https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65456
Bill Schmidt wschmidt at gcc dot gnu.org changed:
What|Removed |Added
Status|REOPENED|RESOLVED
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65456
--- Comment #26 from ro at CeBiTec dot Uni-Bielefeld.DE ro at CeBiTec dot
Uni-Bielefeld.DE ---
--- Comment #25 from ro at CeBiTec dot Uni-Bielefeld.DE ro at CeBiTec dot
Uni-Bielefeld.DE ---
[...]
Not yet: those sparc boxes are slow, and it
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65456
--- Comment #25 from ro at CeBiTec dot Uni-Bielefeld.DE ro at CeBiTec dot
Uni-Bielefeld.DE ---
--- Comment #24 from Bill Schmidt wschmidt at gcc dot gnu.org ---
No, I don't think so. The same change was made in GCC 4.9, and it didn't
cause
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65456
Rainer Orth ro at gcc dot gnu.org changed:
What|Removed |Added
Status|RESOLVED|REOPENED
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65456
--- Comment #23 from Rainer Orth ro at gcc dot gnu.org ---
Created attachment 35456
-- https://gcc.gnu.org/bugzilla/attachment.cgi?id=35456action=edit
bb-slp-32.c.141t.slp2 dump
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65456
--- Comment #24 from Bill Schmidt wschmidt at gcc dot gnu.org ---
No, I don't think so. The same change was made in GCC 4.9, and it didn't cause
it to XPASS there (looking at gcc-testresults). Also, my change restricted the
number of cases for
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65456
--- Comment #19 from Bill Schmidt wschmidt at gcc dot gnu.org ---
Author: wschmidt
Date: Fri Apr 24 13:45:08 2015
New Revision: 222412
URL: https://gcc.gnu.org/viewcvs?rev=222412root=gccview=rev
Log:
[gcc]
2015-04-24 Bill Schmidt
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65456
Bill Schmidt wschmidt at gcc dot gnu.org changed:
What|Removed |Added
Status|ASSIGNED|RESOLVED
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65456
--- Comment #20 from Bill Schmidt wschmidt at gcc dot gnu.org ---
Author: wschmidt
Date: Fri Apr 24 20:17:10 2015
New Revision: 222423
URL: https://gcc.gnu.org/viewcvs?rev=222423root=gccview=rev
Log:
[gcc]
2015-04-24 Bill Schmidt
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65456
--- Comment #18 from Bill Schmidt wschmidt at gcc dot gnu.org ---
Author: wschmidt
Date: Thu Apr 23 21:03:40 2015
New Revision: 222386
URL: https://gcc.gnu.org/viewcvs?rev=222386root=gccview=rev
Log:
[gcc]
2015-04-23 Bill Schmidt
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65456
--- Comment #17 from Bill Schmidt wschmidt at gcc dot gnu.org ---
Author: wschmidt
Date: Thu Apr 23 00:21:39 2015
New Revision: 222349
URL: https://gcc.gnu.org/viewcvs?rev=222349root=gccview=rev
Log:
[gcc]
2015-04-22 Bill Schmidt
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65456
--- Comment #15 from Bill Schmidt wschmidt at gcc dot gnu.org ---
That last bit also needs an update to the table in ISA 2.07 II.2, p. 754.
I'm planning to include the change for the vector alignment piece in my P8
unaligned vectors patch, since
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65456
--- Comment #16 from Bill Schmidt wschmidt at gcc dot gnu.org ---
Proposed patch added to the general P8 unaligned vector patch:
https://gcc.gnu.org/ml/gcc-patches/2015-03/msg01502.html
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65456
--- Comment #12 from Bill Schmidt wschmidt at gcc dot gnu.org ---
The problem is this declaration in rs6000.h, which forces unaligned vector
stores to be scalarized during expand:
/* Define this macro to be the value 1 if unaligned accesses have
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65456
--- Comment #13 from Bill Schmidt wschmidt at gcc dot gnu.org ---
Changing the condition as follows produces a nice tight lxvd2x/stxvd2x loop in
all three places:
#define SLOW_UNALIGNED_ACCESS(MODE, ALIGN) \
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65456
Alan Modra amodra at gmail dot com changed:
What|Removed |Added
CC||amodra at gmail dot
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65456
--- Comment #11 from Bill Schmidt wschmidt at gcc dot gnu.org ---
(In reply to David Edelsohn from comment #10)
I believe that the choice to scalarize is based on the vector cost model.
Hm, that would be interesting. The applied patch changes
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65456
Bill Schmidt wschmidt at gcc dot gnu.org changed:
What|Removed |Added
Status|NEW |ASSIGNED
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65456
--- Comment #10 from David Edelsohn dje at gcc dot gnu.org ---
I believe that the choice to scalarize is based on the vector cost model.
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65456
--- Comment #8 from Bill Schmidt wschmidt at gcc dot gnu.org ---
Pish, those adds shouldn't be the least bit relevant. I'll have a look.
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65456
--- Comment #7 from Anton Blanchard anton at samba dot org ---
Thanks Martin.
Bill: the swaps pass isn't catching our vectorised copy, I guess because of the
adds in the loop:
lxvd2x 0,9,4
addi 28,1,-48
add 6,9,10
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65456
--- Comment #6 from Martin Sebor msebor at gcc dot gnu.org ---
Created attachment 35066
-- https://gcc.gnu.org/bugzilla/attachment.cgi?id=35066action=edit
Assembly emitted by gcc 5.0.0 20150319 after aplying the patch referenced in
comment #5.
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65456
--- Comment #4 from Richard Biener rguenth at gcc dot gnu.org ---
Note the vectorizer has a slight preference to align stores.
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65456
Markus Trippelsdorf trippels at gcc dot gnu.org changed:
What|Removed |Added
CC|
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65456
Markus Trippelsdorf trippels at gcc dot gnu.org changed:
What|Removed |Added
Target Milestone|5.0 |---
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65456
David Edelsohn dje at gcc dot gnu.org changed:
What|Removed |Added
CC||amodra at gcc dot
26 matches
Mail list logo
