[Bug middle-end/89008] [7/8 Regression] O2 and O1 results differ for simple test
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=89008 Bill Schmidt changed: What|Removed |Added Status|NEW |RESOLVED Resolution|--- |FIXED --- Comment #16 from Bill Schmidt --- Latent SLSR bug is now fixed everywhere also.
[Bug middle-end/89008] [7/8 Regression] O2 and O1 results differ for simple test
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=89008 --- Comment #15 from Bill Schmidt --- Author: wschmidt Date: Thu Jan 31 21:55:45 2019 New Revision: 268431 URL: https://gcc.gnu.org/viewcvs?rev=268431&root=gcc&view=rev Log: 2018-01-31 Bill Schmidt Backport from mainline 2018-01-31 Bill Schmidt PR tree-optimization/89008 * gimple-ssa-strength-reduction.c (slsr_process_mul): Don't process anything of the form X * 0. Modified: branches/gcc-7-branch/gcc/ChangeLog branches/gcc-7-branch/gcc/gimple-ssa-strength-reduction.c
[Bug middle-end/89008] [7/8 Regression] O2 and O1 results differ for simple test
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=89008 --- Comment #14 from Bill Schmidt --- Author: wschmidt Date: Thu Jan 31 17:14:36 2019 New Revision: 268425 URL: https://gcc.gnu.org/viewcvs?rev=268425&root=gcc&view=rev Log: 2018-01-31 Bill Schmidt Backport from mainline 2018-01-31 Bill Schmidt PR tree-optimization/89008 * gimple-ssa-strength-reduction.c (slsr_process_mul): Don't process anything of the form X * 0. Modified: branches/gcc-8-branch/gcc/ChangeLog branches/gcc-8-branch/gcc/gimple-ssa-strength-reduction.c
[Bug middle-end/89008] [7/8 Regression] O2 and O1 results differ for simple test
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=89008 --- Comment #13 from Bill Schmidt --- Author: wschmidt Date: Thu Jan 31 13:53:06 2019 New Revision: 268422 URL: https://gcc.gnu.org/viewcvs?rev=268422&root=gcc&view=rev Log: 2018-01-31 Bill Schmidt PR tree-optimization/89008 * gimple-ssa-strength-reduction.c (slsr_process_mul): Don't process anything of the form X * 0. Modified: trunk/gcc/ChangeLog trunk/gcc/gimple-ssa-strength-reduction.c
[Bug middle-end/89008] [7/8 Regression] O2 and O1 results differ for simple test
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=89008 --- Comment #12 from Bill Schmidt --- Yes, fully agree -- I'll take care of that probably tomorrow. Too many meetings this week...
[Bug middle-end/89008] [7/8 Regression] O2 and O1 results differ for simple test
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=89008 --- Comment #10 from Richard Biener --- Author: rguenth Date: Thu Jan 24 11:07:19 2019 New Revision: 268235 URL: https://gcc.gnu.org/viewcvs?rev=268235&root=gcc&view=rev Log: 2019-01-24 Richard Biener Backport from mainline 2019-01-23 Richard Biener PR tree-optimization/89008 * tree-ssa-reassoc.c (eliminate_using_constants): For * 0 do not leave another stray operand. * gcc.dg/torture/pr89008.c: New testcase. Added: branches/gcc-7-branch/gcc/testsuite/gcc.dg/torture/pr89008.c Modified: branches/gcc-7-branch/gcc/ChangeLog branches/gcc-7-branch/gcc/testsuite/ChangeLog branches/gcc-7-branch/gcc/tree-ssa-reassoc.c
[Bug middle-end/89008] [7/8 Regression] O2 and O1 results differ for simple test
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=89008 Richard Biener changed: What|Removed |Added Known to work||7.4.1, 8.2.1 --- Comment #11 from Richard Biener --- "fixed" everywhere, I'm out of the way now, SLSR is still latently buggy (I suggest to simply "ignore" * 0 stmts)
[Bug middle-end/89008] [7/8 Regression] O2 and O1 results differ for simple test
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=89008 --- Comment #9 from Richard Biener --- Bill, you now need r268233 or earlier on the GCC 8 branch to reproduce the latent issue.
[Bug middle-end/89008] [7/8 Regression] O2 and O1 results differ for simple test
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=89008 --- Comment #8 from Richard Biener --- Author: rguenth Date: Thu Jan 24 09:31:51 2019 New Revision: 268234 URL: https://gcc.gnu.org/viewcvs?rev=268234&root=gcc&view=rev Log: 2019-01-24 Richard Biener Backport from mainline 2019-01-23 Richard Biener PR tree-optimization/89008 * tree-ssa-reassoc.c (eliminate_using_constants): For * 0 do not leave another stray operand. * gcc.dg/torture/pr89008.c: New testcase. Added: branches/gcc-8-branch/gcc/testsuite/gcc.dg/torture/pr89008.c Modified: branches/gcc-8-branch/gcc/ChangeLog branches/gcc-8-branch/gcc/testsuite/ChangeLog branches/gcc-8-branch/gcc/tree-ssa-reassoc.c
[Bug middle-end/89008] [7/8 Regression] O2 and O1 results differ for simple test
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=89008 --- Comment #7 from Bill Schmidt --- (In reply to Richard Biener from comment #5) > Created attachment 45506 [details] > patch for the * 0 issue in reassoc > > I am testing a mitigation (and missed optimization fix) in reassoc. Bill, > can you look into the SLSR issue which is still latent? Yes, I'll look into it. * 0, eh? Yeah, I probably didn't think about that. :-) Probably will be a couple of days before I get to it.
