https://gcc.gnu.org/bugzilla/show_bug.cgi?id=88134
--- Comment #8 from Arseny Solokha <asolokha at gmx dot com> --- Aha. Segher, you've helpfully reminded me about -msoft-float w/ your recent commit. So it's not (solely) my misconfiguration once again: https://gcc.godbolt.org/z/sshcIF (it's g++, because gcc for powerpc seems to be broken there; -O1 -ftree-pre instead of -O2, because it's the pass that actually fails; and 1.0 instead of __builtin_huge_val (), because value the second operand doesn't really seem to matter).
