[Bug target/82975] [6/7/8 Regression] ICE in baseness at rtlanal.c:6220
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=82975 --- Comment #6 from ktkachov at gcc dot gnu.org --- Author: ktkachov Date: Tue Dec 19 16:58:22 2017 New Revision: 255830 URL: https://gcc.gnu.org/viewcvs?rev=255830=gcc=rev Log: [arm] PR target/82975: Guard against reg_renumber being NULL in arm.h In this bug we ICE when checking REGNO_OK_FOR_INDEX_P on arm during pre-IRA scheduling. This is because REGNO_OK_FOR_INDEX_P ends up checking the reg_renumber array. Before IRA reg_renumber is NULL and thus we segfault. The fix is to guard the use of reg_renumber in the logic in TEST_REGNO in arm.h. On aarch64, for example, we also guard against the reg_renumber == NULL case. This fixes the ICE. I also remove the part of the comment that muses on when reg_renumber is available as with this patch it should now be safe to use at any point. Bootstrapped and tested on arm-none-linux-gnueabihf. PR target/82975 * config/arm/arm.h (TEST_REGNO): Check reg_renumber is set before accessing it. Adjust comment. * gcc.dg/pr82975.c: New test. Added: trunk/gcc/testsuite/gcc.dg/pr82975.c Modified: trunk/gcc/ChangeLog trunk/gcc/config/arm/arm.h trunk/gcc/testsuite/ChangeLog
[Bug target/82975] [6/7/8 Regression] ICE in baseness at rtlanal.c:6220
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=82975 ktkachov at gcc dot gnu.org changed: What|Removed |Added Status|NEW |ASSIGNED Assignee|unassigned at gcc dot gnu.org |ktkachov at gcc dot gnu.org --- Comment #5 from ktkachov at gcc dot gnu.org --- This is indeed an arm target issue. As part of the computation in rtlanal.c it uses REGNO_OK_FOR_INDEX_P which for arm eventually expands to something referencing reg_renumber. But during sched1 (where we ICE) we haven't run IRA yet to setup reg_renumber, so it's NULL at that point, causing the segfault. We should guard against that case in the definition of REGNO_OK_FOR_INDEX_P and REGNO_MODE_OK_FOR_BASE_P. The aarch64 port, for example, bails out when reg_renumber is not set. Testing a simple patch.
[Bug target/82975] [6/7/8 Regression] ICE in baseness at rtlanal.c:6220
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=82975 ktkachov at gcc dot gnu.org changed: What|Removed |Added Priority|P3 |P2 Last reconfirmed|2017-11-23 00:00:00 |2017-12-18 CC||ktkachov at gcc dot gnu.org Known to work||4.9.4 Summary|ICE in baseness at |[6/7/8 Regression] ICE in |rtlanal.c:6220 |baseness at rtlanal.c:6220 Known to fail||5.4.1, 6.4.1, 7.2.1, 8.0 --- Comment #4 from ktkachov at gcc dot gnu.org --- I've reproduced this ICE with the options Martin provided on arm-none-eabi cross compilers from GCC 5 onwards. 4.9 doesn't ICE for me. I think 4.9 didn't have the autopref scheduling stuff. It's not clear if this is an rtl-optimization or a target issue, but needs to be looked at anyway.