https://gcc.gnu.org/bugzilla/show_bug.cgi?id=88283
Richard Biener <rguenth at gcc dot gnu.org> changed: What |Removed |Added ---------------------------------------------------------------------------- Status|UNCONFIRMED |NEW Last reconfirmed| |2018-12-03 Ever confirmed|0 |1 --- Comment #1 from Richard Biener <rguenth at gcc dot gnu.org> --- #1 0x000000000121f580 in try_improve_iv_set (data=0x7fffffffd820, ivs=0x28a23d0, try_replace_p=0x7fffffffd6f7) at /tmp/trunk/gcc/tree-ssa-loop-ivopts.c:6677 6677 gcc_assert (best_cost == iv_ca_cost (ivs)); (gdb) p best_cost $1 = {cost = 12, complexity = 0, scratch = 0} (gdb) p iv_ca_cost (ivs) $2 = {cost = 10000000, complexity = 10000000, scratch = 10000000} Thus confirmed. Only graphite related as to the input IL to the IVOPTs pass. I suspect we run into some --param and do not expect backing out with maxmimum cost here.