* [Bug middle-end/56098] conditional write through volatile pointer produces unintended read
2013-01-24 18:22 [Bug c/56098] New: conditional write through volatile pointer produces unintended read werner at almesberger dot net
@ 2013-01-25 8:43 ` mikpe at it dot uu.se
2013-01-25 14:43 ` [Bug middle-end/56098] [4.6/4.7/4.8 Regression] " rguenth at gcc dot gnu.org
` (9 subsequent siblings)
10 siblings, 0 replies; 12+ messages in thread
From: mikpe at it dot uu.se @ 2013-01-25 8:43 UTC (permalink / raw)
To: gcc-bugs
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=56098
Mikael Pettersson <mikpe at it dot uu.se> changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |mikpe at it dot uu.se
--- Comment #1 from Mikael Pettersson <mikpe at it dot uu.se> 2013-01-25 08:42:56 UTC ---
gcc 3.3.6 to 4.2.4 generate:
problem:
.LFB2:
movq ptr(%rip), %rax
testl %edi, %edi
movl $1, (%rax)
je .L4
movl $2, (%rax)
.L4:
rep ; ret
which looks Ok to me. From 4.3.6 up to 4.7.2 we get the broken code Werner
showed. 3.2.3 generates different broken code:
problem:
.LFB1:
xorl %eax, %eax
movq ptr(%rip), %rdx
testl %edi, %edi
setne %al
movl $1, (%rdx)
incl %eax
movl %eax, (%rdx)
ret
^ permalink raw reply [flat|nested] 12+ messages in thread
* [Bug middle-end/56098] [4.6/4.7/4.8 Regression] conditional write through volatile pointer produces unintended read
2013-01-24 18:22 [Bug c/56098] New: conditional write through volatile pointer produces unintended read werner at almesberger dot net
2013-01-25 8:43 ` [Bug middle-end/56098] " mikpe at it dot uu.se
@ 2013-01-25 14:43 ` rguenth at gcc dot gnu.org
2013-01-25 15:35 ` jakub at gcc dot gnu.org
` (8 subsequent siblings)
10 siblings, 0 replies; 12+ messages in thread
From: rguenth at gcc dot gnu.org @ 2013-01-25 14:43 UTC (permalink / raw)
To: gcc-bugs
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=56098
Richard Biener <rguenth at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|UNCONFIRMED |NEW
Last reconfirmed| |2013-01-25
CC| |matz at gcc dot gnu.org
Target Milestone|--- |4.6.4
Summary|conditional write through |[4.6/4.7/4.8 Regression]
|volatile pointer produces |conditional write through
|unintended read |volatile pointer produces
| |unintended read
Ever Confirmed|0 |1
--- Comment #2 from Richard Biener <rguenth at gcc dot gnu.org> 2013-01-25 14:42:43 UTC ---
This is cselim at work, preparing the way for if-conversion of the store ...
It shouldn't do this for volatiles of course.
Workaround: -fno-tree-cselim
^ permalink raw reply [flat|nested] 12+ messages in thread
* [Bug middle-end/56098] [4.6/4.7/4.8 Regression] conditional write through volatile pointer produces unintended read
2013-01-24 18:22 [Bug c/56098] New: conditional write through volatile pointer produces unintended read werner at almesberger dot net
2013-01-25 8:43 ` [Bug middle-end/56098] " mikpe at it dot uu.se
2013-01-25 14:43 ` [Bug middle-end/56098] [4.6/4.7/4.8 Regression] " rguenth at gcc dot gnu.org
@ 2013-01-25 15:35 ` jakub at gcc dot gnu.org
2013-01-25 20:04 ` jakub at gcc dot gnu.org
` (7 subsequent siblings)
10 siblings, 0 replies; 12+ messages in thread
From: jakub at gcc dot gnu.org @ 2013-01-25 15:35 UTC (permalink / raw)
To: gcc-bugs
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=56098
Jakub Jelinek <jakub at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|NEW |ASSIGNED
AssignedTo|unassigned at gcc dot |jakub at gcc dot gnu.org
|gnu.org |
--- Comment #3 from Jakub Jelinek <jakub at gcc dot gnu.org> 2013-01-25 15:34:58 UTC ---
Created attachment 29275
--> http://gcc.gnu.org/bugzilla/attachment.cgi?id=29275
gcc48-pr56098.patch
Untested fix.
