[Bug rtl-optimization/92591] ICE in optimize_sc, at modulo-sched.c:1063
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92591 Roman Zhuykov changed: What|Removed |Added Status|ASSIGNED|RESOLVED Resolution|--- |FIXED --- Comment #12 from Roman Zhuykov --- Sorry for mistake in test names (was 92951 instead of 92591). Obvious rename patch committed as r279645, r279646, r279647 into 10/9/8 branches. PR is now fixed in all branches.
[Bug rtl-optimization/92591] ICE in optimize_sc, at modulo-sched.c:1063
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92591 --- Comment #11 from Roman Zhuykov --- Author: zhroma Date: Fri Dec 20 15:07:58 2019 New Revision: 279643 URL: https://gcc.gnu.org/viewcvs?rev=279643&root=gcc&view=rev Log: modulo-sched: backport PR92591 fix from mainline 2019-12-20 Roman Zhuykov modulo-sched: fix branch rescheduling issue (PR92591) PR rtl-optimization/92591 * modulo-sched.c (ps_add_node_check_conflicts): Improve checking for history > 0 case. * params.def (sms-dfa-history): Limit to 16. gcc/testsuite: PR rtl-optimization/92591 * gcc.dg/pr92951-1.c: New test. * gcc.dg/pr92951-2.c: New test. Added: branches/gcc-8-branch/gcc/testsuite/gcc.dg/pr92951-1.c branches/gcc-8-branch/gcc/testsuite/gcc.dg/pr92951-2.c Modified: branches/gcc-8-branch/gcc/ChangeLog branches/gcc-8-branch/gcc/modulo-sched.c branches/gcc-8-branch/gcc/params.def branches/gcc-8-branch/gcc/testsuite/ChangeLog
[Bug rtl-optimization/92591] ICE in optimize_sc, at modulo-sched.c:1063
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92591 --- Comment #10 from Roman Zhuykov --- Author: zhroma Date: Fri Dec 20 15:05:12 2019 New Revision: 279642 URL: https://gcc.gnu.org/viewcvs?rev=279642&root=gcc&view=rev Log: modulo-sched: backport PR92591 fix from mainline 2019-12-20 Roman Zhuykov modulo-sched: fix branch rescheduling issue (PR92591) PR rtl-optimization/92591 * modulo-sched.c (ps_add_node_check_conflicts): Improve checking for history > 0 case. * params.def (sms-dfa-history): Limit to 16. gcc/testsuite: PR rtl-optimization/92591 * gcc.dg/pr92951-1.c: New test. * gcc.dg/pr92951-2.c: New test. Added: branches/gcc-9-branch/gcc/testsuite/gcc.dg/pr92951-1.c branches/gcc-9-branch/gcc/testsuite/gcc.dg/pr92951-2.c Modified: branches/gcc-9-branch/gcc/ChangeLog branches/gcc-9-branch/gcc/modulo-sched.c branches/gcc-9-branch/gcc/params.def branches/gcc-9-branch/gcc/testsuite/ChangeLog
[Bug rtl-optimization/92591] ICE in optimize_sc, at modulo-sched.c:1063
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92591 --- Comment #9 from Roman Zhuykov --- Started discussion in mailing list: https://gcc.gnu.org/ml/gcc-patches/2019-12/msg01223.html
[Bug rtl-optimization/92591] ICE in optimize_sc, at modulo-sched.c:1063
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92591 --- Comment #8 from Arseny Solokha --- Is there a backport pending, or can this PR be closed?
[Bug rtl-optimization/92591] ICE in optimize_sc, at modulo-sched.c:1063
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92591 --- Comment #7 from Roman Zhuykov --- Author: zhroma Date: Fri Dec 13 17:33:38 2019 New Revision: 279377 URL: https://gcc.gnu.org/viewcvs?rev=279377&root=gcc&view=rev Log: modulo-sched: fix branch rescheduling issue (PR92591) PR rtl-optimization/92591 * modulo-sched.c (ps_add_node_check_conflicts): Improve checking for history > 0 case. testsuite: PR rtl-optimization/92591 * gcc.dg/pr92951-1.c: New test. * gcc.dg/pr92951-2.c: New test. Added: trunk/gcc/testsuite/gcc.dg/pr92951-1.c trunk/gcc/testsuite/gcc.dg/pr92951-2.c Modified: trunk/gcc/ChangeLog trunk/gcc/modulo-sched.c trunk/gcc/testsuite/ChangeLog
[Bug rtl-optimization/92591] ICE in optimize_sc, at modulo-sched.c:1063
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92591 --- Comment #6 from Roman Zhuykov --- Patches are still testing. Arm (both 32 and 64 bit) cross compilers are OK, and I want to test some additional scenarios for powerpc. Probably I'll post patches next week.
[Bug rtl-optimization/92591] ICE in optimize_sc, at modulo-sched.c:1063
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92591 --- Comment #5 from Arseny Solokha --- Roman, are you going to submit these patches to the mailing list?
[Bug rtl-optimization/92591] ICE in optimize_sc, at modulo-sched.c:1063
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92591 Roman Zhuykov changed: What|Removed |Added Attachment #47379|0 |1 is obsolete|| --- Comment #4 from Roman Zhuykov --- Created attachment 47386 --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=47386&action=edit Proposed patch v2 Found a mistake in previous patch
[Bug rtl-optimization/92591] ICE in optimize_sc, at modulo-sched.c:1063
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92591 --- Comment #3 from Roman Zhuykov --- Created attachment 47380 --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=47380&action=edit Parameters patch
[Bug rtl-optimization/92591] ICE in optimize_sc, at modulo-sched.c:1063
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92591 --- Comment #2 from Roman Zhuykov --- Created attachment 47379 --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=47379&action=edit Proposed patch
[Bug rtl-optimization/92591] ICE in optimize_sc, at modulo-sched.c:1063
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92591 Roman Zhuykov changed: What|Removed |Added Status|UNCONFIRMED |ASSIGNED Last reconfirmed||2019-11-27 Assignee|unassigned at gcc dot gnu.org |zhroma at gcc dot gnu.org Ever confirmed|0 |1 --- Comment #1 from Roman Zhuykov --- Oops, it seems that non-zero sms-dfa-history is untested since modulo-sched was added in 2004. This bugs is latent since r177235 in 2011, and logic is similar to bug 84032 comment #4. When we schedule some insn into row R we check that DFA can't see any conflicts in rows R-H .. R+H (modulo II), where H=sms-dfa-history. So, when we schedule branch into row R some of its neighbours were not yet scheduled and DFA tells OK. While scheduling neighbours DFA may check some nearby rows, for example, rows [R+1-H .. R+1+H] or something like that, but not exactly [R-H .. R+H]. Later we decide to temporarily remove branch insn from schedule and are pretty sure we could put it back (if rescheduling to another row fails). And this time DFA shows a conflict, because now all branch neighbours are scheduled already. Attached untested patch fixes this by running extra DFA checks for all rows [x-H..x+H] where X is any value in range [R-H,R+H], thus eliminating the issue. Certainly, these checks are expensive enough, so I also prepated a patch about modulo-sched params, which will set max sms-dfa-history value to 16. While at it, the patch also fixes other issues about sms parameters.
