--- Comment #11 from rguenth at gcc dot gnu dot org 2007-04-12 10:16
---
Subject: Bug 31307
Author: rguenth
Date: Thu Apr 12 10:15:53 2007
New Revision: 123736
URL: http://gcc.gnu.org/viewcvs?root=gccview=revrev=123736
Log:
2007-04-12 Richard Guenther [EMAIL PROTECTED]
PR
--- Comment #12 from rguenth at gcc dot gnu dot org 2007-04-12 10:27
---
Fixed.
--
rguenth at gcc dot gnu dot org changed:
What|Removed |Added
--- Comment #13 from michael dot meissner at amd dot com 2007-04-12 20:18
---
How hard would it be to back port the change to 4.1.3 and 4.2?
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=31307
--- Comment #14 from pinskia at gcc dot gnu dot org 2007-04-12 22:32
---
(In reply to comment #13)
How hard would it be to back port the change to 4.1.3 and 4.2?
Why do you want to that, this is not a regression at all. I am tried of people
asking questions like this for missed
--- Comment #15 from rguenth at gcc dot gnu dot org 2007-04-12 22:36
---
4.2.0 wouldn't be too difficult (a svn merge of the change to 4.2.0 branch
succeeds without problems), but 4.1.3 has ineffective store copyprop (see
PR26135). Of course this is not a regression, so a backport is
--- Comment #5 from rguenth at gcc dot gnu dot org 2007-04-10 14:15 ---
There's one missed FRE opportunity in that we do not value-number
VIEW_CONVERT_EXPR__v16qi(D.6423) the same. This is because VIEW_CONVERT_EXPR
is tcc_reference I believe.
Created value VH.19 for VH.17.VH.18 vuses:
--- Comment #6 from rguenth at gcc dot gnu dot org 2007-04-10 14:37 ---
Reduced testcase:
typedef long long __m128i __attribute__ ((__vector_size__ (16),
__may_alias__));
typedef long long __v2di __attribute__ ((__vector_size__ (16)));
typedef char __v16qi __attribute__
--- Comment #7 from rguenth at gcc dot gnu dot org 2007-04-10 15:32 ---
store_copyprop is not able to optimize this because the two array refs
r.dst[0].i use different types for the index zero (one int, one unsigned long)
and one has operand2 and operand3 set but the other not and
--- Comment #8 from rguenth at gcc dot gnu dot org 2007-04-10 16:55 ---
Created an attachment (id=13348)
-- (http://gcc.gnu.org/bugzilla/attachment.cgi?id=13348action=view)
patch
Actually that breaks with some hashing. I have a slightly different approach
in
testing now. We are
--- Comment #9 from pinskia at gcc dot gnu dot org 2007-04-10 21:11 ---
(In reply to comment #8)
Created an attachment (id=13348)
-- (http://gcc.gnu.org/bugzilla/attachment.cgi?id=13348action=view) [edit]
patch
for one part we only set operands 2 and 3 if they are not
--- Comment #10 from rguenth at gcc dot gnu dot org 2007-04-10 21:38
---
Indeed.
--
rguenth at gcc dot gnu dot org changed:
What|Removed |Added
OtherBugsDependingO|
--- Comment #4 from pinskia at gcc dot gnu dot org 2007-03-22 01:38 ---
Inline version:
r.dst[0].i = MEM[base: d];
D.6423 = r.dst[0].i;
D.6449 = __builtin_ia32_paddusb128 (VIEW_CONVERT_EXPR__v16qi(D.6423),
VIEW_CONVERT_EXPR__v16qi(D.6423));
r.dst[0].i =
12 matches
Mail list logo