^ permalink raw reply [flat|nested] 12+ messages in thread
* [Bug middle-end/56098] [4.6/4.7/4.8 Regression] conditional write through volatile pointer produces unintended read
2013-01-24 18:22 [Bug c/56098] New: conditional write through volatile pointer produces unintended read werner at almesberger dot net
` (2 preceding siblings ...)
2013-01-25 15:35 ` jakub at gcc dot gnu.org
@ 2013-01-25 20:04 ` jakub at gcc dot gnu.org
2013-01-25 20:06 ` [Bug middle-end/56098] [4.6/4.7 " jakub at gcc dot gnu.org
` (6 subsequent siblings)
10 siblings, 0 replies; 12+ messages in thread
From: jakub at gcc dot gnu.org @ 2013-01-25 20:04 UTC (permalink / raw)
To: gcc-bugs
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=56098
--- Comment #4 from Jakub Jelinek <jakub at gcc dot gnu.org> 2013-01-25 20:04:01 UTC ---
Author: jakub
Date: Fri Jan 25 20:03:54 2013
New Revision: 195475
URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=195475
Log:
PR tree-optimization/56098
* tree-ssa-phiopt.c (nt_init_block): Don't call add_or_mark_expr
for stmts with volatile ops.
(cond_store_replacement): Don't optimize if assign has volatile ops.
(cond_if_else_store_replacement_1): Don't optimize if either
then_assign or else_assign have volatile ops.
(hoist_adjacent_loads): Don't optimize if either def1 or def2 have
volatile ops.
* gcc.dg/pr56098-1.c: New test.
* gcc.dg/pr56098-2.c: New test.
Added:
trunk/gcc/testsuite/gcc.dg/pr56098-1.c
trunk/gcc/testsuite/gcc.dg/pr56098-2.c
Modified:
trunk/gcc/ChangeLog
trunk/gcc/testsuite/ChangeLog
trunk/gcc/tree-ssa-phiopt.c
^ permalink raw reply [flat|nested] 12+ messages in thread
* [Bug middle-end/56098] [4.6/4.7 Regression] conditional write through volatile pointer produces unintended read
2013-01-24 18:22 [Bug c/56098] New: conditional write through volatile pointer produces unintended read werner at almesberger dot net
` (3 preceding siblings ...)
2013-01-25 20:04 ` jakub at gcc dot gnu.org
@ 2013-01-25 20:06 ` jakub at gcc dot gnu.org
2013-01-25 22:46 ` werner at almesberger dot net
` (5 subsequent siblings)
10 siblings, 0 replies; 12+ messages in thread
From: jakub at gcc dot gnu.org @ 2013-01-25 20:06 UTC (permalink / raw)
To: gcc-bugs
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=56098
Jakub Jelinek <jakub at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Summary|[4.6/4.7/4.8 Regression] |[4.6/4.7 Regression]
|conditional write through |conditional write through
|volatile pointer produces |volatile pointer produces
|unintended read |unintended read
--- Comment #5 from Jakub Jelinek <jakub at gcc dot gnu.org> 2013-01-25 20:05:48 UTC ---
Fixed on the trunk so far.
^ permalink raw reply [flat|nested] 12+ messages in thread
* [Bug middle-end/56098] [4.6/4.7 Regression] conditional write through volatile pointer produces unintended read
2013-01-24 18:22 [Bug c/56098] New: conditional write through volatile pointer produces unintended read werner at almesberger dot net
` (4 preceding siblings ...)
