This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [PATCH] PR tree-optimization/16808: verify_ssa failed: Missingdefinition for SSA_NAME
- From: Jeffrey A Law <law at redhat dot com>
- To: Jie Zhang <zhangjie at magima dot com dot cn>
- Cc: gcc-patches at gcc dot gnu dot org, amacleod at redhat dot com
- Date: Tue, 03 Aug 2004 16:30:44 -0600
- Subject: Re: [PATCH] PR tree-optimization/16808: verify_ssa failed: Missingdefinition for SSA_NAME
- Organization: Red Hat, Inc
- References: <410EDA98.6080503@magima.com.cn> <410FA7EE.2070106@magima.com.cn>
- Reply-to: law at redhat dot com
On Tue, 2004-08-03 at 08:57, Jie Zhang wrote:
> Hi,
>
> I investigated the bootstrap problem of my previous patch. I made a
> mistake that I thought the following two statements
>
> tree *stmtp = bsi_stmt_ptr (bsi);
> tree stmt = *stmtp;
>
> were equivalent to
>
> tree stmt = bsi_stmt (bsi);
> tree *stmtp = &stmt;
>
> Here is a new patch, bootstrapped and checked with c and c++. I also
> simply add the testcase, not its patch. Could someone help to bootstrap
> and check with java and ada? Any comments? Thanks.
>
>
> regards
> --
> Jie
>
>
> 2004-08-03 Jie Zhang <zhangjie@magima.com.cn>
>
> PR tree-optimization/16808
> * tree-ssa-ccp.c (execute_fold_all_builtins): Rename the affected vars
> in V_MAY_DEFS after folding builtins.
> (struct tree_opt_pass pass_fold_builtins): Add TODO_rename_vars in
> todo_flags_finish.
>
> * testsuite/gcc.c-torture/compile/20040803-1.c: New testcase.
This seems like the wrong way to approach this problem and rather
heavy-weight. We're really trying to get away from calling back
into the SSA renamer as often as we currently do.
I suggest waiting until Andrew's immediate uses work is done -- it
may fix this problem as a side effect. At the least it will make
updating the SSA graph trivial for this kind of thing.
Jeff