This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Proposal: GCC core changes for different address spaces
- From: Richard Henderson <rth at redhat dot com>
- To: Martin Koegler <mkoegler at auto dot tuwien dot ac dot at>
- Cc: gcc at gcc dot gnu dot org
- Date: Sat, 23 Apr 2005 11:36:59 -0700
- Subject: Re: Proposal: GCC core changes for different address spaces
- References: <20050423171822.GA7895@ahab.auto.tuwien.ac.at>
On Sat, Apr 23, 2005 at 07:18:22PM +0200, Martin Koegler wrote:
> For implementing the type attributes, I propose:
> Add the field "tree type;" to "struct mem_attrs". This field holds the type, if present,
> or 0, if no type information is available.
>
> To access it, I propose:
> #define MEM_TYPE(RTX) (MEM_ATTRS (RTX) == 0 ? 0 : MEM_ATTRS (RTX)->type)
I'd prefer that there be a flags field, akin to SYMBOL_REF_FLAGS.
You're going to have to check this property quite often; I'd rather
the data be cached somewhere easy to get at, rather than burried in
some data structure 4 levels removed.
The flags field of the mem could be initialized by a target function
that receives the decl or type, as received by set_mem_attributes.
> *What to do, if the reload pass copy the REG_EXPR to a
> MEM_EXPR (set_mem_attrs_from_reg)?
I would expect you to do nothing.
r~