This is the mail archive of the
gcc-bugs@gcc.gnu.org
mailing list for the GCC project.
[Bug debug/36060] [4.3/4.4 Regression] Too big stack requirements of cc1plus during GC
- From: "jakub at gcc dot gnu dot org" <gcc-bugzilla at gcc dot gnu dot org>
- To: gcc-bugs at gcc dot gnu dot org
- Date: 27 Apr 2008 11:06:24 -0000
- Subject: [Bug debug/36060] [4.3/4.4 Regression] Too big stack requirements of cc1plus during GC
- References: <bug-36060-87@http.gcc.gnu.org/bugzilla/>
- Reply-to: gcc-bugzilla at gcc dot gnu dot org
------- Comment #2 from jakub at gcc dot gnu dot org 2008-04-27 11:06 -------
As most of the nested calls are due to following die_sib, out of the first
almost 46000 frames I've uped, there were:
20273 gt_ggc_m_10die_struct ((*x).die_sib);
8878 gt_ggc_m_19VEC_dw_attr_node_gc ((*x).die_attr);
8877 gt_ggc_m_10die_struct ((*x).base.vec[i0].dw_attr_val.v.val_die_ref.die);
7220 gt_ggc_m_10die_struct ((*x).die_child);
the best fix is IMHO to mark die_sib as chain_next. Unfortunately, the DIEs
are linked circularly through the die_sib field, so chain_next doesn't work
properly for that. I'll post a patch which introduces chain_circular
alternative to chain_next. With this patch, cc1plus with this testcase on
ppc64-linux native
compiles even with ulimit -s 500, which is 28 times lower stack usage than
before.
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=36060