https://gcc.gnu.org/bugzilla/show_bug.cgi?id=115458
--- Comment #9 from Robin Dapp <rdapp at gcc dot gnu.org> --- Bisecting further only leads to the commit that introduced the vector ABI. Comparing the dumps with and without vector ABI is very tedious because a lot of things differ. It looks like we cannot create a reload for a masked sub vsub v8,v16,v24,v0.t (which already uses all available LMUL=8 hard regs). This insn is being inlined and disabling inlining gets rid of the ICE. I don't yet understand how the ABI changes spilling possibilities here.