2013-01-25 20:06 ` [Bug middle-end/56098] [4.6/4.7 " jakub at gcc dot gnu.org
@ 2013-01-25 22:46 ` werner at almesberger dot net
2013-01-26 2:10 ` werner at almesberger dot net
` (4 subsequent siblings)
10 siblings, 0 replies; 12+ messages in thread
From: werner at almesberger dot net @ 2013-01-25 22:46 UTC (permalink / raw)
To: gcc-bugs
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=56098
--- Comment #6 from werner at almesberger dot net 2013-01-25 22:46:17 UTC ---
Thanks for the analysis and the fixes ! I'll try them soonish.
Regarding work-arounds, the ones I mentioned for my original code snippet
(i.e., -O1 or -fno-strict-aliasing) aren't sufficient in the following more
general case:
volatile void *p;
#define P (*(volatile int *) (p+8))
void foo(int x)
{
P = 1;
if (x)
P = 2;
}
This is for gcc (Ubuntu/Linaro 4.7.2-2ubuntu1) 4.7.2 on x86-64. The offset (8)
may be architecture-specific. For values < 8, correct code is generated with
-O1 (but not -O2 or higher).
The good news is that -fno-tree-cselim does indeed avoid the bad read in all
cases I've tried, with any optimization level.
^ permalink raw reply [flat|nested] 12+ messages in thread
* [Bug middle-end/56098] [4.6/4.7 Regression] conditional write through volatile pointer produces unintended read
2013-01-24 18:22 [Bug c/56098] New: conditional write through volatile pointer produces unintended read werner at almesberger dot net
` (5 preceding siblings ...)
2013-01-25 22:46 ` werner at almesberger dot net
@ 2013-01-26 2:10 ` werner at almesberger dot net
2013-02-01 14:17 ` jakub at gcc dot gnu.org
` (3 subsequent siblings)
10 siblings, 0 replies; 12+ messages in thread
From: werner at almesberger dot net @ 2013-01-26 2:10 UTC (permalink / raw)
To: gcc-bugs
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=56098
--- Comment #7 from werner at almesberger dot net 2013-01-26 02:09:33 UTC ---
I'm happy to confirm that trunk (for x86-64) now passes all variations of this
theme I tried with flying colors. Thanks again !
^ permalink raw reply [flat|nested] 12+ messages in thread
* [Bug middle-end/56098] [4.6/4.7 Regression] conditional write through volatile pointer produces unintended read
2013-01-24 18:22 [Bug c/56098] New: conditional write through volatile pointer produces unintended read werner at almesberger dot net
` (6 preceding siblings ...)
2013-01-26 2:10 ` werner at almesberger dot net
@ 2013-02-01 14:17 ` jakub at gcc dot gnu.org
2013-02-01 14:38 ` [Bug middle-end/56098] [4.6 " jakub at gcc dot gnu.org
` (2 subsequent siblings)
10 siblings, 0 replies; 12+ messages in thread
From: jakub at gcc dot gnu.org @ 2013-02-01 14:17 UTC (permalink / raw)
To: gcc-bugs
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=56098
--- Comment #8 from Jakub Jelinek <jakub at gcc dot gnu.org> 2013-02-01 14:16:28 UTC ---
Author: jakub
Date: Fri Feb 1 14:16:20 2013
New Revision: 195663
URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=195663
Log:
Backported from mainline
2013-01-25 Jakub Jelinek <jakub@redhat.com>
PR tree-optimization/56098
* tree-ssa-phiopt.c (nt_init_block): Don't call add_or_mark_expr
for stmts with volatile ops.
(cond_store_replacement): Don't optimize if assign has volatile ops.
(cond_if_else_store_replacement_1): Don't optimize if either
then_assign or else_assign have volatile ops.
* gcc.dg/pr56098-1.c: New test.
Added:
branches/gcc-4_7-branch/gcc/testsuite/gcc.dg/pr56098-1.c
Modified:
branches/gcc-4_7-branch/gcc/ChangeLog
branches/gcc-4_7-branch/gcc/testsuite/ChangeLog
branches/gcc-4_7-branch/gcc/tree-ssa-phiopt.c
^ permalink raw reply [flat|nested] 12+ messages in thread
* [Bug middle-end/56098] [4.6 Regression] conditional write through volatile pointer produces unintended read
2013-01-24 18:22 [Bug c/56098] New: conditional write through volatile pointer produces unintended read werner at almesberger dot net
` (7 preceding siblings ...)
