This is the mail archive of the
gcc-bugs@gcc.gnu.org
mailing list for the GCC project.
[Bug target/13722] [3.4/3.5 regression] [ia64] ICE in push_secondary_reload
- From: "wilson at specifixinc dot com" <gcc-bugzilla at gcc dot gnu dot org>
- To: gcc-bugs at gcc dot gnu dot org
- Date: 24 Jan 2004 20:17:52 -0000
- Subject: [Bug target/13722] [3.4/3.5 regression] [ia64] ICE in push_secondary_reload
- References: <20040117230657.13722.schwab@suse.de>
- Reply-to: gcc-bugzilla at gcc dot gnu dot org
------- Additional Comments From wilson at specifixinc dot com 2004-01-24 20:17 -------
Subject: Re: PR 13722 candidate fix
I tried a bootstrap with all HAVE_POST_* macros turned off, and it still
failed.
I tried adding debug_rtx{_list,} calls to ia64_split_tmode_move and I
noticed a few things.
You are adding REG_INC notes when a MEM operand is a dest, but not when
a src. An auto-inc address is still a side-effect even when it is used
in a source operand. This is easy to fix by copying some code, I'll
have to try another bootstrap with this change.
I noticed that we are looking REG_*_P flag bits and alias info for
registers, because you are generating new ones with with gen_rtx_REG
without copying any info over from the old ones. It isn't clear if this
is an optimization problem, I suspect not. I think this is only a
potential performance loss.
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13722