https://gcc.gnu.org/bugzilla/show_bug.cgi?id=93934
--- Comment #16 from joseph at codesourcery dot com <joseph at codesourcery dot com> --- I don't think this bug is anything to do with -fsignaling-nans, for the same reason as applies to bug 58416 and bug 71460. The option -fsignaling-nans is only about correctly handling programs that, in the abstract machine, interpret the bit pattern of a signaling NaN as a floating-point value (i.e. do lvalue-to-rvalue conversion on it in the relevant floating-point type). Thus bug, and those other two, and maybe others, involves a different case: a bit pattern of a signaling NaN occurs somewhere, but is never converted to an rvalue in the abstract machine. In such cases, the bit pattern should be handled correctly even without -fsignaling-nans (and, thus, on x87, it's incorrect to do single-precision or double-precision loads on any bit-pattern where an lvalue-to-rvalue conversion doesn't occur in the abstract machine, even without -fsignaling-nans).
