https://gcc.gnu.org/bugzilla/show_bug.cgi?id=79916
--- Comment #9 from Vladimir Makarov <vmakarov at gcc dot gnu.org> --- I've reproduced one test bug on my machine by using: ./cc1 -I. ../../gcc/gcc/testsuite/gcc.dg/dfp/pr41049.c -fno-expensive-optimizations --param ira-max-conflict-table-size=0 -mpopcntd -O3 I think the fix can be probably in rs6000 machine-dependent code related to LRA and fixing it needs to know LRA very well. So I'll be working on it.
