This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
[PATCH]: Fix PR tree-optimization/20926
- From: Daniel Berlin <dberlin at dberlin dot org>
- To: gcc-patches at gcc dot gnu dot org, dnovillo at redhat dot com
- Date: Mon, 11 Apr 2005 12:51:20 -0400
- Subject: [PATCH]: Fix PR tree-optimization/20926
We can't add_type_alias the var if it has subvars, we need to do the
subvars instead, otherwise, we'll get verify_ssa errors later when we
see the var instead of it's subvars.
Diego, since this is *always* the case, should i simply attempt to move
this logic into add_type_alias itself, so that if you call it with a var
that has subvars, it automatically does the right thing?
I wasn't sure whether i should force the callers to know what they are
doing, or just make the callee handle it for them.
--Dan
2005-04-11 Daniel Berlin <dberlin@dberlin.org>
Fix PR tree-optimization/20926
* tree-ssa-loop-ivopts.c (rewrite_address_base): Handle subvars.
Index: tree-ssa-loop-ivopts.c
===================================================================
RCS file: /cvs/gcc/gcc/gcc/tree-ssa-loop-ivopts.c,v
retrieving revision 2.59
diff -u -p -r2.59 tree-ssa-loop-ivopts.c
--- tree-ssa-loop-ivopts.c 9 Apr 2005 01:37:24 -0000 2.59
+++ tree-ssa-loop-ivopts.c 11 Apr 2005 16:48:48 -0000
@@ -4849,7 +4849,19 @@ rewrite_address_base (block_stmt_iterato
if (tag)
var_ann (new_ptr)->type_mem_tag = tag;
else
- add_type_alias (new_ptr, var);
+ {
+ subvar_t svars;
+ if (var_can_have_subvars (var)
+ && (svars = get_subvars_for_var (var)))
+ {
+ subvar_t sv;
+
+ for (sv = svars; sv; sv = sv->next)
+ add_type_alias (new_ptr, sv->var);
+ }
+ else
+ add_type_alias (new_ptr, var);
+ }
new_name = make_ssa_name (new_ptr, copy);
}