gfortran-9.0.0-alpha20190127 snapshot (r268327) ICEs when compiling the following snippet reduced from gcc/testsuite/gfortran.dg/mvbits_7.f90 w/ -O1 -floop-parallelize-all -fmodulo-sched -fmodulo-sched-allow-regmoves -freorder-blocks-and-partition -ftree-parallelize-loops=2 -fno-inline -fno-move-loop-invariants -fno-sched-pressure: type t integer :: I character(9) :: chr end type type(t) :: x(4,3) call foo (x) contains SUBROUTINE foo (x) TYPE(t) x(4, 3) x%i = x%i * 2 END SUBROUTINE end % powerpc-e300c3-linux-gnu-gfortran-9.0.0-alpha20190127 -O1 -floop-parallelize-all -fmodulo-sched -fmodulo-sched-allow-regmoves -freorder-blocks-and-partition -ftree-parallelize-loops=2 -fno-inline -fno-move-loop-invariants -fno-sched-pressure -c l8jqrtst.f90 during RTL pass: sms f951: internal compiler error: in cfg_layout_redirect_edge_and_branch_force, at cfgrtl.c:4482 0x5a6108 cfg_layout_redirect_edge_and_branch_force /var/tmp/portage/cross-powerpc-e300c3-linux-gnu/gcc-9.0.0_alpha20190127/work/gcc-9-20190127/gcc/cfgrtl.c:4482 0x94c145 redirect_edge_and_branch_force(edge_def*, basic_block_def*) /var/tmp/portage/cross-powerpc-e300c3-linux-gnu/gcc-9.0.0_alpha20190127/work/gcc-9-20190127/gcc/cfghooks.c:486 0x95fb4e cfg_layout_split_edge /var/tmp/portage/cross-powerpc-e300c3-linux-gnu/gcc-9.0.0_alpha20190127/work/gcc-9-20190127/gcc/cfgrtl.c:4742 0x94c46f split_edge(edge_def*) /var/tmp/portage/cross-powerpc-e300c3-linux-gnu/gcc-9.0.0_alpha20190127/work/gcc-9-20190127/gcc/cfghooks.c:648 0xc0aa22 split_edge_and_insert(edge_def*, rtx_insn*) /var/tmp/portage/cross-powerpc-e300c3-linux-gnu/gcc-9.0.0_alpha20190127/work/gcc-9-20190127/gcc/loop-unroll.c:759 0x156ac09 generate_prolog_epilog /var/tmp/portage/cross-powerpc-e300c3-linux-gnu/gcc-9.0.0_alpha20190127/work/gcc-9-20190127/gcc/modulo-sched.c:1159 0x156ac09 sms_schedule /var/tmp/portage/cross-powerpc-e300c3-linux-gnu/gcc-9.0.0_alpha20190127/work/gcc-9-20190127/gcc/modulo-sched.c:1739 0x156b61f execute /var/tmp/portage/cross-powerpc-e300c3-linux-gnu/gcc-9.0.0_alpha20190127/work/gcc-9-20190127/gcc/modulo-sched.c:3337 In fact it can be a duplicate of PR85408 or PR87329 both of which I cannot reproduce anymore, or PR85426 which probably tracks two different issues already.
Started with r254832. Likely latent before.
Yeah, this is the same bug as PR85408. I haven't heard from Honza any comments on https://gcc.gnu.org/bugzilla/show_bug.cgi?id=85408#c2 Honza?
GCC 8.3 has been released.
GCC 8.4.0 has been released, adjusting target milestone.
PR84508 does not seem to be reproducible on trunk now as last commented by Arseny on trunk. Not sure about backporting a fix but does not seem to happen on trunk for now.
(In reply to Nicholas Krause from comment #5) > PR84508 does not seem to be reproducible on trunk now as last commented by > Arseny on trunk. Not sure about backporting a fix but does not seem to > happen on trunk for now. PR85408, you mean. Sure, but why post it here? It is PR89116, not PR85408. The latter is tracked in PR90040 by the modulo scheduler maintainer, and he's perfectly aware of that. But one generally cannot simply close random PRs they seemingly cannot reproduce anymore (and it seems you even didn't try to reproduce it yourself), well-established GCC development process doesn't work that way, and the whole issue w/ the modulo scheduler is not so trivial anyway, so all those PRs are there for a reason.
(In reply to Arseny Solokha from comment #6) > (In reply to Nicholas Krause from comment #5) > > PR84508 does not seem to be reproducible on trunk now as last commented by > > Arseny on trunk. Not sure about backporting a fix but does not seem to > > happen on trunk for now. > > PR85408, you mean. Sure, but why post it here? It is PR89116, not PR85408. > The latter is tracked in PR90040 by the modulo scheduler maintainer, and > he's perfectly aware of that. But one generally cannot simply close random > PRs they seemingly cannot reproduce anymore (and it seems you even didn't > try to reproduce it yourself), well-established GCC development process > doesn't work that way, and the whole issue w/ the modulo scheduler is not so > trivial anyway, so all those PRs are there for a reason. Sorry I see it now that this does block 90040 so it makes more sense now. My mistake as I thought this was just a duplicate and not a set of bugs.
GCC 8 branch is being closed.
GCC 9.4 is being released, retargeting bugs to GCC 9.5.
GCC 9 branch is being closed
GCC 10.4 is being released, retargeting bugs to GCC 10.5.
GCC 10 branch is being closed.