[Bug middle-end/89008] [7/8 Regression] O2 and O1 results differ for simple test
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=89008 --- Comment #6 from Richard Biener --- Author: rguenth Date: Wed Jan 23 14:02:13 2019 New Revision: 268186 URL: https://gcc.gnu.org/viewcvs?rev=268186&root=gcc&view=rev Log: 2019-01-23 Richard Biener PR tree-optimization/89008 * tree-ssa-reassoc.c (eliminate_using_constants): For * 0 do not leave another stray operand. * gcc.dg/torture/pr89008.c: New testcase. Added: trunk/gcc/testsuite/gcc.dg/torture/pr89008.c Modified: trunk/gcc/ChangeLog trunk/gcc/testsuite/ChangeLog trunk/gcc/tree-ssa-reassoc.c
[Bug middle-end/89008] [7/8 Regression] O2 and O1 results differ for simple test
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=89008 --- Comment #5 from Richard Biener --- Created attachment 45506 --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=45506&action=edit patch for the * 0 issue in reassoc I am testing a mitigation (and missed optimization fix) in reassoc. Bill, can you look into the SLSR issue which is still latent?
[Bug middle-end/89008] [7/8 Regression] O2 and O1 results differ for simple test
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=89008
--- Comment #4 from Richard Biener ---
(In reply to Martin Liška from comment #3)
> Started with r236440 and was fixed with r263875.
Thanks Martin - this means both the * 0 added by reassoc and the SLSR issue
are latent.
Found * 0, removing all other ops
Width = 1 was chosen for reassociation
Transforming _12 = _10 * 83886080;
into _12 = b_lsm.9_34 * 0;
GIMPLE testcase that shows the above on trunk:
unsigned long a;
unsigned b;
long int __GIMPLE (startwith("reassoc"))
f ()
{
unsigned int _4;
unsigned int _6;
unsigned int _10;
unsigned int _12;
unsigned int _19;
unsigned int _33;
unsigned int _36;
unsigned long g_21;
unsigned int b_lsm9_34;
bb_2:
b_lsm9_34 = b;
_36 = b_lsm9_34 * 83886080u;
_33 = _36 * 83886080u;
_4 = _33 * 83886080u;
_10 = _4 * 83886080u;
_12 = _10 * 83886080u;
_19 = _12 * 2u;
_6 = -_19;
g_21 = (long unsigned int) _6;
a = g_21;
return 0l;
}
[Bug middle-end/89008] [7/8 Regression] O2 and O1 results differ for simple test
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=89008 --- Comment #3 from Martin Liška --- Started with r236440 and was fixed with r263875.
[Bug middle-end/89008] [7/8 Regression] O2 and O1 results differ for simple test
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=89008 Richard Biener changed: What|Removed |Added CC||wschmidt at gcc dot gnu.org --- Comment #2 from Richard Biener --- Eventually the issue also just became latent. SLSR does Replacing: _7 = _12 * 83886080; With: _7 = _12; Replacing: _19 = _12 * 2; With: _19 = _2; where the latter is bogus replacing 0 with 4. Eventually SLSR is confused by _12 = b_lsm.9_34 * 0; being present in the IL, not sure. IL before SLSR: _12 = b_lsm.9_34 * 0; ... _2 = _12 + 4; ... _19 = _12 * 2; _6 = -_19; g_21 = (long unsigned int) _6; a = g_21; possibly checks some multiple-of which breaks down with zero. On trunk the * 0 is gone (possibly the "fix").
[Bug middle-end/89008] [7/8 Regression] O2 and O1 results differ for simple test
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=89008
Richard Biener changed:
What|Removed |Added
Priority|P3 |P2
Status|UNCONFIRMED |NEW
Known to work||6.4.0, 9.0
Keywords||needs-bisection, wrong-code
Last reconfirmed||2019-01-23
CC||marxin at gcc dot gnu.org
Ever confirmed|0 |1
Summary|O2 and O1 results differ|[7/8 Regression] O2 and O1
|for simple test |results differ for simple
||test
Target Milestone|--- |7.5
Known to fail||7.4.0, 8.2.0
--- Comment #1 from Richard Biener ---
Confirmed. Seems to work on trunk, disabling SLSR fixes it. We end up with
f ()
{
[local count: 14598063]:
b = 0;
c = 4;
e = 6;
d = 6;
a = 4294967292;
return 0;
so I guess a fix needs backporting. Martin, can you bisect the fix?
