--
pinskia at gcc dot gnu dot org changed:
What|Removed |Added
Target Milestone|--- |4.3.0
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=31995
--- Comment #8 from pinskia at gcc dot gnu dot org 2007-05-22 00:09 ---
Subject: Bug 31995
Author: pinskia
Date: Mon May 21 23:08:54 2007
New Revision: 124927
URL: http://gcc.gnu.org/viewcvs?root=gccview=revrev=124927
Log:
2007-05-21 Andrew Pinski [EMAIL PROTECTED]
PR
--- Comment #7 from pinskia at gcc dot gnu dot org 2007-05-22 00:09 ---
Fixed.
--
pinskia at gcc dot gnu dot org changed:
What|Removed |Added
Status|ASSIGNED
--- Comment #2 from pinskia at gcc dot gnu dot org 2007-05-20 09:40 ---
Created an attachment (id=13587)
-- (http://gcc.gnu.org/bugzilla/attachment.cgi?id=13587action=view)
Patch which fixes the problem
I think this patch is correct, the issue is two fold, one we don't consider
--- Comment #3 from pinskia at gcc dot gnu dot org 2007-05-20 10:28 ---
There were no regressions in the vect.exp testsuite on powerpc-darwin. Plus a
quick s/x+i+1/y+i+1/ on this testcase gives the correct answer of
t.c:19: note: not vectorized: can't determine dependence between
--- Comment #4 from pinskia at gcc dot gnu dot org 2007-05-20 21:44 ---
Created an attachment (id=13591)
-- (http://gcc.gnu.org/bugzilla/attachment.cgi?id=13591action=view)
new patch
This new patch changes one little thing dealing with how with indirect
references are handled in
--- Comment #5 from pinskia at gcc dot gnu dot org 2007-05-20 23:03 ---
(From update of attachment 13591)
I take this back, this can cause an ICE with some code (which seems like I
should add a testcase).
--
pinskia at gcc dot gnu dot org changed:
What|Removed
--- Comment #6 from pinskia at gcc dot gnu dot org 2007-05-20 23:06 ---
(From update of attachment 13587)
This patch is safer and allows us to add more to build_classic_dist_vector_1.
Like when the two access functions are constants, just take the difference.
(which might be needed
--- Comment #1 from pinskia at gcc dot gnu dot org 2007-05-20 05:34 ---
A quick note we can determine the depency between p-a[i] and p-a[i+1].
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=31995