This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [sa]: Revert partial def stuff, use fake variables instead


Hello,

> >>>analysis are not affected by the way addresses are expressed, and having
> >>>to perform magic just to keep the results the same does not seem right
> >>>to me.
> >>
> >>What exactly are you proposing you should have to do in order for
> >>everything else to know you haven't changed anything?
> >
> >See my previous mail to Diego for concrete proposal (if it is too hard
> >to understand, I will prepare a complete patch in the evening; it should
> >not take too much time).
> 
> I was planning on doing more or less what you've proposed as part of the 
> structure aliasing stuff, but it still wouldn't help your problem.
> 
> ARRAY_REF doesn't have a tag associated with it because ARRAY_REF's 
> can't be of pointers (they are always of array_types), and thus, there 
> is no reason to have a tag there.
> 
> So you'd still have to create the tag in that case anyway.

why? If the information necessary for creating the virtual operands is
moved to the memory reference, nobody will need the tag for the pointer
(except perhaps for alias analysis, but it is obviously able to cope with
pointers without a tag).

Zdenek


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]