2013-02-01 14:17 ` jakub at gcc dot gnu.org
@ 2013-02-01 14:38 ` jakub at gcc dot gnu.org
2013-02-27 19:47 ` pinskia at gcc dot gnu.org
2013-04-03 18:20 ` jakub at gcc dot gnu.org
10 siblings, 0 replies; 12+ messages in thread
From: jakub at gcc dot gnu.org @ 2013-02-01 14:38 UTC (permalink / raw)
To: gcc-bugs
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=56098
Jakub Jelinek <jakub at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Summary|[4.6/4.7 Regression] |[4.6 Regression]
|conditional write through |conditional write through
|volatile pointer produces |volatile pointer produces
|unintended read |unintended read
--- Comment #9 from Jakub Jelinek <jakub at gcc dot gnu.org> 2013-02-01 14:38:06 UTC ---
Fixed for 4.7.3+ too.
^ permalink raw reply [flat|nested] 12+ messages in thread
* [Bug middle-end/56098] [4.6 Regression] conditional write through volatile pointer produces unintended read
2013-01-24 18:22 [Bug c/56098] New: conditional write through volatile pointer produces unintended read werner at almesberger dot net
` (8 preceding siblings ...)
2013-02-01 14:38 ` [Bug middle-end/56098] [4.6 " jakub at gcc dot gnu.org
@ 2013-02-27 19:47 ` pinskia at gcc dot gnu.org
2013-04-03 18:20 ` jakub at gcc dot gnu.org
10 siblings, 0 replies; 12+ messages in thread
From: pinskia at gcc dot gnu.org @ 2013-02-27 19:47 UTC (permalink / raw)
To: gcc-bugs
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=56098
Andrew Pinski <pinskia at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |joff at embeddedarm dot com
--- Comment #10 from Andrew Pinski <pinskia at gcc dot gnu.org> 2013-02-27 19:46:16 UTC ---
*** Bug 56476 has been marked as a duplicate of this bug. ***
^ permalink raw reply [flat|nested] 12+ messages in thread
* [Bug middle-end/56098] [4.6 Regression] conditional write through volatile pointer produces unintended read
2013-01-24 18:22 [Bug c/56098] New: conditional write through volatile pointer produces unintended read werner at almesberger dot net
` (9 preceding siblings ...)
2013-02-27 19:47 ` pinskia at gcc dot gnu.org
@ 2013-04-03 18:20 ` jakub at gcc dot gnu.org
10 siblings, 0 replies; 12+ messages in thread
From: jakub at gcc dot gnu.org @ 2013-04-03 18:20 UTC (permalink / raw)
To: gcc-bugs
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=56098
Jakub Jelinek <jakub at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|ASSIGNED |RESOLVED
Resolution| |FIXED
--- Comment #11 from Jakub Jelinek <jakub at gcc dot gnu.org> 2013-04-03 18:20:54 UTC ---
Author: jakub
Date: Wed Apr 3 18:02:57 2013
New Revision: 197449
URL: http://gcc.gnu.org/viewcvs?rev=197449&root=gcc&view=rev
Log:
Backported from mainline
2013-01-25 Jakub Jelinek <jakub@redhat.com>
PR tree-optimization/56098
* tree-ssa-phiopt.c (nt_init_block): Don't call add_or_mark_expr
for stmts with volatile ops.
(cond_store_replacement): Don't optimize if assign has volatile ops.
(cond_if_else_store_replacement_1): Don't optimize if either
then_assign or else_assign have volatile ops.
* gcc.dg/pr56098-1.c: New test.
Added:
branches/gcc-4_6-branch/gcc/testsuite/gcc.dg/pr56098-1.c
Modified:
branches/gcc-4_6-branch/gcc/ChangeLog
branches/gcc-4_6-branch/gcc/testsuite/ChangeLog
branches/gcc-4_6-branch/gcc/tree-ssa-phiopt.c
^ permalink raw reply [flat|nested] 12+ messages in thread