This is the mail archive of the
mailing list for the GCC project.
Re: MEM_EXPRs vs MEM_DECLs
- From: Richard Henderson <rth at redhat dot com>
- To: Richard Kenner <kenner at vlsi1 dot ultra dot nyu dot edu>
- Cc: gcc-patches at gcc dot gnu dot org
- Date: Tue, 4 Dec 2001 09:49:56 -0800
- Subject: Re: MEM_EXPRs vs MEM_DECLs
- References: <10112041124.AA15280@vlsi1.ultra.nyu.edu>
On Tue, Dec 04, 2001 at 06:24:43AM -0500, Richard Kenner wrote:
> So when we address a field, we at a variable offset (e.g., there's an array
> reference in there), we have a NULL offset relative to the decl, but do have
> one relative to the field. adjust_address updates the *last* offset.
I think it'd be useful to have adjust_address update all of the
offsets. This way we could pre-compute the offsets all the way
up the chain, saving time during alias analysis.
It would also allow us to discover the decl and the decl's offset
after the fact.
> Then the code to check for overlap is basically the same for decls and
> fields: take the offset of the decl or field and add to the offset in the
> memref and see is they confict. This also handles the case of overlapping
> fields just like we handle overlapping decls.