This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [PATCH][LTO] Stream zeroness of alias-sets
- From: Diego Novillo <dnovillo at google dot com>
- To: Richard Guenther <rguenther at suse dot de>
- Cc: gcc-patches at gcc dot gnu dot org
- Date: Thu, 6 Aug 2009 12:01:23 -0400
- Subject: Re: [PATCH][LTO] Stream zeroness of alias-sets
- References: <alpine.LNX.2.00.0908061747030.16347@zhemvz.fhfr.qr>
On Thu, Aug 6, 2009 at 11:49, Richard Guenther<rguenther@suse.de> wrote:
> This was supposed to fix the libstdc++ fails but appearantly doesn't.
> The C++ FE has a hack for its unwillingness to properly treat
> pointer-to-member function structs by assigning them alias-set zero.
Bummer.
> LTO better should preserve alias-set zero hacks from the frontends,
> so we stream that state out now.
Agreed.
> *************** unpack_ts_type_value_fields (struct bitp
> *** 1601,1606 ****
> --- 1601,1607 ----
> Â ÂTYPE_USER_ALIGN (expr) = (unsigned) bp_unpack_value (bp, 1);
> Â ÂTYPE_READONLY (expr) = (unsigned) bp_unpack_value (bp, 1);
> Â ÂTYPE_ALIGN (expr) = (unsigned) bp_unpack_value (bp, HOST_BITS_PER_INT);
> + Â TYPE_ALIAS_SET (expr) = bp_unpack_value (bp, HOST_BITS_PER_INT);
Need type cast to alias_set_type, but that shouldn't make a lot
of difference for 0 or -1.
It looks fine, even if it doesn't fix the libstdc++ problems.
Diego.