Hello everyone,


I heard when define a struct, it is important to make the size of the struct memory aligned.

For example,

struct foo is better than struct goo, even if we only use member p.

Allocate 3 bytes to align with memory. But I do not know why make memory aligned is better. Any reasons? (for example, prevent memory from running out?)


thanks in advance,
George

struct goo is better than struct foo. Leave padding and alignment issues to the compiler, it knows better than you what to do.

Anyway, the reason why aligned memory is sometimes better is because some processors are optimized for aligned access of words (it's the common case), and faster access for aligned bytes comes naturally from that.

In days gone by that was important, but I'm not so sure now.

The reason for such an approach in the structures was (and still is) that numeric calculations were performed faster if they started on a word boundary, hence padding was occasionally required. This goes back to my days of Fortran, Pascal and PL/1 where structures would declare individual bits, and so it was possible for an integer to be declared at a very inconvenient location.

Looking at C the language does the padding automatically for you. If you define a structure with a char, an int, a char and an int, use sizeof to see the size of that structure. Now define a structure with two intsfollowed by two chars and use sizeof again. The first structure will add padding so that the ints are on the proper boundaries, whilst the second doesn't need the padding.

Internal padding in the structure is platform dependent. Under ordinary circumstances you should not worry about alignment, since the compiler will handle it automatically.

However, if you have lot of elements in a structure definition and have the urge to save memory, then grouping like types together will usually reduce the amount of padding that occurs. (Do some diagnostics and printf the starting addresses of items in the structure to see if there are any improvements.)

There are various obscure reasons for it. The memory pipelines, internal and external caches and so-forth in a modern CPU are "wide," so they actually transfer a chunk of memory to or from the CPU each time. Compiler-writers are aware of the capabilities and the preferences of various types of hardware and they do a lot of things for you.

About the only place where you might need to give such things "conscious thought" is when you are matching your code to the data-format of some external data source or file. In that case, you might need to get very specific about exactly which byte goes where, whether they're "big-endian" or "little-endian," and more.