not either or. Compiler, system architecture and system conventions have an effect on how memory allocation works.

For one, the system architecture includes the size of pointers. So a compiler can't use more memory than a system provides, and can't address more bytes or have larger pointers than the CPU supports. (I mean actual RAM, and real addresses. Of course it can use Virtual Memory-like constructs to remap the addresses the programmer uses to something that CPU can handle).

Similarly, usually operating system vendors have conventions what a function call should be like: Whether parameters are passed on the stack or in registers. Whether parameters that are larger than a certain size should be copied onto the stack or provided as a pointer to a stack/heap object. If you want to call system functions, you will have to use these conventions.

However, beyond that, it is left to the compiler to decide details. Eg many Pascal compilers pass additional hidden pointers to enclosing scope into built-in functions defined in the same language. As long as you know code has been written using the same language, you can agree on differences in calling conventions within the limits the system architecture permits. Eg if you know that your software has to run on similar CPUs with stricter requirements, you can choose to apply these strict requirements even on CPUs that don't need it (like compiling code so everything is at least 2-byte-aligned so your MC68040 code will also run on the older MC68000 which had that requirement).