Some computer systems have alignment rules that must be taken into consideration when constructing complex types. The complex types may be organized such that all base types start on an offset which is a multiple of their won sizeFor example, the offset of a 16-bit short, within a complex type could be 0, 16, 32, 48 or any other multiple of 16.
My question is, how can I determine whether there is word alignment role exsiting in my computer system?
Thanks.
This is not a computer dependency, it is a language/compiler issue. Check your compilers documentation.
Wave upon wave of demented avengers march cheerfully out of obscurity unto the dream.
Really? But in some documents I find it is computer system dependent. Could you please recommend some documents or information about it? Thanks.Originally posted by adrianxw
This is not a computer dependency, it is a language/compiler issue. Check your compilers documentation.
Computers have "natural" boundaries. When a 32 bit machine loads a 32 bit register, it does it quickly, so many compilers align variables on 32 bit boundaries. However, some languages are specified to byte align data.
A Fortran COMMON block for example may begin on a 32 bit boundary, but within the COMMON block, the individual data elements are adjacent, (hence with Fortran coding, it is good practice to list the elements in size order). It HAS to be that way because you can declare the data differently in different functions, (i.e. In function a() your common block has a simple 32 bit integer, in function b() the common block is declared as a 4 byte array). You cannot allow the compiler to optimise that, because you do not know if an external library has been optimised in the same way.
It takes more time and fiddling about to get the data onto non natural boundaries, and is therefore avoided. I can select the packing of my structures using compiler directives or pragmas, from MSDN...
>>>
Depending on the/Zp compiler option or the pack pragma, intervening space can be introduced to align member data on word or doubleword boundaries.
<<<
... I am unaware of any general documents, all these types of thing tend to be vendor specific.
Wave upon wave of demented avengers march cheerfully out of obscurity unto the dream.
But I need to directly access each attribute of the complex types and modify the values of them. So I must know the address of each attribute before I modify it. How can I determine whether the word alignment role existing before I compute the address?Originally posted by Salem
Yeah, and the compiler takes care of it for you, so you don't have to.
Last edited by corn; 06-19-2003 at 03:03 AM.
>>> modify the values of them.
As Salem says, the compiler will sort that out. Whilst modifying the values, you do not need to know if the member x is a byte, a word or a double word. If you want to know how it's done, then I would say again...Code:Class.x = 5;
>>> Check your compilers documentation.
... if you want to change it, look up the options and/or pragmas available to you, these will also be found in in your compilers documentation. I would advise against changing the alignment unless you have a really good reason, because the compilers optimisation/applications performance can take a real hit.
Wave upon wave of demented avengers march cheerfully out of obscurity unto the dream.
adrianxw:
Yes, it is a easy way to modify value using "Class.x = 5;". However, the problem is that I don't know the name of these attributes when I modify them.
Suppose there are a class like this:
I design a dll to modify the values of these objects. Before that, the intial address of this object instance and the structure of that object is known. So I need compute the address of each attribute and then modify it. It's obvious that I cannot do it as:Code:class AObject{ int a; double b; char c; }; AObject A1=new AObject();
I have to know whether I need take the word alignment rule into account when I compute the address. But how?Code:A1.b=0.5;
Thanks.
All of your suggestions are helpful.
