Re: Documentation for __aligned__ out of date?

[Jim Wilson <wilson at specifixinc dot com>]

Daniel Jacobowitz wrote:
> drow@nevyn:/big/codesourcery% echo 'int x __attribute__((__aligned__(1)));' >qqqq.c
>         .comm   x,4,1

That is a bug in the implementation of the aligned attribute. 
Apparently one that has been around for a long time unnoticed.  I tried 
2.8.1 and it fails there also.

If you try this with a structure type, you will see that it works as 
documented.

The reason for not letting aligned decrease alignment is because that 
will result in code that will fail on any target that requires aligned 
memory references.  The only way that we can support unaligned memory 
references is with the packed attribute, so we allow one to reduce 
alignment only if packed is used.  If you want to reduce alignment to 1, 
you use packed by itself.  If you want to reduce alignment to a value 
other than 1, then you use packed in conjunction with aligned.

One might ask why the aligned attribute doesn't give a warning when you 
give a too small value.  I think the reasoning here is that you only 
used aligned when you want a minimum alignment, and if the target 
happens to give you a larger alignment, then there is no problem, and 
hence no need to give a warning or error.  If you want a specific 
alignment, then you have to use packed with or without aligned.
--
Jim Wilson, GNU Tools Support, http://www.SpecifixInc.com