This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Recent removal of the 'aux' field in tree_common
- From: Diego Novillo <dnovillo at redhat dot com>
- To: Geoff Keating <geoffk at geoffk dot org>
- Cc: gcc at gcc dot gnu dot org
- Date: Thu, 28 Feb 2002 12:17:59 -0500
- Subject: Re: Recent removal of the 'aux' field in tree_common
- Organization: Red Hat Canada
- References: <20020228160600.GA5835@tornado.toronto.redhat.com> <jmsn7l8s6t.fsf@desire.geoffk.org>
On Thu, 28 Feb 2002, Geoff Keating wrote:
> The idea behind the PCH code is that it runs through every data
> structure in the compiler, and writes them to disk. This means that
> it needs to know where every pointer is and the size of every object.
> It also has to read all the data structures from disk, which means it
> needs to know how to allocate every kind of data structure, and it
> needs to know how to de-allocate any structure that is replaced.
>
Thanks. So, distributing the annotation fields as I proposed in
my original message, shouldn't upset PCH, right? ie,
> struct tree_ann_def
> {
> /* Basic block that contains this tree. */
> basic_block bb;
>
> /* For _DECL trees, list of references made to this variable. */
> varray_type refs;
>
> /* Most recent definition for this symbol. Used when placing FUD
> chains. */
> varref currdef;
>
> /* Immediately enclosing compound statement to which this tree belongs. */
> tree compound_stmt;
> };
>
> [ ...]
>
> - tree_ann.bb and tree_ann.compound_stmt should go in statement
> nodes. Those are the only tree nodes that can be mapped to a
> basic block.
>
> - tree_ann.refs and tree_ann.currdef should go in decl nodes.
Thanks. Diego.