This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.

Index Nav:	[Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav:	[Date Prev] [Date Next]	[Thread Prev] [Thread Next]
Other format:	[Raw text]

[PATCH, i386]: Fix PR target/46051: [4.6 Regression] ICE: in extract_insn, at recog.c:2110 with -mavx -ftree-vectorize

From: Uros Bizjak <ubizjak at gmail dot com>
To: gcc-patches at gcc dot gnu dot org
Cc: "H.J. Lu" <hjl dot tools at gmail dot com>, Richard Henderson <rth at redhat dot com>, Richard Guenther <rguenther at suse dot de>
Date: Sun, 17 Oct 2010 22:42:26 +0200
Subject: [PATCH, i386]: Fix PR target/46051: [4.6 Regression] ICE: in extract_insn, at recog.c:2110 with -mavx -ftree-vectorize

On Sun, Oct 17, 2010 at 6:36 PM, H.J. Lu <hjl.tools@gmail.com> wrote:

>>> Please fix the expanders that generate wrong patterns instead, i.e.
>>> vec_interleave_lowv4df.
>>
>> Something like attached, but there are probably more instances to fix.
>
> It works. ?Thanks.

There were similar problems with v8sf vec_interleave expanders.
Attached patch fixes them all.

2010-10-17  Uros Bizjak  <ubizjak@gmail.com>

	PR target/46051
	* config/i386/sse.md (vec_interleave_highv4df): Fix third RTX of
	generated sequence to match *avx_vperm2f128<mode>_nozero.
	(vec_interleave_lowv4df): Ditto.
	(vec_interleave_highv8sf): Ditto.
	(vec_interleave_lowv8sf): Ditto.

testsuite/ChangeLog:

2010-10-17  Uros Bizjak  <ubizjak@gmail.com>
	    H.J. Lu  <hongjiu.lu@intel.com>

	PR target/46051
	* gcc.target/i386/pr46051.c: New test.

Patch was tested on x86_64-pc-linux-gnu {,-m32} and committed to mainline SVN.

Uros.

Attachment: p.diff.txt
Description: Text document

Index Nav:	[Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav:	[Date Prev] [Date Next]	[Thread Prev] [Thread Next]